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 };
33 
34 /**
35  * struct the_nilfs - struct to supervise multiple nilfs mount points
36  * @ns_flags: flags
37  * @ns_flushed_device: flag indicating if all volatile data was flushed
38  * @ns_sb: back pointer to super block instance
39  * @ns_bdev: block device
40  * @ns_sem: semaphore for shared states
41  * @ns_snapshot_mount_mutex: mutex to protect snapshot mounts
42  * @ns_sbh: buffer heads of on-disk super blocks
43  * @ns_sbp: pointers to super block data
44  * @ns_sbwtime: previous write time of super block
45  * @ns_sbwcount: write count of super block
46  * @ns_sbsize: size of valid data in super block
47  * @ns_mount_state: file system state
48  * @ns_sb_update_freq: interval of periodical update of superblocks (in seconds)
49  * @ns_seg_seq: segment sequence counter
50  * @ns_segnum: index number of the latest full segment.
51  * @ns_nextnum: index number of the full segment index to be used next
52  * @ns_pseg_offset: offset of next partial segment in the current full segment
53  * @ns_cno: next checkpoint number
54  * @ns_ctime: write time of the last segment
55  * @ns_nongc_ctime: write time of the last segment not for cleaner operation
56  * @ns_ndirtyblks: Number of dirty data blocks
57  * @ns_last_segment_lock: lock protecting fields for the latest segment
58  * @ns_last_pseg: start block number of the latest segment
59  * @ns_last_seq: sequence value of the latest segment
60  * @ns_last_cno: checkpoint number of the latest segment
61  * @ns_prot_seq: least sequence number of segments which must not be reclaimed
62  * @ns_prev_seq: base sequence number used to decide if advance log cursor
63  * @ns_writer: log writer
64  * @ns_segctor_sem: semaphore protecting log write
65  * @ns_dat: DAT file inode
66  * @ns_cpfile: checkpoint file inode
67  * @ns_sufile: segusage file inode
68  * @ns_cptree: rb-tree of all mounted checkpoints (nilfs_root)
69  * @ns_cptree_lock: lock protecting @ns_cptree
70  * @ns_dirty_files: list of dirty files
71  * @ns_inode_lock: lock protecting @ns_dirty_files
72  * @ns_gc_inodes: dummy inodes to keep live blocks
73  * @ns_next_generation: next generation number for inodes
74  * @ns_next_gen_lock: lock protecting @ns_next_generation
75  * @ns_mount_opt: mount options
76  * @ns_resuid: uid for reserved blocks
77  * @ns_resgid: gid for reserved blocks
78  * @ns_interval: checkpoint creation interval
79  * @ns_watermark: watermark for the number of dirty buffers
80  * @ns_blocksize_bits: bit length of block size
81  * @ns_blocksize: block size
82  * @ns_nsegments: number of segments in filesystem
83  * @ns_blocks_per_segment: number of blocks per segment
84  * @ns_r_segments_percentage: reserved segments percentage
85  * @ns_nrsvsegs: number of reserved segments
86  * @ns_first_data_block: block number of first data block
87  * @ns_inode_size: size of on-disk inode
88  * @ns_first_ino: first not-special inode number
89  * @ns_crc_seed: seed value of CRC32 calculation
90  * @ns_dev_kobj: /sys/fs/<nilfs>/<device>
91  * @ns_dev_kobj_unregister: completion state
92  * @ns_dev_subgroups: <device> subgroups pointer
93  */
94 struct the_nilfs {
95 	unsigned long		ns_flags;
96 	int			ns_flushed_device;
97 
98 	struct super_block     *ns_sb;
99 	struct block_device    *ns_bdev;
100 	struct rw_semaphore	ns_sem;
101 	struct mutex		ns_snapshot_mount_mutex;
102 
103 	/*
104 	 * used for
105 	 * - loading the latest checkpoint exclusively.
106 	 * - allocating a new full segment.
107 	 */
108 	struct buffer_head     *ns_sbh[2];
109 	struct nilfs_super_block *ns_sbp[2];
110 	time64_t		ns_sbwtime;
111 	unsigned int		ns_sbwcount;
112 	unsigned int		ns_sbsize;
113 	unsigned int		ns_mount_state;
114 	unsigned int		ns_sb_update_freq;
115 
116 	/*
117 	 * The following fields are updated by a writable FS-instance.
118 	 * These fields are protected by ns_segctor_sem outside load_nilfs().
119 	 */
120 	u64			ns_seg_seq;
121 	__u64			ns_segnum;
122 	__u64			ns_nextnum;
123 	unsigned long		ns_pseg_offset;
124 	__u64			ns_cno;
125 	time64_t		ns_ctime;
126 	time64_t		ns_nongc_ctime;
127 	atomic_t		ns_ndirtyblks;
128 
129 	/*
130 	 * The following fields hold information on the latest partial segment
131 	 * written to disk with a super root.  These fields are protected by
132 	 * ns_last_segment_lock.
133 	 */
134 	spinlock_t		ns_last_segment_lock;
135 	sector_t		ns_last_pseg;
136 	u64			ns_last_seq;
137 	__u64			ns_last_cno;
138 	u64			ns_prot_seq;
139 	u64			ns_prev_seq;
140 
141 	struct nilfs_sc_info   *ns_writer;
142 	struct rw_semaphore	ns_segctor_sem;
143 
144 	/*
145 	 * Following fields are lock free except for the period before
146 	 * the_nilfs is initialized.
147 	 */
148 	struct inode	       *ns_dat;
149 	struct inode	       *ns_cpfile;
150 	struct inode	       *ns_sufile;
151 
152 	/* Checkpoint tree */
153 	struct rb_root		ns_cptree;
154 	spinlock_t		ns_cptree_lock;
155 
156 	/* Dirty inode list */
157 	struct list_head	ns_dirty_files;
158 	spinlock_t		ns_inode_lock;
159 
160 	/* GC inode list */
161 	struct list_head	ns_gc_inodes;
162 
163 	/* Inode allocator */
164 	u32			ns_next_generation;
165 	spinlock_t		ns_next_gen_lock;
166 
167 	/* Mount options */
168 	unsigned long		ns_mount_opt;
169 
170 	uid_t			ns_resuid;
171 	gid_t			ns_resgid;
172 	unsigned long		ns_interval;
173 	unsigned long		ns_watermark;
174 
175 	/* Disk layout information (static) */
176 	unsigned int		ns_blocksize_bits;
177 	unsigned int		ns_blocksize;
178 	unsigned long		ns_nsegments;
179 	unsigned long		ns_blocks_per_segment;
180 	unsigned long		ns_r_segments_percentage;
181 	unsigned long		ns_nrsvsegs;
182 	unsigned long		ns_first_data_block;
183 	int			ns_inode_size;
184 	int			ns_first_ino;
185 	u32			ns_crc_seed;
186 
187 	/* /sys/fs/<nilfs>/<device> */
188 	struct kobject ns_dev_kobj;
189 	struct completion ns_dev_kobj_unregister;
190 	struct nilfs_sysfs_dev_subgroups *ns_dev_subgroups;
191 };
192 
193 #define THE_NILFS_FNS(bit, name)					\
194 static inline void set_nilfs_##name(struct the_nilfs *nilfs)		\
195 {									\
196 	set_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);			\
197 }									\
198 static inline void clear_nilfs_##name(struct the_nilfs *nilfs)		\
199 {									\
200 	clear_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);			\
201 }									\
202 static inline int nilfs_##name(struct the_nilfs *nilfs)			\
203 {									\
204 	return test_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);		\
205 }
206 
207 THE_NILFS_FNS(INIT, init)
208 THE_NILFS_FNS(DISCONTINUED, discontinued)
209 THE_NILFS_FNS(GC_RUNNING, gc_running)
210 THE_NILFS_FNS(SB_DIRTY, sb_dirty)
211 
212 /*
213  * Mount option operations
214  */
215 #define nilfs_clear_opt(nilfs, opt)  \
216 	((nilfs)->ns_mount_opt &= ~NILFS_MOUNT_##opt)
217 #define nilfs_set_opt(nilfs, opt)  \
218 	((nilfs)->ns_mount_opt |= NILFS_MOUNT_##opt)
219 #define nilfs_test_opt(nilfs, opt) ((nilfs)->ns_mount_opt & NILFS_MOUNT_##opt)
220 #define nilfs_write_opt(nilfs, mask, opt)				\
221 	((nilfs)->ns_mount_opt =					\
222 		(((nilfs)->ns_mount_opt & ~NILFS_MOUNT_##mask) |	\
223 		 NILFS_MOUNT_##opt))					\
224 
225 /**
226  * struct nilfs_root - nilfs root object
227  * @cno: checkpoint number
228  * @rb_node: red-black tree node
229  * @count: refcount of this structure
230  * @nilfs: nilfs object
231  * @ifile: inode file
232  * @inodes_count: number of inodes
233  * @blocks_count: number of blocks
234  * @snapshot_kobj: /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot>
235  * @snapshot_kobj_unregister: completion state for kernel object
236  */
237 struct nilfs_root {
238 	__u64 cno;
239 	struct rb_node rb_node;
240 
241 	refcount_t count;
242 	struct the_nilfs *nilfs;
243 	struct inode *ifile;
244 
245 	atomic64_t inodes_count;
246 	atomic64_t blocks_count;
247 
248 	/* /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot> */
249 	struct kobject snapshot_kobj;
250 	struct completion snapshot_kobj_unregister;
251 };
252 
253 /* Special checkpoint number */
254 #define NILFS_CPTREE_CURRENT_CNO	0
255 
256 /* Minimum interval of periodical update of superblocks (in seconds) */
257 #define NILFS_SB_FREQ		10
258 
nilfs_sb_need_update(struct the_nilfs * nilfs)259 static inline int nilfs_sb_need_update(struct the_nilfs *nilfs)
260 {
261 	u64 t = ktime_get_real_seconds();
262 
263 	return t < nilfs->ns_sbwtime ||
264 		t > nilfs->ns_sbwtime + nilfs->ns_sb_update_freq;
265 }
266 
nilfs_sb_will_flip(struct the_nilfs * nilfs)267 static inline int nilfs_sb_will_flip(struct the_nilfs *nilfs)
268 {
269 	int flip_bits = nilfs->ns_sbwcount & 0x0FL;
270 
271 	return (flip_bits != 0x08 && flip_bits != 0x0F);
272 }
273 
274 void nilfs_set_last_segment(struct the_nilfs *, sector_t, u64, __u64);
275 struct the_nilfs *alloc_nilfs(struct super_block *sb);
276 void destroy_nilfs(struct the_nilfs *nilfs);
277 int init_nilfs(struct the_nilfs *nilfs, struct super_block *sb, char *data);
278 int load_nilfs(struct the_nilfs *nilfs, struct super_block *sb);
279 unsigned long nilfs_nrsvsegs(struct the_nilfs *nilfs, unsigned long nsegs);
280 void nilfs_set_nsegments(struct the_nilfs *nilfs, unsigned long nsegs);
281 int nilfs_discard_segments(struct the_nilfs *, __u64 *, size_t);
282 int nilfs_count_free_blocks(struct the_nilfs *, sector_t *);
283 struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno);
284 struct nilfs_root *nilfs_find_or_create_root(struct the_nilfs *nilfs,
285 					     __u64 cno);
286 void nilfs_put_root(struct nilfs_root *root);
287 int nilfs_near_disk_full(struct the_nilfs *);
288 void nilfs_fall_back_super_block(struct the_nilfs *);
289 void nilfs_swap_super_block(struct the_nilfs *);
290 
291 
nilfs_get_root(struct nilfs_root * root)292 static inline void nilfs_get_root(struct nilfs_root *root)
293 {
294 	refcount_inc(&root->count);
295 }
296 
nilfs_valid_fs(struct the_nilfs * nilfs)297 static inline int nilfs_valid_fs(struct the_nilfs *nilfs)
298 {
299 	unsigned int valid_fs;
300 
301 	down_read(&nilfs->ns_sem);
302 	valid_fs = (nilfs->ns_mount_state & NILFS_VALID_FS);
303 	up_read(&nilfs->ns_sem);
304 	return valid_fs;
305 }
306 
307 static inline void
nilfs_get_segment_range(struct the_nilfs * nilfs,__u64 segnum,sector_t * seg_start,sector_t * seg_end)308 nilfs_get_segment_range(struct the_nilfs *nilfs, __u64 segnum,
309 			sector_t *seg_start, sector_t *seg_end)
310 {
311 	*seg_start = (sector_t)nilfs->ns_blocks_per_segment * segnum;
312 	*seg_end = *seg_start + nilfs->ns_blocks_per_segment - 1;
313 	if (segnum == 0)
314 		*seg_start = nilfs->ns_first_data_block;
315 }
316 
317 static inline sector_t
nilfs_get_segment_start_blocknr(struct the_nilfs * nilfs,__u64 segnum)318 nilfs_get_segment_start_blocknr(struct the_nilfs *nilfs, __u64 segnum)
319 {
320 	return (segnum == 0) ? nilfs->ns_first_data_block :
321 		(sector_t)nilfs->ns_blocks_per_segment * segnum;
322 }
323 
324 static inline __u64
nilfs_get_segnum_of_block(struct the_nilfs * nilfs,sector_t blocknr)325 nilfs_get_segnum_of_block(struct the_nilfs *nilfs, sector_t blocknr)
326 {
327 	sector_t segnum = blocknr;
328 
329 	sector_div(segnum, nilfs->ns_blocks_per_segment);
330 	return segnum;
331 }
332 
333 static inline void
nilfs_terminate_segment(struct the_nilfs * nilfs,sector_t seg_start,sector_t seg_end)334 nilfs_terminate_segment(struct the_nilfs *nilfs, sector_t seg_start,
335 			sector_t seg_end)
336 {
337 	/* terminate the current full segment (used in case of I/O-error) */
338 	nilfs->ns_pseg_offset = seg_end - seg_start + 1;
339 }
340 
nilfs_shift_to_next_segment(struct the_nilfs * nilfs)341 static inline void nilfs_shift_to_next_segment(struct the_nilfs *nilfs)
342 {
343 	/* move forward with a full segment */
344 	nilfs->ns_segnum = nilfs->ns_nextnum;
345 	nilfs->ns_pseg_offset = 0;
346 	nilfs->ns_seg_seq++;
347 }
348 
nilfs_last_cno(struct the_nilfs * nilfs)349 static inline __u64 nilfs_last_cno(struct the_nilfs *nilfs)
350 {
351 	__u64 cno;
352 
353 	spin_lock(&nilfs->ns_last_segment_lock);
354 	cno = nilfs->ns_last_cno;
355 	spin_unlock(&nilfs->ns_last_segment_lock);
356 	return cno;
357 }
358 
nilfs_segment_is_active(struct the_nilfs * nilfs,__u64 n)359 static inline int nilfs_segment_is_active(struct the_nilfs *nilfs, __u64 n)
360 {
361 	return n == nilfs->ns_segnum || n == nilfs->ns_nextnum;
362 }
363 
nilfs_flush_device(struct the_nilfs * nilfs)364 static inline int nilfs_flush_device(struct the_nilfs *nilfs)
365 {
366 	int err;
367 
368 	if (!nilfs_test_opt(nilfs, BARRIER) || nilfs->ns_flushed_device)
369 		return 0;
370 
371 	nilfs->ns_flushed_device = 1;
372 	/*
373 	 * the store to ns_flushed_device must not be reordered after
374 	 * blkdev_issue_flush().
375 	 */
376 	smp_wmb();
377 
378 	err = blkdev_issue_flush(nilfs->ns_bdev);
379 	if (err != -EIO)
380 		err = 0;
381 	return err;
382 }
383 
384 #endif /* _THE_NILFS_H */
385