xref: /linux/fs/ceph/super.h (revision f8324e20f8289dffc646d64366332e05eaacab25)
1 #ifndef _FS_CEPH_SUPER_H
2 #define _FS_CEPH_SUPER_H
3 
4 #include "ceph_debug.h"
5 
6 #include <asm/unaligned.h>
7 #include <linux/backing-dev.h>
8 #include <linux/completion.h>
9 #include <linux/exportfs.h>
10 #include <linux/fs.h>
11 #include <linux/mempool.h>
12 #include <linux/pagemap.h>
13 #include <linux/wait.h>
14 #include <linux/writeback.h>
15 #include <linux/slab.h>
16 
17 #include "types.h"
18 #include "messenger.h"
19 #include "msgpool.h"
20 #include "mon_client.h"
21 #include "mds_client.h"
22 #include "osd_client.h"
23 #include "ceph_fs.h"
24 
25 /* f_type in struct statfs */
26 #define CEPH_SUPER_MAGIC 0x00c36400
27 
28 /* large granularity for statfs utilization stats to facilitate
29  * large volume sizes on 32-bit machines. */
30 #define CEPH_BLOCK_SHIFT   20  /* 1 MB */
31 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
32 
33 /*
34  * mount options
35  */
36 #define CEPH_OPT_FSID             (1<<0)
37 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
38 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
39 #define CEPH_OPT_DIRSTAT          (1<<4) /* funky `cat dirname` for stats */
40 #define CEPH_OPT_RBYTES           (1<<5) /* dir st_bytes = rbytes */
41 #define CEPH_OPT_NOCRC            (1<<6) /* no data crc on writes */
42 #define CEPH_OPT_NOASYNCREADDIR   (1<<7) /* no dcache readdir */
43 
44 #define CEPH_OPT_DEFAULT   (CEPH_OPT_RBYTES)
45 
46 #define ceph_set_opt(client, opt) \
47 	(client)->mount_args->flags |= CEPH_OPT_##opt;
48 #define ceph_test_opt(client, opt) \
49 	(!!((client)->mount_args->flags & CEPH_OPT_##opt))
50 
51 
52 struct ceph_mount_args {
53 	int sb_flags;
54 	int flags;
55 	struct ceph_fsid fsid;
56 	struct ceph_entity_addr my_addr;
57 	int num_mon;
58 	struct ceph_entity_addr *mon_addr;
59 	int mount_timeout;
60 	int osd_idle_ttl;
61 	int osd_timeout;
62 	int osd_keepalive_timeout;
63 	int wsize;
64 	int rsize;            /* max readahead */
65 	int congestion_kb;    /* max writeback in flight */
66 	int caps_wanted_delay_min, caps_wanted_delay_max;
67 	int cap_release_safety;
68 	int max_readdir;       /* max readdir result (entires) */
69 	int max_readdir_bytes; /* max readdir result (bytes) */
70 	char *snapdir_name;   /* default ".snap" */
71 	char *name;
72 	char *secret;
73 };
74 
75 /*
76  * defaults
77  */
78 #define CEPH_MOUNT_TIMEOUT_DEFAULT  60
79 #define CEPH_OSD_TIMEOUT_DEFAULT    60  /* seconds */
80 #define CEPH_OSD_KEEPALIVE_DEFAULT  5
81 #define CEPH_OSD_IDLE_TTL_DEFAULT    60
82 #define CEPH_MOUNT_RSIZE_DEFAULT    (512*1024) /* readahead */
83 #define CEPH_MAX_READDIR_DEFAULT    1024
84 #define CEPH_MAX_READDIR_BYTES_DEFAULT    (512*1024)
85 
86 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
87 #define CEPH_MSG_MAX_DATA_LEN	(16*1024*1024)
88 
89 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
90 #define CEPH_AUTH_NAME_DEFAULT   "guest"
91 /*
92  * Delay telling the MDS we no longer want caps, in case we reopen
93  * the file.  Delay a minimum amount of time, even if we send a cap
94  * message for some other reason.  Otherwise, take the oppotunity to
95  * update the mds to avoid sending another message later.
96  */
97 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
98 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
99 
100 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT        (CEPH_CAPS_PER_RELEASE * 4)
101 
102 /* mount state */
103 enum {
104 	CEPH_MOUNT_MOUNTING,
105 	CEPH_MOUNT_MOUNTED,
106 	CEPH_MOUNT_UNMOUNTING,
107 	CEPH_MOUNT_UNMOUNTED,
108 	CEPH_MOUNT_SHUTDOWN,
109 };
110 
111 /*
112  * subtract jiffies
113  */
114 static inline unsigned long time_sub(unsigned long a, unsigned long b)
115 {
116 	BUG_ON(time_after(b, a));
117 	return (long)a - (long)b;
118 }
119 
120 /*
121  * per-filesystem client state
122  *
123  * possibly shared by multiple mount points, if they are
124  * mounting the same ceph filesystem/cluster.
125  */
126 struct ceph_client {
127 	struct ceph_fsid fsid;
128 	bool have_fsid;
129 
130 	struct mutex mount_mutex;       /* serialize mount attempts */
131 	struct ceph_mount_args *mount_args;
132 
133 	struct super_block *sb;
134 
135 	unsigned long mount_state;
136 	wait_queue_head_t auth_wq;
137 
138 	int auth_err;
139 
140 	int min_caps;                  /* min caps i added */
141 
142 	struct ceph_messenger *msgr;   /* messenger instance */
143 	struct ceph_mon_client monc;
144 	struct ceph_mds_client mdsc;
145 	struct ceph_osd_client osdc;
146 
147 	/* writeback */
148 	mempool_t *wb_pagevec_pool;
149 	struct workqueue_struct *wb_wq;
150 	struct workqueue_struct *pg_inv_wq;
151 	struct workqueue_struct *trunc_wq;
152 	atomic_long_t writeback_count;
153 
154 	struct backing_dev_info backing_dev_info;
155 
156 #ifdef CONFIG_DEBUG_FS
157 	struct dentry *debugfs_monmap;
158 	struct dentry *debugfs_mdsmap, *debugfs_osdmap;
159 	struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
160 	struct dentry *debugfs_congestion_kb;
161 	struct dentry *debugfs_bdi;
162 #endif
163 };
164 
165 /*
166  * File i/o capability.  This tracks shared state with the metadata
167  * server that allows us to cache or writeback attributes or to read
168  * and write data.  For any given inode, we should have one or more
169  * capabilities, one issued by each metadata server, and our
170  * cumulative access is the OR of all issued capabilities.
171  *
172  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
173  * session capability lists.
174  */
175 struct ceph_cap {
176 	struct ceph_inode_info *ci;
177 	struct rb_node ci_node;          /* per-ci cap tree */
178 	struct ceph_mds_session *session;
179 	struct list_head session_caps;   /* per-session caplist */
180 	int mds;
181 	u64 cap_id;       /* unique cap id (mds provided) */
182 	int issued;       /* latest, from the mds */
183 	int implemented;  /* implemented superset of issued (for revocation) */
184 	int mds_wanted;
185 	u32 seq, issue_seq, mseq;
186 	u32 cap_gen;      /* active/stale cycle */
187 	unsigned long last_used;
188 	struct list_head caps_item;
189 };
190 
191 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
192 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
193 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
194 
195 /*
196  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
197  * we first complete any in-process sync writes and writeback any dirty
198  * data before flushing the snapped state (tracked here) back to the MDS.
199  */
200 struct ceph_cap_snap {
201 	atomic_t nref;
202 	struct ceph_inode_info *ci;
203 	struct list_head ci_item, flushing_item;
204 
205 	u64 follows, flush_tid;
206 	int issued, dirty;
207 	struct ceph_snap_context *context;
208 
209 	mode_t mode;
210 	uid_t uid;
211 	gid_t gid;
212 
213 	void *xattr_blob;
214 	int xattr_len;
215 	u64 xattr_version;
216 
217 	u64 size;
218 	struct timespec mtime, atime, ctime;
219 	u64 time_warp_seq;
220 	int writing;   /* a sync write is still in progress */
221 	int dirty_pages;     /* dirty pages awaiting writeback */
222 };
223 
224 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
225 {
226 	if (atomic_dec_and_test(&capsnap->nref))
227 		kfree(capsnap);
228 }
229 
230 /*
231  * The frag tree describes how a directory is fragmented, potentially across
232  * multiple metadata servers.  It is also used to indicate points where
233  * metadata authority is delegated, and whether/where metadata is replicated.
234  *
235  * A _leaf_ frag will be present in the i_fragtree IFF there is
236  * delegation info.  That is, if mds >= 0 || ndist > 0.
237  */
238 #define CEPH_MAX_DIRFRAG_REP 4
239 
240 struct ceph_inode_frag {
241 	struct rb_node node;
242 
243 	/* fragtree state */
244 	u32 frag;
245 	int split_by;         /* i.e. 2^(split_by) children */
246 
247 	/* delegation and replication info */
248 	int mds;              /* -1 if same authority as parent */
249 	int ndist;            /* >0 if replicated */
250 	int dist[CEPH_MAX_DIRFRAG_REP];
251 };
252 
253 /*
254  * We cache inode xattrs as an encoded blob until they are first used,
255  * at which point we parse them into an rbtree.
256  */
257 struct ceph_inode_xattr {
258 	struct rb_node node;
259 
260 	const char *name;
261 	int name_len;
262 	const char *val;
263 	int val_len;
264 	int dirty;
265 
266 	int should_free_name;
267 	int should_free_val;
268 };
269 
270 struct ceph_inode_xattrs_info {
271 	/*
272 	 * (still encoded) xattr blob. we avoid the overhead of parsing
273 	 * this until someone actually calls getxattr, etc.
274 	 *
275 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
276 	 * NULL means we don't know.
277 	*/
278 	struct ceph_buffer *blob, *prealloc_blob;
279 
280 	struct rb_root index;
281 	bool dirty;
282 	int count;
283 	int names_size;
284 	int vals_size;
285 	u64 version, index_version;
286 };
287 
288 /*
289  * Ceph inode.
290  */
291 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
292 #define CEPH_I_NODELAY   4  /* do not delay cap release */
293 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
294 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
295 
296 struct ceph_inode_info {
297 	struct ceph_vino i_vino;   /* ceph ino + snap */
298 
299 	u64 i_version;
300 	u32 i_time_warp_seq;
301 
302 	unsigned i_ceph_flags;
303 	unsigned long i_release_count;
304 
305 	struct ceph_file_layout i_layout;
306 	char *i_symlink;
307 
308 	/* for dirs */
309 	struct timespec i_rctime;
310 	u64 i_rbytes, i_rfiles, i_rsubdirs;
311 	u64 i_files, i_subdirs;
312 	u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
313 
314 	struct rb_root i_fragtree;
315 	struct mutex i_fragtree_mutex;
316 
317 	struct ceph_inode_xattrs_info i_xattrs;
318 
319 	/* capabilities.  protected _both_ by i_lock and cap->session's
320 	 * s_mutex. */
321 	struct rb_root i_caps;           /* cap list */
322 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
323 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
324 	struct list_head i_dirty_item, i_flushing_item;
325 	u64 i_cap_flush_seq;
326 	/* we need to track cap writeback on a per-cap-bit basis, to allow
327 	 * overlapping, pipelined cap flushes to the mds.  we can probably
328 	 * reduce the tid to 8 bits if we're concerned about inode size. */
329 	u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
330 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
331 	unsigned long i_hold_caps_min; /* jiffies */
332 	unsigned long i_hold_caps_max; /* jiffies */
333 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
334 	int i_cap_exporting_mds;         /* to handle cap migration between */
335 	unsigned i_cap_exporting_mseq;   /*  mds's. */
336 	unsigned i_cap_exporting_issued;
337 	struct ceph_cap_reservation i_cap_migration_resv;
338 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
339 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 */
340 	unsigned i_snap_caps;           /* cap bits for snapped files */
341 
342 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
343 
344 	u32 i_truncate_seq;        /* last truncate to smaller size */
345 	u64 i_truncate_size;       /*  and the size we last truncated down to */
346 	int i_truncate_pending;    /*  still need to call vmtruncate */
347 
348 	u64 i_max_size;            /* max file size authorized by mds */
349 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
350 	u64 i_wanted_max_size;     /* offset we'd like to write too */
351 	u64 i_requested_max_size;  /* max_size we've requested */
352 
353 	/* held references to caps */
354 	int i_pin_ref;
355 	int i_rd_ref, i_rdcache_ref, i_wr_ref;
356 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
357 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
358 	u32 i_rdcache_gen;      /* we increment this each time we get
359 				   FILE_CACHE.  If it's non-zero, we
360 				   _may_ have cached pages. */
361 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
362 
363 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
364 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
365 	spinlock_t i_unsafe_lock;
366 
367 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
368 	int i_snap_realm_counter; /* snap realm (if caps) */
369 	struct list_head i_snap_realm_item;
370 	struct list_head i_snap_flush_item;
371 
372 	struct work_struct i_wb_work;  /* writeback work */
373 	struct work_struct i_pg_inv_work;  /* page invalidation work */
374 
375 	struct work_struct i_vmtruncate_work;
376 
377 	struct inode vfs_inode; /* at end */
378 };
379 
380 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
381 {
382 	return container_of(inode, struct ceph_inode_info, vfs_inode);
383 }
384 
385 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
386 {
387 	struct ceph_inode_info *ci = ceph_inode(inode);
388 
389 	spin_lock(&inode->i_lock);
390 	ci->i_ceph_flags &= ~mask;
391 	spin_unlock(&inode->i_lock);
392 }
393 
394 static inline void ceph_i_set(struct inode *inode, unsigned mask)
395 {
396 	struct ceph_inode_info *ci = ceph_inode(inode);
397 
398 	spin_lock(&inode->i_lock);
399 	ci->i_ceph_flags |= mask;
400 	spin_unlock(&inode->i_lock);
401 }
402 
403 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
404 {
405 	struct ceph_inode_info *ci = ceph_inode(inode);
406 	bool r;
407 
408 	smp_mb();
409 	r = (ci->i_ceph_flags & mask) == mask;
410 	return r;
411 }
412 
413 
414 /* find a specific frag @f */
415 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
416 						u32 f);
417 
418 /*
419  * choose fragment for value @v.  copy frag content to pfrag, if leaf
420  * exists
421  */
422 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
423 			    struct ceph_inode_frag *pfrag,
424 			    int *found);
425 
426 /*
427  * Ceph dentry state
428  */
429 struct ceph_dentry_info {
430 	struct ceph_mds_session *lease_session;
431 	u32 lease_gen, lease_shared_gen;
432 	u32 lease_seq;
433 	unsigned long lease_renew_after, lease_renew_from;
434 	struct list_head lru;
435 	struct dentry *dentry;
436 	u64 time;
437 	u64 offset;
438 };
439 
440 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
441 {
442 	return (struct ceph_dentry_info *)dentry->d_fsdata;
443 }
444 
445 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
446 {
447 	return ((loff_t)frag << 32) | (loff_t)off;
448 }
449 
450 /*
451  * ino_t is <64 bits on many architectures, blech.
452  *
453  * don't include snap in ino hash, at least for now.
454  */
455 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
456 {
457 	ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
458 #if BITS_PER_LONG == 32
459 	ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
460 	if (!ino)
461 		ino = 1;
462 #endif
463 	return ino;
464 }
465 
466 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
467 {
468 	ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
469 	inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
470 	return 0;
471 }
472 
473 static inline struct ceph_vino ceph_vino(struct inode *inode)
474 {
475 	return ceph_inode(inode)->i_vino;
476 }
477 
478 /* for printf-style formatting */
479 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
480 
481 static inline u64 ceph_ino(struct inode *inode)
482 {
483 	return ceph_inode(inode)->i_vino.ino;
484 }
485 static inline u64 ceph_snap(struct inode *inode)
486 {
487 	return ceph_inode(inode)->i_vino.snap;
488 }
489 
490 static inline int ceph_ino_compare(struct inode *inode, void *data)
491 {
492 	struct ceph_vino *pvino = (struct ceph_vino *)data;
493 	struct ceph_inode_info *ci = ceph_inode(inode);
494 	return ci->i_vino.ino == pvino->ino &&
495 		ci->i_vino.snap == pvino->snap;
496 }
497 
498 static inline struct inode *ceph_find_inode(struct super_block *sb,
499 					    struct ceph_vino vino)
500 {
501 	ino_t t = ceph_vino_to_ino(vino);
502 	return ilookup5(sb, t, ceph_ino_compare, &vino);
503 }
504 
505 
506 /*
507  * caps helpers
508  */
509 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
510 {
511 	return !RB_EMPTY_ROOT(&ci->i_caps);
512 }
513 
514 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
515 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
516 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
517 				    struct ceph_cap *cap);
518 
519 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
520 {
521 	int issued;
522 	spin_lock(&ci->vfs_inode.i_lock);
523 	issued = __ceph_caps_issued(ci, NULL);
524 	spin_unlock(&ci->vfs_inode.i_lock);
525 	return issued;
526 }
527 
528 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
529 					int touch)
530 {
531 	int r;
532 	spin_lock(&ci->vfs_inode.i_lock);
533 	r = __ceph_caps_issued_mask(ci, mask, touch);
534 	spin_unlock(&ci->vfs_inode.i_lock);
535 	return r;
536 }
537 
538 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
539 {
540 	return ci->i_dirty_caps | ci->i_flushing_caps;
541 }
542 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
543 
544 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
545 extern int __ceph_caps_used(struct ceph_inode_info *ci);
546 
547 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
548 
549 /*
550  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
551  */
552 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
553 {
554 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
555 	if (w & CEPH_CAP_FILE_BUFFER)
556 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
557 	return w;
558 }
559 
560 /* what the mds thinks we want */
561 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
562 
563 extern void ceph_caps_init(void);
564 extern void ceph_caps_finalize(void);
565 extern void ceph_adjust_min_caps(int delta);
566 extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need);
567 extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx);
568 extern void ceph_reservation_status(struct ceph_client *client,
569 				    int *total, int *avail, int *used,
570 				    int *reserved, int *min);
571 
572 static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
573 {
574 	return (struct ceph_client *)inode->i_sb->s_fs_info;
575 }
576 
577 static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
578 {
579 	return (struct ceph_client *)sb->s_fs_info;
580 }
581 
582 
583 /*
584  * we keep buffered readdir results attached to file->private_data
585  */
586 struct ceph_file_info {
587 	int fmode;     /* initialized on open */
588 
589 	/* readdir: position within the dir */
590 	u32 frag;
591 	struct ceph_mds_request *last_readdir;
592 	int at_end;
593 
594 	/* readdir: position within a frag */
595 	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
596 	u64 next_offset;       /* offset of next chunk (last_name's + 1) */
597 	char *last_name;       /* last entry in previous chunk */
598 	struct dentry *dentry; /* next dentry (for dcache readdir) */
599 	unsigned long dir_release_count;
600 
601 	/* used for -o dirstat read() on directory thing */
602 	char *dir_info;
603 	int dir_info_len;
604 };
605 
606 
607 
608 /*
609  * snapshots
610  */
611 
612 /*
613  * A "snap context" is the set of existing snapshots when we
614  * write data.  It is used by the OSD to guide its COW behavior.
615  *
616  * The ceph_snap_context is refcounted, and attached to each dirty
617  * page, indicating which context the dirty data belonged when it was
618  * dirtied.
619  */
620 struct ceph_snap_context {
621 	atomic_t nref;
622 	u64 seq;
623 	int num_snaps;
624 	u64 snaps[];
625 };
626 
627 static inline struct ceph_snap_context *
628 ceph_get_snap_context(struct ceph_snap_context *sc)
629 {
630 	/*
631 	printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
632 	       atomic_read(&sc->nref)+1);
633 	*/
634 	if (sc)
635 		atomic_inc(&sc->nref);
636 	return sc;
637 }
638 
639 static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
640 {
641 	if (!sc)
642 		return;
643 	/*
644 	printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
645 	       atomic_read(&sc->nref)-1);
646 	*/
647 	if (atomic_dec_and_test(&sc->nref)) {
648 		/*printk(" deleting snap_context %p\n", sc);*/
649 		kfree(sc);
650 	}
651 }
652 
653 /*
654  * A "snap realm" describes a subset of the file hierarchy sharing
655  * the same set of snapshots that apply to it.  The realms themselves
656  * are organized into a hierarchy, such that children inherit (some of)
657  * the snapshots of their parents.
658  *
659  * All inodes within the realm that have capabilities are linked into a
660  * per-realm list.
661  */
662 struct ceph_snap_realm {
663 	u64 ino;
664 	atomic_t nref;
665 	struct rb_node node;
666 
667 	u64 created, seq;
668 	u64 parent_ino;
669 	u64 parent_since;   /* snapid when our current parent became so */
670 
671 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
672 	int num_prior_parent_snaps;   /*  had prior to parent_since */
673 	u64 *snaps;                   /* snaps specific to this realm */
674 	int num_snaps;
675 
676 	struct ceph_snap_realm *parent;
677 	struct list_head children;       /* list of child realms */
678 	struct list_head child_item;
679 
680 	struct list_head empty_item;     /* if i have ref==0 */
681 
682 	/* the current set of snaps for this realm */
683 	struct ceph_snap_context *cached_context;
684 
685 	struct list_head inodes_with_caps;
686 	spinlock_t inodes_with_caps_lock;
687 };
688 
689 
690 
691 /*
692  * calculate the number of pages a given length and offset map onto,
693  * if we align the data.
694  */
695 static inline int calc_pages_for(u64 off, u64 len)
696 {
697 	return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
698 		(off >> PAGE_CACHE_SHIFT);
699 }
700 
701 
702 
703 /* snap.c */
704 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
705 					       u64 ino);
706 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
707 				struct ceph_snap_realm *realm);
708 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
709 				struct ceph_snap_realm *realm);
710 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
711 				  void *p, void *e, bool deletion);
712 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
713 			     struct ceph_mds_session *session,
714 			     struct ceph_msg *msg);
715 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
716 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
717 				  struct ceph_cap_snap *capsnap);
718 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
719 
720 /*
721  * a cap_snap is "pending" if it is still awaiting an in-progress
722  * sync write (that may/may not still update size, mtime, etc.).
723  */
724 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
725 {
726 	return !list_empty(&ci->i_cap_snaps) &&
727 		list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
728 			   ci_item)->writing;
729 }
730 
731 
732 /* super.c */
733 extern struct kmem_cache *ceph_inode_cachep;
734 extern struct kmem_cache *ceph_cap_cachep;
735 extern struct kmem_cache *ceph_dentry_cachep;
736 extern struct kmem_cache *ceph_file_cachep;
737 
738 extern const char *ceph_msg_type_name(int type);
739 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
740 
741 #define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \
742 	"%02x%02x%02x%02x%02x%02x"
743 #define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \
744 		(f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7],    \
745 		(f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11],  \
746 		(f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15]
747 
748 /* inode.c */
749 extern const struct inode_operations ceph_file_iops;
750 
751 extern struct inode *ceph_alloc_inode(struct super_block *sb);
752 extern void ceph_destroy_inode(struct inode *inode);
753 
754 extern struct inode *ceph_get_inode(struct super_block *sb,
755 				    struct ceph_vino vino);
756 extern struct inode *ceph_get_snapdir(struct inode *parent);
757 extern int ceph_fill_file_size(struct inode *inode, int issued,
758 			       u32 truncate_seq, u64 truncate_size, u64 size);
759 extern void ceph_fill_file_time(struct inode *inode, int issued,
760 				u64 time_warp_seq, struct timespec *ctime,
761 				struct timespec *mtime, struct timespec *atime);
762 extern int ceph_fill_trace(struct super_block *sb,
763 			   struct ceph_mds_request *req,
764 			   struct ceph_mds_session *session);
765 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
766 				    struct ceph_mds_session *session);
767 
768 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
769 
770 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
771 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
772 extern void ceph_queue_vmtruncate(struct inode *inode);
773 
774 extern void ceph_queue_invalidate(struct inode *inode);
775 extern void ceph_queue_writeback(struct inode *inode);
776 
777 extern int ceph_do_getattr(struct inode *inode, int mask);
778 extern int ceph_permission(struct inode *inode, int mask);
779 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
780 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
781 			struct kstat *stat);
782 
783 /* xattr.c */
784 extern int ceph_setxattr(struct dentry *, const char *, const void *,
785 			 size_t, int);
786 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
787 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
788 extern int ceph_removexattr(struct dentry *, const char *);
789 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
790 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
791 
792 /* caps.c */
793 extern const char *ceph_cap_string(int c);
794 extern void ceph_handle_caps(struct ceph_mds_session *session,
795 			     struct ceph_msg *msg);
796 extern int ceph_add_cap(struct inode *inode,
797 			struct ceph_mds_session *session, u64 cap_id,
798 			int fmode, unsigned issued, unsigned wanted,
799 			unsigned cap, unsigned seq, u64 realmino, int flags,
800 			struct ceph_cap_reservation *caps_reservation);
801 extern void __ceph_remove_cap(struct ceph_cap *cap);
802 static inline void ceph_remove_cap(struct ceph_cap *cap)
803 {
804 	struct inode *inode = &cap->ci->vfs_inode;
805 	spin_lock(&inode->i_lock);
806 	__ceph_remove_cap(cap);
807 	spin_unlock(&inode->i_lock);
808 }
809 extern void ceph_put_cap(struct ceph_cap *cap);
810 
811 extern void ceph_queue_caps_release(struct inode *inode);
812 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
813 extern int ceph_fsync(struct file *file, int datasync);
814 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
815 				    struct ceph_mds_session *session);
816 extern int ceph_get_cap_mds(struct inode *inode);
817 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
818 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
819 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
820 				       struct ceph_snap_context *snapc);
821 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
822 			       struct ceph_mds_session **psession);
823 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
824 			    struct ceph_mds_session *session);
825 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
826 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
827 
828 extern int ceph_encode_inode_release(void **p, struct inode *inode,
829 				     int mds, int drop, int unless, int force);
830 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
831 				      int mds, int drop, int unless);
832 
833 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
834 			 int *got, loff_t endoff);
835 
836 /* for counting open files by mode */
837 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
838 {
839 	ci->i_nr_by_mode[mode]++;
840 }
841 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
842 
843 /* addr.c */
844 extern const struct address_space_operations ceph_aops;
845 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
846 
847 /* file.c */
848 extern const struct file_operations ceph_file_fops;
849 extern const struct address_space_operations ceph_aops;
850 extern int ceph_open(struct inode *inode, struct file *file);
851 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
852 				       struct nameidata *nd, int mode,
853 				       int locked_dir);
854 extern int ceph_release(struct inode *inode, struct file *filp);
855 extern void ceph_release_page_vector(struct page **pages, int num_pages);
856 
857 /* dir.c */
858 extern const struct file_operations ceph_dir_fops;
859 extern const struct inode_operations ceph_dir_iops;
860 extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
861 	ceph_snapdir_dentry_ops;
862 
863 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
864 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
865 					 struct dentry *dentry, int err);
866 
867 extern void ceph_dentry_lru_add(struct dentry *dn);
868 extern void ceph_dentry_lru_touch(struct dentry *dn);
869 extern void ceph_dentry_lru_del(struct dentry *dn);
870 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
871 
872 /*
873  * our d_ops vary depending on whether the inode is live,
874  * snapshotted (read-only), or a virtual ".snap" directory.
875  */
876 int ceph_init_dentry(struct dentry *dentry);
877 
878 
879 /* ioctl.c */
880 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
881 
882 /* export.c */
883 extern const struct export_operations ceph_export_ops;
884 
885 /* debugfs.c */
886 extern int ceph_debugfs_init(void);
887 extern void ceph_debugfs_cleanup(void);
888 extern int ceph_debugfs_client_init(struct ceph_client *client);
889 extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
890 
891 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
892 {
893 	if (dentry && dentry->d_parent)
894 		return dentry->d_parent->d_inode;
895 
896 	return NULL;
897 }
898 
899 #endif /* _FS_CEPH_SUPER_H */
900