xref: /linux/fs/nilfs2/the_nilfs.h (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
1 /*
2  * the_nilfs.h - the_nilfs shared structure.
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  * Written by Ryusuke Konishi.
17  *
18  */
19 
20 #ifndef _THE_NILFS_H
21 #define _THE_NILFS_H
22 
23 #include <linux/types.h>
24 #include <linux/buffer_head.h>
25 #include <linux/rbtree.h>
26 #include <linux/fs.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/slab.h>
30 
31 struct nilfs_sc_info;
32 struct nilfs_sysfs_dev_subgroups;
33 
34 /* the_nilfs struct */
35 enum {
36 	THE_NILFS_INIT = 0,     /* Information from super_block is set */
37 	THE_NILFS_DISCONTINUED,	/* 'next' pointer chain has broken */
38 	THE_NILFS_GC_RUNNING,	/* gc process is running */
39 	THE_NILFS_SB_DIRTY,	/* super block is dirty */
40 };
41 
42 /**
43  * struct the_nilfs - struct to supervise multiple nilfs mount points
44  * @ns_flags: flags
45  * @ns_flushed_device: flag indicating if all volatile data was flushed
46  * @ns_sb: back pointer to super block instance
47  * @ns_bdev: block device
48  * @ns_sem: semaphore for shared states
49  * @ns_snapshot_mount_mutex: mutex to protect snapshot mounts
50  * @ns_sbh: buffer heads of on-disk super blocks
51  * @ns_sbp: pointers to super block data
52  * @ns_sbwtime: previous write time of super block
53  * @ns_sbwcount: write count of super block
54  * @ns_sbsize: size of valid data in super block
55  * @ns_mount_state: file system state
56  * @ns_sb_update_freq: interval of periodical update of superblocks (in seconds)
57  * @ns_seg_seq: segment sequence counter
58  * @ns_segnum: index number of the latest full segment.
59  * @ns_nextnum: index number of the full segment index to be used next
60  * @ns_pseg_offset: offset of next partial segment in the current full segment
61  * @ns_cno: next checkpoint number
62  * @ns_ctime: write time of the last segment
63  * @ns_nongc_ctime: write time of the last segment not for cleaner operation
64  * @ns_ndirtyblks: Number of dirty data blocks
65  * @ns_last_segment_lock: lock protecting fields for the latest segment
66  * @ns_last_pseg: start block number of the latest segment
67  * @ns_last_seq: sequence value of the latest segment
68  * @ns_last_cno: checkpoint number of the latest segment
69  * @ns_prot_seq: least sequence number of segments which must not be reclaimed
70  * @ns_prev_seq: base sequence number used to decide if advance log cursor
71  * @ns_writer: log writer
72  * @ns_segctor_sem: semaphore protecting log write
73  * @ns_dat: DAT file inode
74  * @ns_cpfile: checkpoint file inode
75  * @ns_sufile: segusage file inode
76  * @ns_cptree: rb-tree of all mounted checkpoints (nilfs_root)
77  * @ns_cptree_lock: lock protecting @ns_cptree
78  * @ns_dirty_files: list of dirty files
79  * @ns_inode_lock: lock protecting @ns_dirty_files
80  * @ns_gc_inodes: dummy inodes to keep live blocks
81  * @ns_next_generation: next generation number for inodes
82  * @ns_next_gen_lock: lock protecting @ns_next_generation
83  * @ns_mount_opt: mount options
84  * @ns_resuid: uid for reserved blocks
85  * @ns_resgid: gid for reserved blocks
86  * @ns_interval: checkpoint creation interval
87  * @ns_watermark: watermark for the number of dirty buffers
88  * @ns_blocksize_bits: bit length of block size
89  * @ns_blocksize: block size
90  * @ns_nsegments: number of segments in filesystem
91  * @ns_blocks_per_segment: number of blocks per segment
92  * @ns_r_segments_percentage: reserved segments percentage
93  * @ns_nrsvsegs: number of reserved segments
94  * @ns_first_data_block: block number of first data block
95  * @ns_inode_size: size of on-disk inode
96  * @ns_first_ino: first not-special inode number
97  * @ns_crc_seed: seed value of CRC32 calculation
98  * @ns_dev_kobj: /sys/fs/<nilfs>/<device>
99  * @ns_dev_kobj_unregister: completion state
100  * @ns_dev_subgroups: <device> subgroups pointer
101  */
102 struct the_nilfs {
103 	unsigned long		ns_flags;
104 	int			ns_flushed_device;
105 
106 	struct super_block     *ns_sb;
107 	struct block_device    *ns_bdev;
108 	struct rw_semaphore	ns_sem;
109 	struct mutex		ns_snapshot_mount_mutex;
110 
111 	/*
112 	 * used for
113 	 * - loading the latest checkpoint exclusively.
114 	 * - allocating a new full segment.
115 	 */
116 	struct buffer_head     *ns_sbh[2];
117 	struct nilfs_super_block *ns_sbp[2];
118 	time_t			ns_sbwtime;
119 	unsigned int		ns_sbwcount;
120 	unsigned int		ns_sbsize;
121 	unsigned int		ns_mount_state;
122 	unsigned int		ns_sb_update_freq;
123 
124 	/*
125 	 * The following fields are updated by a writable FS-instance.
126 	 * These fields are protected by ns_segctor_sem outside load_nilfs().
127 	 */
128 	u64			ns_seg_seq;
129 	__u64			ns_segnum;
130 	__u64			ns_nextnum;
131 	unsigned long		ns_pseg_offset;
132 	__u64			ns_cno;
133 	time_t			ns_ctime;
134 	time_t			ns_nongc_ctime;
135 	atomic_t		ns_ndirtyblks;
136 
137 	/*
138 	 * The following fields hold information on the latest partial segment
139 	 * written to disk with a super root.  These fields are protected by
140 	 * ns_last_segment_lock.
141 	 */
142 	spinlock_t		ns_last_segment_lock;
143 	sector_t		ns_last_pseg;
144 	u64			ns_last_seq;
145 	__u64			ns_last_cno;
146 	u64			ns_prot_seq;
147 	u64			ns_prev_seq;
148 
149 	struct nilfs_sc_info   *ns_writer;
150 	struct rw_semaphore	ns_segctor_sem;
151 
152 	/*
153 	 * Following fields are lock free except for the period before
154 	 * the_nilfs is initialized.
155 	 */
156 	struct inode	       *ns_dat;
157 	struct inode	       *ns_cpfile;
158 	struct inode	       *ns_sufile;
159 
160 	/* Checkpoint tree */
161 	struct rb_root		ns_cptree;
162 	spinlock_t		ns_cptree_lock;
163 
164 	/* Dirty inode list */
165 	struct list_head	ns_dirty_files;
166 	spinlock_t		ns_inode_lock;
167 
168 	/* GC inode list */
169 	struct list_head	ns_gc_inodes;
170 
171 	/* Inode allocator */
172 	u32			ns_next_generation;
173 	spinlock_t		ns_next_gen_lock;
174 
175 	/* Mount options */
176 	unsigned long		ns_mount_opt;
177 
178 	uid_t			ns_resuid;
179 	gid_t			ns_resgid;
180 	unsigned long		ns_interval;
181 	unsigned long		ns_watermark;
182 
183 	/* Disk layout information (static) */
184 	unsigned int		ns_blocksize_bits;
185 	unsigned int		ns_blocksize;
186 	unsigned long		ns_nsegments;
187 	unsigned long		ns_blocks_per_segment;
188 	unsigned long		ns_r_segments_percentage;
189 	unsigned long		ns_nrsvsegs;
190 	unsigned long		ns_first_data_block;
191 	int			ns_inode_size;
192 	int			ns_first_ino;
193 	u32			ns_crc_seed;
194 
195 	/* /sys/fs/<nilfs>/<device> */
196 	struct kobject ns_dev_kobj;
197 	struct completion ns_dev_kobj_unregister;
198 	struct nilfs_sysfs_dev_subgroups *ns_dev_subgroups;
199 };
200 
201 #define THE_NILFS_FNS(bit, name)					\
202 static inline void set_nilfs_##name(struct the_nilfs *nilfs)		\
203 {									\
204 	set_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);			\
205 }									\
206 static inline void clear_nilfs_##name(struct the_nilfs *nilfs)		\
207 {									\
208 	clear_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);			\
209 }									\
210 static inline int nilfs_##name(struct the_nilfs *nilfs)			\
211 {									\
212 	return test_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);		\
213 }
214 
215 THE_NILFS_FNS(INIT, init)
216 THE_NILFS_FNS(DISCONTINUED, discontinued)
217 THE_NILFS_FNS(GC_RUNNING, gc_running)
218 THE_NILFS_FNS(SB_DIRTY, sb_dirty)
219 
220 /*
221  * Mount option operations
222  */
223 #define nilfs_clear_opt(nilfs, opt)  \
224 	((nilfs)->ns_mount_opt &= ~NILFS_MOUNT_##opt)
225 #define nilfs_set_opt(nilfs, opt)  \
226 	((nilfs)->ns_mount_opt |= NILFS_MOUNT_##opt)
227 #define nilfs_test_opt(nilfs, opt) ((nilfs)->ns_mount_opt & NILFS_MOUNT_##opt)
228 #define nilfs_write_opt(nilfs, mask, opt)				\
229 	((nilfs)->ns_mount_opt =					\
230 		(((nilfs)->ns_mount_opt & ~NILFS_MOUNT_##mask) |	\
231 		 NILFS_MOUNT_##opt))					\
232 
233 /**
234  * struct nilfs_root - nilfs root object
235  * @cno: checkpoint number
236  * @rb_node: red-black tree node
237  * @count: refcount of this structure
238  * @nilfs: nilfs object
239  * @ifile: inode file
240  * @inodes_count: number of inodes
241  * @blocks_count: number of blocks
242  * @snapshot_kobj: /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot>
243  * @snapshot_kobj_unregister: completion state for kernel object
244  */
245 struct nilfs_root {
246 	__u64 cno;
247 	struct rb_node rb_node;
248 
249 	atomic_t count;
250 	struct the_nilfs *nilfs;
251 	struct inode *ifile;
252 
253 	atomic64_t inodes_count;
254 	atomic64_t blocks_count;
255 
256 	/* /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot> */
257 	struct kobject snapshot_kobj;
258 	struct completion snapshot_kobj_unregister;
259 };
260 
261 /* Special checkpoint number */
262 #define NILFS_CPTREE_CURRENT_CNO	0
263 
264 /* Minimum interval of periodical update of superblocks (in seconds) */
265 #define NILFS_SB_FREQ		10
266 
267 static inline int nilfs_sb_need_update(struct the_nilfs *nilfs)
268 {
269 	u64 t = get_seconds();
270 
271 	return t < nilfs->ns_sbwtime ||
272 		t > nilfs->ns_sbwtime + nilfs->ns_sb_update_freq;
273 }
274 
275 static inline int nilfs_sb_will_flip(struct the_nilfs *nilfs)
276 {
277 	int flip_bits = nilfs->ns_sbwcount & 0x0FL;
278 
279 	return (flip_bits != 0x08 && flip_bits != 0x0F);
280 }
281 
282 void nilfs_set_last_segment(struct the_nilfs *, sector_t, u64, __u64);
283 struct the_nilfs *alloc_nilfs(struct super_block *sb);
284 void destroy_nilfs(struct the_nilfs *nilfs);
285 int init_nilfs(struct the_nilfs *nilfs, struct super_block *sb, char *data);
286 int load_nilfs(struct the_nilfs *nilfs, struct super_block *sb);
287 unsigned long nilfs_nrsvsegs(struct the_nilfs *nilfs, unsigned long nsegs);
288 void nilfs_set_nsegments(struct the_nilfs *nilfs, unsigned long nsegs);
289 int nilfs_discard_segments(struct the_nilfs *, __u64 *, size_t);
290 int nilfs_count_free_blocks(struct the_nilfs *, sector_t *);
291 struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno);
292 struct nilfs_root *nilfs_find_or_create_root(struct the_nilfs *nilfs,
293 					     __u64 cno);
294 void nilfs_put_root(struct nilfs_root *root);
295 int nilfs_near_disk_full(struct the_nilfs *);
296 void nilfs_fall_back_super_block(struct the_nilfs *);
297 void nilfs_swap_super_block(struct the_nilfs *);
298 
299 
300 static inline void nilfs_get_root(struct nilfs_root *root)
301 {
302 	atomic_inc(&root->count);
303 }
304 
305 static inline int nilfs_valid_fs(struct the_nilfs *nilfs)
306 {
307 	unsigned int valid_fs;
308 
309 	down_read(&nilfs->ns_sem);
310 	valid_fs = (nilfs->ns_mount_state & NILFS_VALID_FS);
311 	up_read(&nilfs->ns_sem);
312 	return valid_fs;
313 }
314 
315 static inline void
316 nilfs_get_segment_range(struct the_nilfs *nilfs, __u64 segnum,
317 			sector_t *seg_start, sector_t *seg_end)
318 {
319 	*seg_start = (sector_t)nilfs->ns_blocks_per_segment * segnum;
320 	*seg_end = *seg_start + nilfs->ns_blocks_per_segment - 1;
321 	if (segnum == 0)
322 		*seg_start = nilfs->ns_first_data_block;
323 }
324 
325 static inline sector_t
326 nilfs_get_segment_start_blocknr(struct the_nilfs *nilfs, __u64 segnum)
327 {
328 	return (segnum == 0) ? nilfs->ns_first_data_block :
329 		(sector_t)nilfs->ns_blocks_per_segment * segnum;
330 }
331 
332 static inline __u64
333 nilfs_get_segnum_of_block(struct the_nilfs *nilfs, sector_t blocknr)
334 {
335 	sector_t segnum = blocknr;
336 
337 	sector_div(segnum, nilfs->ns_blocks_per_segment);
338 	return segnum;
339 }
340 
341 static inline void
342 nilfs_terminate_segment(struct the_nilfs *nilfs, sector_t seg_start,
343 			sector_t seg_end)
344 {
345 	/* terminate the current full segment (used in case of I/O-error) */
346 	nilfs->ns_pseg_offset = seg_end - seg_start + 1;
347 }
348 
349 static inline void nilfs_shift_to_next_segment(struct the_nilfs *nilfs)
350 {
351 	/* move forward with a full segment */
352 	nilfs->ns_segnum = nilfs->ns_nextnum;
353 	nilfs->ns_pseg_offset = 0;
354 	nilfs->ns_seg_seq++;
355 }
356 
357 static inline __u64 nilfs_last_cno(struct the_nilfs *nilfs)
358 {
359 	__u64 cno;
360 
361 	spin_lock(&nilfs->ns_last_segment_lock);
362 	cno = nilfs->ns_last_cno;
363 	spin_unlock(&nilfs->ns_last_segment_lock);
364 	return cno;
365 }
366 
367 static inline int nilfs_segment_is_active(struct the_nilfs *nilfs, __u64 n)
368 {
369 	return n == nilfs->ns_segnum || n == nilfs->ns_nextnum;
370 }
371 
372 static inline int nilfs_flush_device(struct the_nilfs *nilfs)
373 {
374 	int err;
375 
376 	if (!nilfs_test_opt(nilfs, BARRIER) || nilfs->ns_flushed_device)
377 		return 0;
378 
379 	nilfs->ns_flushed_device = 1;
380 	/*
381 	 * the store to ns_flushed_device must not be reordered after
382 	 * blkdev_issue_flush().
383 	 */
384 	smp_wmb();
385 
386 	err = blkdev_issue_flush(nilfs->ns_bdev, GFP_KERNEL, NULL);
387 	if (err != -EIO)
388 		err = 0;
389 	return err;
390 }
391 
392 #endif /* _THE_NILFS_H */
393