xref: /linux/fs/ceph/super.h (revision b3b77c8caef1750ebeea1054e39e358550ea9f55)
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/slab.h>
14 #include <linux/wait.h>
15 #include <linux/writeback.h>
16 #include <linux/slab.h>
17 
18 #include "types.h"
19 #include "messenger.h"
20 #include "msgpool.h"
21 #include "mon_client.h"
22 #include "mds_client.h"
23 #include "osd_client.h"
24 #include "ceph_fs.h"
25 
26 /* f_type in struct statfs */
27 #define CEPH_SUPER_MAGIC 0x00c36400
28 
29 /* large granularity for statfs utilization stats to facilitate
30  * large volume sizes on 32-bit machines. */
31 #define CEPH_BLOCK_SHIFT   20  /* 1 MB */
32 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
33 
34 /*
35  * mount options
36  */
37 #define CEPH_OPT_FSID             (1<<0)
38 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
39 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
40 #define CEPH_OPT_DIRSTAT          (1<<4) /* funky `cat dirname` for stats */
41 #define CEPH_OPT_RBYTES           (1<<5) /* dir st_bytes = rbytes */
42 #define CEPH_OPT_NOCRC            (1<<6) /* no data crc on writes */
43 #define CEPH_OPT_NOASYNCREADDIR   (1<<7) /* no dcache readdir */
44 
45 #define CEPH_OPT_DEFAULT   (CEPH_OPT_RBYTES)
46 
47 #define ceph_set_opt(client, opt) \
48 	(client)->mount_args->flags |= CEPH_OPT_##opt;
49 #define ceph_test_opt(client, opt) \
50 	(!!((client)->mount_args->flags & CEPH_OPT_##opt))
51 
52 
53 struct ceph_mount_args {
54 	int sb_flags;
55 	int flags;
56 	struct ceph_fsid fsid;
57 	struct ceph_entity_addr my_addr;
58 	int num_mon;
59 	struct ceph_entity_addr *mon_addr;
60 	int mount_timeout;
61 	int osd_idle_ttl;
62 	int osd_timeout;
63 	int osd_keepalive_timeout;
64 	int wsize;
65 	int rsize;            /* max readahead */
66 	int congestion_kb;    /* max writeback in flight */
67 	int caps_wanted_delay_min, caps_wanted_delay_max;
68 	int cap_release_safety;
69 	int max_readdir;       /* max readdir result (entires) */
70 	int max_readdir_bytes; /* max readdir result (bytes) */
71 	char *snapdir_name;   /* default ".snap" */
72 	char *name;
73 	char *secret;
74 };
75 
76 /*
77  * defaults
78  */
79 #define CEPH_MOUNT_TIMEOUT_DEFAULT  60
80 #define CEPH_OSD_TIMEOUT_DEFAULT    60  /* seconds */
81 #define CEPH_OSD_KEEPALIVE_DEFAULT  5
82 #define CEPH_OSD_IDLE_TTL_DEFAULT    60
83 #define CEPH_MOUNT_RSIZE_DEFAULT    (512*1024) /* readahead */
84 #define CEPH_MAX_READDIR_DEFAULT    1024
85 #define CEPH_MAX_READDIR_BYTES_DEFAULT    (512*1024)
86 
87 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
88 #define CEPH_MSG_MAX_DATA_LEN	(16*1024*1024)
89 
90 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
91 #define CEPH_AUTH_NAME_DEFAULT   "guest"
92 /*
93  * Delay telling the MDS we no longer want caps, in case we reopen
94  * the file.  Delay a minimum amount of time, even if we send a cap
95  * message for some other reason.  Otherwise, take the oppotunity to
96  * update the mds to avoid sending another message later.
97  */
98 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
99 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
100 
101 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT        (CEPH_CAPS_PER_RELEASE * 4)
102 
103 /* mount state */
104 enum {
105 	CEPH_MOUNT_MOUNTING,
106 	CEPH_MOUNT_MOUNTED,
107 	CEPH_MOUNT_UNMOUNTING,
108 	CEPH_MOUNT_UNMOUNTED,
109 	CEPH_MOUNT_SHUTDOWN,
110 };
111 
112 /*
113  * subtract jiffies
114  */
115 static inline unsigned long time_sub(unsigned long a, unsigned long b)
116 {
117 	BUG_ON(time_after(b, a));
118 	return (long)a - (long)b;
119 }
120 
121 /*
122  * per-filesystem client state
123  *
124  * possibly shared by multiple mount points, if they are
125  * mounting the same ceph filesystem/cluster.
126  */
127 struct ceph_client {
128 	struct ceph_fsid fsid;
129 	bool have_fsid;
130 
131 	struct mutex mount_mutex;       /* serialize mount attempts */
132 	struct ceph_mount_args *mount_args;
133 
134 	struct super_block *sb;
135 
136 	unsigned long mount_state;
137 	wait_queue_head_t auth_wq;
138 
139 	int auth_err;
140 
141 	int min_caps;                  /* min caps i added */
142 
143 	struct ceph_messenger *msgr;   /* messenger instance */
144 	struct ceph_mon_client monc;
145 	struct ceph_mds_client mdsc;
146 	struct ceph_osd_client osdc;
147 
148 	/* writeback */
149 	mempool_t *wb_pagevec_pool;
150 	struct workqueue_struct *wb_wq;
151 	struct workqueue_struct *pg_inv_wq;
152 	struct workqueue_struct *trunc_wq;
153 	atomic_long_t writeback_count;
154 
155 	struct backing_dev_info backing_dev_info;
156 
157 #ifdef CONFIG_DEBUG_FS
158 	struct dentry *debugfs_monmap;
159 	struct dentry *debugfs_mdsmap, *debugfs_osdmap;
160 	struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
161 	struct dentry *debugfs_congestion_kb;
162 	struct dentry *debugfs_bdi;
163 #endif
164 };
165 
166 /*
167  * File i/o capability.  This tracks shared state with the metadata
168  * server that allows us to cache or writeback attributes or to read
169  * and write data.  For any given inode, we should have one or more
170  * capabilities, one issued by each metadata server, and our
171  * cumulative access is the OR of all issued capabilities.
172  *
173  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
174  * session capability lists.
175  */
176 struct ceph_cap {
177 	struct ceph_inode_info *ci;
178 	struct rb_node ci_node;          /* per-ci cap tree */
179 	struct ceph_mds_session *session;
180 	struct list_head session_caps;   /* per-session caplist */
181 	int mds;
182 	u64 cap_id;       /* unique cap id (mds provided) */
183 	int issued;       /* latest, from the mds */
184 	int implemented;  /* implemented superset of issued (for revocation) */
185 	int mds_wanted;
186 	u32 seq, issue_seq, mseq;
187 	u32 cap_gen;      /* active/stale cycle */
188 	unsigned long last_used;
189 	struct list_head caps_item;
190 };
191 
192 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
193 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
194 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
195 
196 /*
197  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
198  * we first complete any in-process sync writes and writeback any dirty
199  * data before flushing the snapped state (tracked here) back to the MDS.
200  */
201 struct ceph_cap_snap {
202 	atomic_t nref;
203 	struct ceph_inode_info *ci;
204 	struct list_head ci_item, flushing_item;
205 
206 	u64 follows, flush_tid;
207 	int issued, dirty;
208 	struct ceph_snap_context *context;
209 
210 	mode_t mode;
211 	uid_t uid;
212 	gid_t gid;
213 
214 	void *xattr_blob;
215 	int xattr_len;
216 	u64 xattr_version;
217 
218 	u64 size;
219 	struct timespec mtime, atime, ctime;
220 	u64 time_warp_seq;
221 	int writing;   /* a sync write is still in progress */
222 	int dirty_pages;     /* dirty pages awaiting writeback */
223 };
224 
225 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
226 {
227 	if (atomic_dec_and_test(&capsnap->nref))
228 		kfree(capsnap);
229 }
230 
231 /*
232  * The frag tree describes how a directory is fragmented, potentially across
233  * multiple metadata servers.  It is also used to indicate points where
234  * metadata authority is delegated, and whether/where metadata is replicated.
235  *
236  * A _leaf_ frag will be present in the i_fragtree IFF there is
237  * delegation info.  That is, if mds >= 0 || ndist > 0.
238  */
239 #define CEPH_MAX_DIRFRAG_REP 4
240 
241 struct ceph_inode_frag {
242 	struct rb_node node;
243 
244 	/* fragtree state */
245 	u32 frag;
246 	int split_by;         /* i.e. 2^(split_by) children */
247 
248 	/* delegation and replication info */
249 	int mds;              /* -1 if same authority as parent */
250 	int ndist;            /* >0 if replicated */
251 	int dist[CEPH_MAX_DIRFRAG_REP];
252 };
253 
254 /*
255  * We cache inode xattrs as an encoded blob until they are first used,
256  * at which point we parse them into an rbtree.
257  */
258 struct ceph_inode_xattr {
259 	struct rb_node node;
260 
261 	const char *name;
262 	int name_len;
263 	const char *val;
264 	int val_len;
265 	int dirty;
266 
267 	int should_free_name;
268 	int should_free_val;
269 };
270 
271 struct ceph_inode_xattrs_info {
272 	/*
273 	 * (still encoded) xattr blob. we avoid the overhead of parsing
274 	 * this until someone actually calls getxattr, etc.
275 	 *
276 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
277 	 * NULL means we don't know.
278 	*/
279 	struct ceph_buffer *blob, *prealloc_blob;
280 
281 	struct rb_root index;
282 	bool dirty;
283 	int count;
284 	int names_size;
285 	int vals_size;
286 	u64 version, index_version;
287 };
288 
289 /*
290  * Ceph inode.
291  */
292 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
293 #define CEPH_I_NODELAY   4  /* do not delay cap release */
294 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
295 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
296 
297 struct ceph_inode_info {
298 	struct ceph_vino i_vino;   /* ceph ino + snap */
299 
300 	u64 i_version;
301 	u32 i_time_warp_seq;
302 
303 	unsigned i_ceph_flags;
304 	unsigned long i_release_count;
305 
306 	struct ceph_file_layout i_layout;
307 	char *i_symlink;
308 
309 	/* for dirs */
310 	struct timespec i_rctime;
311 	u64 i_rbytes, i_rfiles, i_rsubdirs;
312 	u64 i_files, i_subdirs;
313 	u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
314 
315 	struct rb_root i_fragtree;
316 	struct mutex i_fragtree_mutex;
317 
318 	struct ceph_inode_xattrs_info i_xattrs;
319 
320 	/* capabilities.  protected _both_ by i_lock and cap->session's
321 	 * s_mutex. */
322 	struct rb_root i_caps;           /* cap list */
323 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
324 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
325 	struct list_head i_dirty_item, i_flushing_item;
326 	u64 i_cap_flush_seq;
327 	/* we need to track cap writeback on a per-cap-bit basis, to allow
328 	 * overlapping, pipelined cap flushes to the mds.  we can probably
329 	 * reduce the tid to 8 bits if we're concerned about inode size. */
330 	u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
331 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
332 	unsigned long i_hold_caps_min; /* jiffies */
333 	unsigned long i_hold_caps_max; /* jiffies */
334 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
335 	int i_cap_exporting_mds;         /* to handle cap migration between */
336 	unsigned i_cap_exporting_mseq;   /*  mds's. */
337 	unsigned i_cap_exporting_issued;
338 	struct ceph_cap_reservation i_cap_migration_resv;
339 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
340 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 */
341 	unsigned i_snap_caps;           /* cap bits for snapped files */
342 
343 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
344 
345 	u32 i_truncate_seq;        /* last truncate to smaller size */
346 	u64 i_truncate_size;       /*  and the size we last truncated down to */
347 	int i_truncate_pending;    /*  still need to call vmtruncate */
348 
349 	u64 i_max_size;            /* max file size authorized by mds */
350 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
351 	u64 i_wanted_max_size;     /* offset we'd like to write too */
352 	u64 i_requested_max_size;  /* max_size we've requested */
353 
354 	/* held references to caps */
355 	int i_pin_ref;
356 	int i_rd_ref, i_rdcache_ref, i_wr_ref;
357 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
358 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
359 	u32 i_rdcache_gen;      /* we increment this each time we get
360 				   FILE_CACHE.  If it's non-zero, we
361 				   _may_ have cached pages. */
362 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
363 
364 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
365 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
366 	spinlock_t i_unsafe_lock;
367 
368 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
369 	int i_snap_realm_counter; /* snap realm (if caps) */
370 	struct list_head i_snap_realm_item;
371 	struct list_head i_snap_flush_item;
372 
373 	struct work_struct i_wb_work;  /* writeback work */
374 	struct work_struct i_pg_inv_work;  /* page invalidation work */
375 
376 	struct work_struct i_vmtruncate_work;
377 
378 	struct inode vfs_inode; /* at end */
379 };
380 
381 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
382 {
383 	return container_of(inode, struct ceph_inode_info, vfs_inode);
384 }
385 
386 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
387 {
388 	struct ceph_inode_info *ci = ceph_inode(inode);
389 
390 	spin_lock(&inode->i_lock);
391 	ci->i_ceph_flags &= ~mask;
392 	spin_unlock(&inode->i_lock);
393 }
394 
395 static inline void ceph_i_set(struct inode *inode, unsigned mask)
396 {
397 	struct ceph_inode_info *ci = ceph_inode(inode);
398 
399 	spin_lock(&inode->i_lock);
400 	ci->i_ceph_flags |= mask;
401 	spin_unlock(&inode->i_lock);
402 }
403 
404 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
405 {
406 	struct ceph_inode_info *ci = ceph_inode(inode);
407 	bool r;
408 
409 	smp_mb();
410 	r = (ci->i_ceph_flags & mask) == mask;
411 	return r;
412 }
413 
414 
415 /* find a specific frag @f */
416 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
417 						u32 f);
418 
419 /*
420  * choose fragment for value @v.  copy frag content to pfrag, if leaf
421  * exists
422  */
423 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
424 			    struct ceph_inode_frag *pfrag,
425 			    int *found);
426 
427 /*
428  * Ceph dentry state
429  */
430 struct ceph_dentry_info {
431 	struct ceph_mds_session *lease_session;
432 	u32 lease_gen, lease_shared_gen;
433 	u32 lease_seq;
434 	unsigned long lease_renew_after, lease_renew_from;
435 	struct list_head lru;
436 	struct dentry *dentry;
437 	u64 time;
438 	u64 offset;
439 };
440 
441 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
442 {
443 	return (struct ceph_dentry_info *)dentry->d_fsdata;
444 }
445 
446 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
447 {
448 	return ((loff_t)frag << 32) | (loff_t)off;
449 }
450 
451 /*
452  * ino_t is <64 bits on many architectures, blech.
453  *
454  * don't include snap in ino hash, at least for now.
455  */
456 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
457 {
458 	ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
459 #if BITS_PER_LONG == 32
460 	ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
461 	if (!ino)
462 		ino = 1;
463 #endif
464 	return ino;
465 }
466 
467 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
468 {
469 	ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
470 	inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
471 	return 0;
472 }
473 
474 static inline struct ceph_vino ceph_vino(struct inode *inode)
475 {
476 	return ceph_inode(inode)->i_vino;
477 }
478 
479 /* for printf-style formatting */
480 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
481 
482 static inline u64 ceph_ino(struct inode *inode)
483 {
484 	return ceph_inode(inode)->i_vino.ino;
485 }
486 static inline u64 ceph_snap(struct inode *inode)
487 {
488 	return ceph_inode(inode)->i_vino.snap;
489 }
490 
491 static inline int ceph_ino_compare(struct inode *inode, void *data)
492 {
493 	struct ceph_vino *pvino = (struct ceph_vino *)data;
494 	struct ceph_inode_info *ci = ceph_inode(inode);
495 	return ci->i_vino.ino == pvino->ino &&
496 		ci->i_vino.snap == pvino->snap;
497 }
498 
499 static inline struct inode *ceph_find_inode(struct super_block *sb,
500 					    struct ceph_vino vino)
501 {
502 	ino_t t = ceph_vino_to_ino(vino);
503 	return ilookup5(sb, t, ceph_ino_compare, &vino);
504 }
505 
506 
507 /*
508  * caps helpers
509  */
510 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
511 {
512 	return !RB_EMPTY_ROOT(&ci->i_caps);
513 }
514 
515 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
516 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
517 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
518 				    struct ceph_cap *cap);
519 
520 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
521 {
522 	int issued;
523 	spin_lock(&ci->vfs_inode.i_lock);
524 	issued = __ceph_caps_issued(ci, NULL);
525 	spin_unlock(&ci->vfs_inode.i_lock);
526 	return issued;
527 }
528 
529 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
530 					int touch)
531 {
532 	int r;
533 	spin_lock(&ci->vfs_inode.i_lock);
534 	r = __ceph_caps_issued_mask(ci, mask, touch);
535 	spin_unlock(&ci->vfs_inode.i_lock);
536 	return r;
537 }
538 
539 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
540 {
541 	return ci->i_dirty_caps | ci->i_flushing_caps;
542 }
543 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
544 
545 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
546 extern int __ceph_caps_used(struct ceph_inode_info *ci);
547 
548 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
549 
550 /*
551  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
552  */
553 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
554 {
555 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
556 	if (w & CEPH_CAP_FILE_BUFFER)
557 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
558 	return w;
559 }
560 
561 /* what the mds thinks we want */
562 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
563 
564 extern void ceph_caps_init(void);
565 extern void ceph_caps_finalize(void);
566 extern void ceph_adjust_min_caps(int delta);
567 extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need);
568 extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx);
569 extern void ceph_reservation_status(struct ceph_client *client,
570 				    int *total, int *avail, int *used,
571 				    int *reserved, int *min);
572 
573 static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
574 {
575 	return (struct ceph_client *)inode->i_sb->s_fs_info;
576 }
577 
578 static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
579 {
580 	return (struct ceph_client *)sb->s_fs_info;
581 }
582 
583 
584 /*
585  * we keep buffered readdir results attached to file->private_data
586  */
587 struct ceph_file_info {
588 	int fmode;     /* initialized on open */
589 
590 	/* readdir: position within the dir */
591 	u32 frag;
592 	struct ceph_mds_request *last_readdir;
593 	int at_end;
594 
595 	/* readdir: position within a frag */
596 	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
597 	u64 next_offset;       /* offset of next chunk (last_name's + 1) */
598 	char *last_name;       /* last entry in previous chunk */
599 	struct dentry *dentry; /* next dentry (for dcache readdir) */
600 	unsigned long dir_release_count;
601 
602 	/* used for -o dirstat read() on directory thing */
603 	char *dir_info;
604 	int dir_info_len;
605 };
606 
607 
608 
609 /*
610  * snapshots
611  */
612 
613 /*
614  * A "snap context" is the set of existing snapshots when we
615  * write data.  It is used by the OSD to guide its COW behavior.
616  *
617  * The ceph_snap_context is refcounted, and attached to each dirty
618  * page, indicating which context the dirty data belonged when it was
619  * dirtied.
620  */
621 struct ceph_snap_context {
622 	atomic_t nref;
623 	u64 seq;
624 	int num_snaps;
625 	u64 snaps[];
626 };
627 
628 static inline struct ceph_snap_context *
629 ceph_get_snap_context(struct ceph_snap_context *sc)
630 {
631 	/*
632 	printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
633 	       atomic_read(&sc->nref)+1);
634 	*/
635 	if (sc)
636 		atomic_inc(&sc->nref);
637 	return sc;
638 }
639 
640 static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
641 {
642 	if (!sc)
643 		return;
644 	/*
645 	printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
646 	       atomic_read(&sc->nref)-1);
647 	*/
648 	if (atomic_dec_and_test(&sc->nref)) {
649 		/*printk(" deleting snap_context %p\n", sc);*/
650 		kfree(sc);
651 	}
652 }
653 
654 /*
655  * A "snap realm" describes a subset of the file hierarchy sharing
656  * the same set of snapshots that apply to it.  The realms themselves
657  * are organized into a hierarchy, such that children inherit (some of)
658  * the snapshots of their parents.
659  *
660  * All inodes within the realm that have capabilities are linked into a
661  * per-realm list.
662  */
663 struct ceph_snap_realm {
664 	u64 ino;
665 	atomic_t nref;
666 	struct rb_node node;
667 
668 	u64 created, seq;
669 	u64 parent_ino;
670 	u64 parent_since;   /* snapid when our current parent became so */
671 
672 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
673 	int num_prior_parent_snaps;   /*  had prior to parent_since */
674 	u64 *snaps;                   /* snaps specific to this realm */
675 	int num_snaps;
676 
677 	struct ceph_snap_realm *parent;
678 	struct list_head children;       /* list of child realms */
679 	struct list_head child_item;
680 
681 	struct list_head empty_item;     /* if i have ref==0 */
682 
683 	/* the current set of snaps for this realm */
684 	struct ceph_snap_context *cached_context;
685 
686 	struct list_head inodes_with_caps;
687 	spinlock_t inodes_with_caps_lock;
688 };
689 
690 
691 
692 /*
693  * calculate the number of pages a given length and offset map onto,
694  * if we align the data.
695  */
696 static inline int calc_pages_for(u64 off, u64 len)
697 {
698 	return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
699 		(off >> PAGE_CACHE_SHIFT);
700 }
701 
702 
703 
704 /* snap.c */
705 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
706 					       u64 ino);
707 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
708 				struct ceph_snap_realm *realm);
709 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
710 				struct ceph_snap_realm *realm);
711 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
712 				  void *p, void *e, bool deletion);
713 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
714 			     struct ceph_mds_session *session,
715 			     struct ceph_msg *msg);
716 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
717 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
718 				  struct ceph_cap_snap *capsnap);
719 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
720 
721 /*
722  * a cap_snap is "pending" if it is still awaiting an in-progress
723  * sync write (that may/may not still update size, mtime, etc.).
724  */
725 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
726 {
727 	return !list_empty(&ci->i_cap_snaps) &&
728 		list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
729 			   ci_item)->writing;
730 }
731 
732 
733 /* super.c */
734 extern struct kmem_cache *ceph_inode_cachep;
735 extern struct kmem_cache *ceph_cap_cachep;
736 extern struct kmem_cache *ceph_dentry_cachep;
737 extern struct kmem_cache *ceph_file_cachep;
738 
739 extern const char *ceph_msg_type_name(int type);
740 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
741 
742 #define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \
743 	"%02x%02x%02x%02x%02x%02x"
744 #define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \
745 		(f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7],    \
746 		(f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11],  \
747 		(f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15]
748 
749 /* inode.c */
750 extern const struct inode_operations ceph_file_iops;
751 
752 extern struct inode *ceph_alloc_inode(struct super_block *sb);
753 extern void ceph_destroy_inode(struct inode *inode);
754 
755 extern struct inode *ceph_get_inode(struct super_block *sb,
756 				    struct ceph_vino vino);
757 extern struct inode *ceph_get_snapdir(struct inode *parent);
758 extern int ceph_fill_file_size(struct inode *inode, int issued,
759 			       u32 truncate_seq, u64 truncate_size, u64 size);
760 extern void ceph_fill_file_time(struct inode *inode, int issued,
761 				u64 time_warp_seq, struct timespec *ctime,
762 				struct timespec *mtime, struct timespec *atime);
763 extern int ceph_fill_trace(struct super_block *sb,
764 			   struct ceph_mds_request *req,
765 			   struct ceph_mds_session *session);
766 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
767 				    struct ceph_mds_session *session);
768 
769 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
770 
771 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
772 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
773 extern void ceph_queue_vmtruncate(struct inode *inode);
774 
775 extern void ceph_queue_invalidate(struct inode *inode);
776 extern void ceph_queue_writeback(struct inode *inode);
777 
778 extern int ceph_do_getattr(struct inode *inode, int mask);
779 extern int ceph_permission(struct inode *inode, int mask);
780 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
781 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
782 			struct kstat *stat);
783 
784 /* xattr.c */
785 extern int ceph_setxattr(struct dentry *, const char *, const void *,
786 			 size_t, int);
787 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
788 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
789 extern int ceph_removexattr(struct dentry *, const char *);
790 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
791 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
792 
793 /* caps.c */
794 extern const char *ceph_cap_string(int c);
795 extern void ceph_handle_caps(struct ceph_mds_session *session,
796 			     struct ceph_msg *msg);
797 extern int ceph_add_cap(struct inode *inode,
798 			struct ceph_mds_session *session, u64 cap_id,
799 			int fmode, unsigned issued, unsigned wanted,
800 			unsigned cap, unsigned seq, u64 realmino, int flags,
801 			struct ceph_cap_reservation *caps_reservation);
802 extern void __ceph_remove_cap(struct ceph_cap *cap);
803 static inline void ceph_remove_cap(struct ceph_cap *cap)
804 {
805 	struct inode *inode = &cap->ci->vfs_inode;
806 	spin_lock(&inode->i_lock);
807 	__ceph_remove_cap(cap);
808 	spin_unlock(&inode->i_lock);
809 }
810 extern void ceph_put_cap(struct ceph_cap *cap);
811 
812 extern void ceph_queue_caps_release(struct inode *inode);
813 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
814 extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync);
815 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
816 				    struct ceph_mds_session *session);
817 extern int ceph_get_cap_mds(struct inode *inode);
818 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
819 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
820 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
821 				       struct ceph_snap_context *snapc);
822 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
823 			       struct ceph_mds_session **psession);
824 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
825 			    struct ceph_mds_session *session);
826 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
827 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
828 
829 extern int ceph_encode_inode_release(void **p, struct inode *inode,
830 				     int mds, int drop, int unless, int force);
831 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
832 				      int mds, int drop, int unless);
833 
834 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
835 			 int *got, loff_t endoff);
836 
837 /* for counting open files by mode */
838 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
839 {
840 	ci->i_nr_by_mode[mode]++;
841 }
842 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
843 
844 /* addr.c */
845 extern const struct address_space_operations ceph_aops;
846 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
847 
848 /* file.c */
849 extern const struct file_operations ceph_file_fops;
850 extern const struct address_space_operations ceph_aops;
851 extern int ceph_open(struct inode *inode, struct file *file);
852 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
853 				       struct nameidata *nd, int mode,
854 				       int locked_dir);
855 extern int ceph_release(struct inode *inode, struct file *filp);
856 extern void ceph_release_page_vector(struct page **pages, int num_pages);
857 
858 /* dir.c */
859 extern const struct file_operations ceph_dir_fops;
860 extern const struct inode_operations ceph_dir_iops;
861 extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
862 	ceph_snapdir_dentry_ops;
863 
864 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
865 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
866 					 struct dentry *dentry, int err);
867 
868 extern void ceph_dentry_lru_add(struct dentry *dn);
869 extern void ceph_dentry_lru_touch(struct dentry *dn);
870 extern void ceph_dentry_lru_del(struct dentry *dn);
871 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
872 
873 /*
874  * our d_ops vary depending on whether the inode is live,
875  * snapshotted (read-only), or a virtual ".snap" directory.
876  */
877 int ceph_init_dentry(struct dentry *dentry);
878 
879 
880 /* ioctl.c */
881 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
882 
883 /* export.c */
884 extern const struct export_operations ceph_export_ops;
885 
886 /* debugfs.c */
887 extern int ceph_debugfs_init(void);
888 extern void ceph_debugfs_cleanup(void);
889 extern int ceph_debugfs_client_init(struct ceph_client *client);
890 extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
891 
892 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
893 {
894 	if (dentry && dentry->d_parent)
895 		return dentry->d_parent->d_inode;
896 
897 	return NULL;
898 }
899 
900 #endif /* _FS_CEPH_SUPER_H */
901