2 * gcinode.c - dummy inodes to buffer blocks for garbage collection
4 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
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.
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.
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
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>.
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.
31 * Since NILFS2 keeps up multiple checkpoints/snapshots across 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 separately 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.
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.
45 #include <linux/buffer_head.h>
46 #include <linux/mpage.h>
47 #include <linux/hash.h>
48 #include <linux/slab.h>
49 #include <linux/swap.h>
58 static const struct address_space_operations def_gcinode_aops
= {
59 .sync_page
= block_sync_page
,
63 * nilfs_gccache_submit_read_data() - add data buffer and submit read request
65 * @blkoff - dummy offset treated as the key for the page cache
66 * @pbn - physical block number of the block
67 * @vbn - virtual block number of the block, 0 for non-virtual block
68 * @out_bh - indirect pointer to a buffer_head struct to receive the results
70 * Description: nilfs_gccache_submit_read_data() registers the data buffer
71 * specified by @pbn to the GC pagecache with the key @blkoff.
72 * This function sets @vbn (@pbn if @vbn is zero) in b_blocknr of the buffer.
74 * Return Value: On success, 0 is returned. On Error, one of the following
75 * negative error code is returned.
79 * %-ENOMEM - Insufficient amount of memory available.
81 * %-ENOENT - The block specified with @pbn does not exist.
83 int nilfs_gccache_submit_read_data(struct inode
*inode
, sector_t blkoff
,
84 sector_t pbn
, __u64 vbn
,
85 struct buffer_head
**out_bh
)
87 struct buffer_head
*bh
;
90 bh
= nilfs_grab_buffer(inode
, inode
->i_mapping
, blkoff
, 0);
94 if (buffer_uptodate(bh
))
98 struct inode
*dat_inode
= NILFS_I_NILFS(inode
)->ns_dat
;
99 /* use original dat, not gc dat. */
100 err
= nilfs_dat_translate(dat_inode
, vbn
, &pbn
);
101 if (unlikely(err
)) { /* -EIO, -ENOMEM, -ENOENT */
108 if (buffer_uptodate(bh
)) {
113 if (!buffer_mapped(bh
)) {
114 bh
->b_bdev
= NILFS_I_NILFS(inode
)->ns_bdev
;
115 set_buffer_mapped(bh
);
118 bh
->b_end_io
= end_buffer_read_sync
;
128 unlock_page(bh
->b_page
);
129 page_cache_release(bh
->b_page
);
134 * nilfs_gccache_submit_read_node() - add node buffer and submit read request
136 * @pbn - physical block number for the block
137 * @vbn - virtual block number for the block
138 * @out_bh - indirect pointer to a buffer_head struct to receive the results
140 * Description: nilfs_gccache_submit_read_node() registers the node buffer
141 * specified by @vbn to the GC pagecache. @pbn can be supplied by the
142 * caller to avoid translation of the disk block address.
144 * Return Value: On success, 0 is returned. On Error, one of the following
145 * negative error code is returned.
149 * %-ENOMEM - Insufficient amount of memory available.
151 int nilfs_gccache_submit_read_node(struct inode
*inode
, sector_t pbn
,
152 __u64 vbn
, struct buffer_head
**out_bh
)
156 ret
= nilfs_btnode_submit_block(&NILFS_I(inode
)->i_btnode_cache
,
157 vbn
? : pbn
, pbn
, READ
, out_bh
, &pbn
);
158 if (ret
== -EEXIST
) /* internal code (cache hit) */
163 int nilfs_gccache_wait_and_mark_dirty(struct buffer_head
*bh
)
166 if (!buffer_uptodate(bh
))
168 if (buffer_dirty(bh
))
171 if (buffer_nilfs_node(bh
)) {
172 if (nilfs_btree_broken_node_block(bh
)) {
173 clear_buffer_uptodate(bh
);
176 nilfs_btnode_mark_dirty(bh
);
178 nilfs_mdt_mark_buffer_dirty(bh
);
184 * nilfs_init_gccache() - allocate and initialize gc_inode hash table
187 * Return Value: On success, 0.
188 * On error, a negative error code is returned.
190 int nilfs_init_gccache(struct the_nilfs
*nilfs
)
194 BUG_ON(nilfs
->ns_gc_inodes_h
);
196 INIT_LIST_HEAD(&nilfs
->ns_gc_inodes
);
198 nilfs
->ns_gc_inodes_h
=
199 kmalloc(sizeof(struct hlist_head
) * NILFS_GCINODE_HASH_SIZE
,
201 if (nilfs
->ns_gc_inodes_h
== NULL
)
204 for (loop
= 0; loop
< NILFS_GCINODE_HASH_SIZE
; loop
++)
205 INIT_HLIST_HEAD(&nilfs
->ns_gc_inodes_h
[loop
]);
210 * nilfs_destroy_gccache() - free gc_inode hash table
213 void nilfs_destroy_gccache(struct the_nilfs
*nilfs
)
215 if (nilfs
->ns_gc_inodes_h
) {
216 nilfs_remove_all_gcinode(nilfs
);
217 kfree(nilfs
->ns_gc_inodes_h
);
218 nilfs
->ns_gc_inodes_h
= NULL
;
222 static struct inode
*alloc_gcinode(struct the_nilfs
*nilfs
, ino_t ino
,
226 struct nilfs_inode_info
*ii
;
228 inode
= nilfs_mdt_new_common(nilfs
, NULL
, ino
, GFP_NOFS
, 0);
234 inode
->i_mapping
->a_ops
= &def_gcinode_aops
;
239 ii
->i_state
= 1 << NILFS_I_GCINODE
;
241 nilfs_bmap_init_gc(ii
->i_bmap
);
246 static unsigned long ihash(ino_t ino
, __u64 cno
)
248 return hash_long((unsigned long)((ino
<< 2) + cno
),
249 NILFS_GCINODE_HASH_BITS
);
253 * nilfs_gc_iget() - find or create gc inode with specified (ino,cno)
255 struct inode
*nilfs_gc_iget(struct the_nilfs
*nilfs
, ino_t ino
, __u64 cno
)
257 struct hlist_head
*head
= nilfs
->ns_gc_inodes_h
+ ihash(ino
, cno
);
258 struct hlist_node
*node
;
261 hlist_for_each_entry(inode
, node
, head
, i_hash
) {
262 if (inode
->i_ino
== ino
&& NILFS_I(inode
)->i_cno
== cno
)
266 inode
= alloc_gcinode(nilfs
, ino
, cno
);
268 hlist_add_head(&inode
->i_hash
, head
);
269 list_add(&NILFS_I(inode
)->i_dirty
, &nilfs
->ns_gc_inodes
);
275 * nilfs_clear_gcinode() - clear and free a gc inode
277 void nilfs_clear_gcinode(struct inode
*inode
)
279 nilfs_mdt_destroy(inode
);
283 * nilfs_remove_all_gcinode() - remove all inodes from the_nilfs
285 void nilfs_remove_all_gcinode(struct the_nilfs
*nilfs
)
287 struct hlist_head
*head
= nilfs
->ns_gc_inodes_h
;
288 struct hlist_node
*node
, *n
;
292 for (loop
= 0; loop
< NILFS_GCINODE_HASH_SIZE
; loop
++, head
++) {
293 hlist_for_each_entry_safe(inode
, node
, n
, head
, i_hash
) {
294 hlist_del_init(&inode
->i_hash
);
295 list_del_init(&NILFS_I(inode
)->i_dirty
);
296 nilfs_clear_gcinode(inode
); /* might sleep */