1 /* SPDX-License-Identifier: GPL-2.0+ */
2 /*
3 * the_nilfs shared structure.
4 *
5 * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
6 *
7 * Written by Ryusuke Konishi.
8 *
9 */
10
11 #ifndef _THE_NILFS_H
12 #define _THE_NILFS_H
13
14 #include <linux/types.h>
15 #include <linux/buffer_head.h>
16 #include <linux/rbtree.h>
17 #include <linux/fs.h>
18 #include <linux/blkdev.h>
19 #include <linux/backing-dev.h>
20 #include <linux/slab.h>
21 #include <linux/refcount.h>
22
23 struct nilfs_sc_info;
24 struct nilfs_sysfs_dev_subgroups;
25
26 /* the_nilfs struct */
27 enum {
28 THE_NILFS_INIT = 0, /* Information from super_block is set */
29 THE_NILFS_DISCONTINUED, /* 'next' pointer chain has broken */
30 THE_NILFS_GC_RUNNING, /* gc process is running */
31 THE_NILFS_SB_DIRTY, /* super block is dirty */
32 THE_NILFS_PURGING, /* disposing dirty files for cleanup */
33 };
34
35 /**
36 * struct the_nilfs - struct to supervise multiple nilfs mount points
37 * @ns_flags: flags
38 * @ns_flushed_device: flag indicating if all volatile data was flushed
39 * @ns_sb: back pointer to super block instance
40 * @ns_bdev: block device
41 * @ns_sem: semaphore for shared states
42 * @ns_snapshot_mount_mutex: mutex to protect snapshot mounts
43 * @ns_sbh: buffer heads of on-disk super blocks
44 * @ns_sbp: pointers to super block data
45 * @ns_sbwtime: previous write time of super block
46 * @ns_sbwcount: write count of super block
47 * @ns_sbsize: size of valid data in super block
48 * @ns_mount_state: file system state
49 * @ns_sb_update_freq: interval of periodical update of superblocks (in seconds)
50 * @ns_seg_seq: segment sequence counter
51 * @ns_segnum: index number of the latest full segment.
52 * @ns_nextnum: index number of the full segment index to be used next
53 * @ns_pseg_offset: offset of next partial segment in the current full segment
54 * @ns_cno: next checkpoint number
55 * @ns_ctime: write time of the last segment
56 * @ns_nongc_ctime: write time of the last segment not for cleaner operation
57 * @ns_ndirtyblks: Number of dirty data blocks
58 * @ns_last_segment_lock: lock protecting fields for the latest segment
59 * @ns_last_pseg: start block number of the latest segment
60 * @ns_last_seq: sequence value of the latest segment
61 * @ns_last_cno: checkpoint number of the latest segment
62 * @ns_prot_seq: least sequence number of segments which must not be reclaimed
63 * @ns_prev_seq: base sequence number used to decide if advance log cursor
64 * @ns_writer: log writer
65 * @ns_segctor_sem: semaphore protecting log write
66 * @ns_dat: DAT file inode
67 * @ns_cpfile: checkpoint file inode
68 * @ns_sufile: segusage file inode
69 * @ns_cptree: rb-tree of all mounted checkpoints (nilfs_root)
70 * @ns_cptree_lock: lock protecting @ns_cptree
71 * @ns_dirty_files: list of dirty files
72 * @ns_inode_lock: lock protecting @ns_dirty_files
73 * @ns_gc_inodes: dummy inodes to keep live blocks
74 * @ns_next_generation: next generation number for inodes
75 * @ns_next_gen_lock: lock protecting @ns_next_generation
76 * @ns_mount_opt: mount options
77 * @ns_resuid: uid for reserved blocks
78 * @ns_resgid: gid for reserved blocks
79 * @ns_interval: checkpoint creation interval
80 * @ns_watermark: watermark for the number of dirty buffers
81 * @ns_blocksize_bits: bit length of block size
82 * @ns_blocksize: block size
83 * @ns_nsegments: number of segments in filesystem
84 * @ns_blocks_per_segment: number of blocks per segment
85 * @ns_r_segments_percentage: reserved segments percentage
86 * @ns_nrsvsegs: number of reserved segments
87 * @ns_first_data_block: block number of first data block
88 * @ns_inode_size: size of on-disk inode
89 * @ns_first_ino: first not-special inode number
90 * @ns_crc_seed: seed value of CRC32 calculation
91 * @ns_dev_kobj: /sys/fs/<nilfs>/<device>
92 * @ns_dev_kobj_unregister: completion state
93 * @ns_dev_subgroups: <device> subgroups pointer
94 */
95 struct the_nilfs {
96 unsigned long ns_flags;
97 int ns_flushed_device;
98
99 struct super_block *ns_sb;
100 struct block_device *ns_bdev;
101 struct rw_semaphore ns_sem;
102 struct mutex ns_snapshot_mount_mutex;
103
104 /*
105 * used for
106 * - loading the latest checkpoint exclusively.
107 * - allocating a new full segment.
108 */
109 struct buffer_head *ns_sbh[2];
110 struct nilfs_super_block *ns_sbp[2];
111 time64_t ns_sbwtime;
112 unsigned int ns_sbwcount;
113 unsigned int ns_sbsize;
114 unsigned int ns_mount_state;
115 unsigned int ns_sb_update_freq;
116
117 /*
118 * The following fields are updated by a writable FS-instance.
119 * These fields are protected by ns_segctor_sem outside load_nilfs().
120 */
121 u64 ns_seg_seq;
122 __u64 ns_segnum;
123 __u64 ns_nextnum;
124 unsigned long ns_pseg_offset;
125 __u64 ns_cno;
126 time64_t ns_ctime;
127 time64_t ns_nongc_ctime;
128 atomic_t ns_ndirtyblks;
129
130 /*
131 * The following fields hold information on the latest partial segment
132 * written to disk with a super root. These fields are protected by
133 * ns_last_segment_lock.
134 */
135 spinlock_t ns_last_segment_lock;
136 sector_t ns_last_pseg;
137 u64 ns_last_seq;
138 __u64 ns_last_cno;
139 u64 ns_prot_seq;
140 u64 ns_prev_seq;
141
142 struct nilfs_sc_info *ns_writer;
143 struct rw_semaphore ns_segctor_sem;
144
145 /*
146 * Following fields are lock free except for the period before
147 * the_nilfs is initialized.
148 */
149 struct inode *ns_dat;
150 struct inode *ns_cpfile;
151 struct inode *ns_sufile;
152
153 /* Checkpoint tree */
154 struct rb_root ns_cptree;
155 spinlock_t ns_cptree_lock;
156
157 /* Dirty inode list */
158 struct list_head ns_dirty_files;
159 spinlock_t ns_inode_lock;
160
161 /* GC inode list */
162 struct list_head ns_gc_inodes;
163
164 /* Inode allocator */
165 u32 ns_next_generation;
166 spinlock_t ns_next_gen_lock;
167
168 /* Mount options */
169 unsigned long ns_mount_opt;
170
171 uid_t ns_resuid;
172 gid_t ns_resgid;
173 unsigned long ns_interval;
174 unsigned long ns_watermark;
175
176 /* Disk layout information (static) */
177 unsigned int ns_blocksize_bits;
178 unsigned int ns_blocksize;
179 unsigned long ns_nsegments;
180 unsigned long ns_blocks_per_segment;
181 unsigned long ns_r_segments_percentage;
182 unsigned long ns_nrsvsegs;
183 unsigned long ns_first_data_block;
184 int ns_inode_size;
185 unsigned int ns_first_ino;
186 u32 ns_crc_seed;
187
188 /* /sys/fs/<nilfs>/<device> */
189 struct kobject ns_dev_kobj;
190 struct completion ns_dev_kobj_unregister;
191 struct nilfs_sysfs_dev_subgroups *ns_dev_subgroups;
192 };
193
194 #define THE_NILFS_FNS(bit, name) \
195 static inline void set_nilfs_##name(struct the_nilfs *nilfs) \
196 { \
197 set_bit(THE_NILFS_##bit, &(nilfs)->ns_flags); \
198 } \
199 static inline void clear_nilfs_##name(struct the_nilfs *nilfs) \
200 { \
201 clear_bit(THE_NILFS_##bit, &(nilfs)->ns_flags); \
202 } \
203 static inline int nilfs_##name(struct the_nilfs *nilfs) \
204 { \
205 return test_bit(THE_NILFS_##bit, &(nilfs)->ns_flags); \
206 }
207
208 THE_NILFS_FNS(INIT, init)
209 THE_NILFS_FNS(DISCONTINUED, discontinued)
210 THE_NILFS_FNS(GC_RUNNING, gc_running)
211 THE_NILFS_FNS(SB_DIRTY, sb_dirty)
212 THE_NILFS_FNS(PURGING, purging)
213
214 /*
215 * Mount option operations
216 */
217 #define nilfs_clear_opt(nilfs, opt) \
218 ((nilfs)->ns_mount_opt &= ~NILFS_MOUNT_##opt)
219 #define nilfs_set_opt(nilfs, opt) \
220 ((nilfs)->ns_mount_opt |= NILFS_MOUNT_##opt)
221 #define nilfs_test_opt(nilfs, opt) ((nilfs)->ns_mount_opt & NILFS_MOUNT_##opt)
222
223 /**
224 * struct nilfs_root - nilfs root object
225 * @cno: checkpoint number
226 * @rb_node: red-black tree node
227 * @count: refcount of this structure
228 * @nilfs: nilfs object
229 * @ifile: inode file
230 * @inodes_count: number of inodes
231 * @blocks_count: number of blocks
232 * @snapshot_kobj: /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot>
233 * @snapshot_kobj_unregister: completion state for kernel object
234 */
235 struct nilfs_root {
236 __u64 cno;
237 struct rb_node rb_node;
238
239 refcount_t count;
240 struct the_nilfs *nilfs;
241 struct inode *ifile;
242
243 atomic64_t inodes_count;
244 atomic64_t blocks_count;
245
246 /* /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot> */
247 struct kobject snapshot_kobj;
248 struct completion snapshot_kobj_unregister;
249 };
250
251 /* Special checkpoint number */
252 #define NILFS_CPTREE_CURRENT_CNO 0
253
254 /* Minimum interval of periodical update of superblocks (in seconds) */
255 #define NILFS_SB_FREQ 10
256
257 static inline int nilfs_sb_need_update(struct the_nilfs *nilfs)
258 {
259 u64 t = ktime_get_real_seconds();
260
261 return t < nilfs->ns_sbwtime ||
262 t > nilfs->ns_sbwtime + nilfs->ns_sb_update_freq;
263 }
264
265 static inline int nilfs_sb_will_flip(struct the_nilfs *nilfs)
266 {
267 int flip_bits = nilfs->ns_sbwcount & 0x0FL;
268
269 return (flip_bits != 0x08 && flip_bits != 0x0F);
270 }
271
272 void nilfs_set_last_segment(struct the_nilfs *, sector_t, u64, __u64);
273 struct the_nilfs *alloc_nilfs(struct super_block *sb);
274 void destroy_nilfs(struct the_nilfs *nilfs);
275 int init_nilfs(struct the_nilfs *nilfs, struct super_block *sb);
276 int load_nilfs(struct the_nilfs *nilfs, struct super_block *sb);
277 unsigned long nilfs_nrsvsegs(struct the_nilfs *nilfs, unsigned long nsegs);
278 void nilfs_set_nsegments(struct the_nilfs *nilfs, unsigned long nsegs);
279 int nilfs_discard_segments(struct the_nilfs *, __u64 *, size_t);
280 int nilfs_count_free_blocks(struct the_nilfs *, sector_t *);
281 struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno);
282 struct nilfs_root *nilfs_find_or_create_root(struct the_nilfs *nilfs,
283 __u64 cno);
284 void nilfs_put_root(struct nilfs_root *root);
285 int nilfs_near_disk_full(struct the_nilfs *);
286 void nilfs_fall_back_super_block(struct the_nilfs *);
287 void nilfs_swap_super_block(struct the_nilfs *);
288
289
290 static inline void nilfs_get_root(struct nilfs_root *root)
291 {
292 refcount_inc(&root->count);
293 }
294
295 static inline int nilfs_valid_fs(struct the_nilfs *nilfs)
296 {
297 unsigned int valid_fs;
298
299 down_read(&nilfs->ns_sem);
300 valid_fs = (nilfs->ns_mount_state & NILFS_VALID_FS);
301 up_read(&nilfs->ns_sem);
302 return valid_fs;
303 }
304
305 static inline void
306 nilfs_get_segment_range(struct the_nilfs *nilfs, __u64 segnum,
307 sector_t *seg_start, sector_t *seg_end)
308 {
309 *seg_start = (sector_t)nilfs->ns_blocks_per_segment * segnum;
310 *seg_end = *seg_start + nilfs->ns_blocks_per_segment - 1;
311 if (segnum == 0)
312 *seg_start = nilfs->ns_first_data_block;
313 }
314
315 static inline sector_t
316 nilfs_get_segment_start_blocknr(struct the_nilfs *nilfs, __u64 segnum)
317 {
318 return (segnum == 0) ? nilfs->ns_first_data_block :
319 (sector_t)nilfs->ns_blocks_per_segment * segnum;
320 }
321
322 static inline __u64
323 nilfs_get_segnum_of_block(struct the_nilfs *nilfs, sector_t blocknr)
324 {
325 sector_t segnum = blocknr;
326
327 sector_div(segnum, nilfs->ns_blocks_per_segment);
328 return segnum;
329 }
330
331 static inline void
332 nilfs_terminate_segment(struct the_nilfs *nilfs, sector_t seg_start,
333 sector_t seg_end)
334 {
335 /* terminate the current full segment (used in case of I/O-error) */
336 nilfs->ns_pseg_offset = seg_end - seg_start + 1;
337 }
338
339 static inline void nilfs_shift_to_next_segment(struct the_nilfs *nilfs)
340 {
341 /* move forward with a full segment */
342 nilfs->ns_segnum = nilfs->ns_nextnum;
343 nilfs->ns_pseg_offset = 0;
344 nilfs->ns_seg_seq++;
345 }
346
347 static inline __u64 nilfs_last_cno(struct the_nilfs *nilfs)
348 {
349 __u64 cno;
350
351 spin_lock(&nilfs->ns_last_segment_lock);
352 cno = nilfs->ns_last_cno;
353 spin_unlock(&nilfs->ns_last_segment_lock);
354 return cno;
355 }
356
357 static inline int nilfs_segment_is_active(struct the_nilfs *nilfs, __u64 n)
358 {
359 return n == nilfs->ns_segnum || n == nilfs->ns_nextnum;
360 }
361
362 static inline int nilfs_flush_device(struct the_nilfs *nilfs)
363 {
364 int err;
365
366 if (!nilfs_test_opt(nilfs, BARRIER) || nilfs->ns_flushed_device)
367 return 0;
368
369 nilfs->ns_flushed_device = 1;
370 /*
371 * the store to ns_flushed_device must not be reordered after
372 * blkdev_issue_flush().
373 */
374 smp_wmb();
375
376 err = blkdev_issue_flush(nilfs->ns_bdev);
377 if (err != -EIO)
378 err = 0;
379 return err;
380 }
381
382 #endif /* _THE_NILFS_H */
383
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.