xref: /linux/fs/ceph/super.h (revision ae00d4f37f4df56821331deb1028748110dd6dc9)
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  * Supported features
35  */
36 #define CEPH_FEATURE_SUPPORTED CEPH_FEATURE_NOSRCADDR | CEPH_FEATURE_FLOCK
37 #define CEPH_FEATURE_REQUIRED  CEPH_FEATURE_NOSRCADDR
38 
39 /*
40  * mount options
41  */
42 #define CEPH_OPT_FSID             (1<<0)
43 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
44 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
45 #define CEPH_OPT_DIRSTAT          (1<<4) /* funky `cat dirname` for stats */
46 #define CEPH_OPT_RBYTES           (1<<5) /* dir st_bytes = rbytes */
47 #define CEPH_OPT_NOCRC            (1<<6) /* no data crc on writes */
48 #define CEPH_OPT_NOASYNCREADDIR   (1<<7) /* no dcache readdir */
49 
50 #define CEPH_OPT_DEFAULT   (CEPH_OPT_RBYTES)
51 
52 #define ceph_set_opt(client, opt) \
53 	(client)->mount_args->flags |= CEPH_OPT_##opt;
54 #define ceph_test_opt(client, opt) \
55 	(!!((client)->mount_args->flags & CEPH_OPT_##opt))
56 
57 
58 struct ceph_mount_args {
59 	int sb_flags;
60 	int flags;
61 	struct ceph_fsid fsid;
62 	struct ceph_entity_addr my_addr;
63 	int num_mon;
64 	struct ceph_entity_addr *mon_addr;
65 	int mount_timeout;
66 	int osd_idle_ttl;
67 	int osd_timeout;
68 	int osd_keepalive_timeout;
69 	int wsize;
70 	int rsize;            /* max readahead */
71 	int congestion_kb;    /* max writeback in flight */
72 	int caps_wanted_delay_min, caps_wanted_delay_max;
73 	int cap_release_safety;
74 	int max_readdir;       /* max readdir result (entires) */
75 	int max_readdir_bytes; /* max readdir result (bytes) */
76 	char *snapdir_name;   /* default ".snap" */
77 	char *name;
78 	char *secret;
79 };
80 
81 /*
82  * defaults
83  */
84 #define CEPH_MOUNT_TIMEOUT_DEFAULT  60
85 #define CEPH_OSD_TIMEOUT_DEFAULT    60  /* seconds */
86 #define CEPH_OSD_KEEPALIVE_DEFAULT  5
87 #define CEPH_OSD_IDLE_TTL_DEFAULT    60
88 #define CEPH_MOUNT_RSIZE_DEFAULT    (512*1024) /* readahead */
89 #define CEPH_MAX_READDIR_DEFAULT    1024
90 #define CEPH_MAX_READDIR_BYTES_DEFAULT    (512*1024)
91 
92 #define CEPH_MSG_MAX_FRONT_LEN	(16*1024*1024)
93 #define CEPH_MSG_MAX_DATA_LEN	(16*1024*1024)
94 
95 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
96 #define CEPH_AUTH_NAME_DEFAULT   "guest"
97 /*
98  * Delay telling the MDS we no longer want caps, in case we reopen
99  * the file.  Delay a minimum amount of time, even if we send a cap
100  * message for some other reason.  Otherwise, take the oppotunity to
101  * update the mds to avoid sending another message later.
102  */
103 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
104 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
105 
106 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT        (CEPH_CAPS_PER_RELEASE * 4)
107 
108 /* mount state */
109 enum {
110 	CEPH_MOUNT_MOUNTING,
111 	CEPH_MOUNT_MOUNTED,
112 	CEPH_MOUNT_UNMOUNTING,
113 	CEPH_MOUNT_UNMOUNTED,
114 	CEPH_MOUNT_SHUTDOWN,
115 };
116 
117 /*
118  * subtract jiffies
119  */
120 static inline unsigned long time_sub(unsigned long a, unsigned long b)
121 {
122 	BUG_ON(time_after(b, a));
123 	return (long)a - (long)b;
124 }
125 
126 /*
127  * per-filesystem client state
128  *
129  * possibly shared by multiple mount points, if they are
130  * mounting the same ceph filesystem/cluster.
131  */
132 struct ceph_client {
133 	struct ceph_fsid fsid;
134 	bool have_fsid;
135 
136 	struct mutex mount_mutex;       /* serialize mount attempts */
137 	struct ceph_mount_args *mount_args;
138 
139 	struct super_block *sb;
140 
141 	unsigned long mount_state;
142 	wait_queue_head_t auth_wq;
143 
144 	int auth_err;
145 
146 	int min_caps;                  /* min caps i added */
147 
148 	struct ceph_messenger *msgr;   /* messenger instance */
149 	struct ceph_mon_client monc;
150 	struct ceph_mds_client mdsc;
151 	struct ceph_osd_client osdc;
152 
153 	/* writeback */
154 	mempool_t *wb_pagevec_pool;
155 	struct workqueue_struct *wb_wq;
156 	struct workqueue_struct *pg_inv_wq;
157 	struct workqueue_struct *trunc_wq;
158 	atomic_long_t writeback_count;
159 
160 	struct backing_dev_info backing_dev_info;
161 
162 #ifdef CONFIG_DEBUG_FS
163 	struct dentry *debugfs_monmap;
164 	struct dentry *debugfs_mdsmap, *debugfs_osdmap;
165 	struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
166 	struct dentry *debugfs_congestion_kb;
167 	struct dentry *debugfs_bdi;
168 #endif
169 };
170 
171 /*
172  * File i/o capability.  This tracks shared state with the metadata
173  * server that allows us to cache or writeback attributes or to read
174  * and write data.  For any given inode, we should have one or more
175  * capabilities, one issued by each metadata server, and our
176  * cumulative access is the OR of all issued capabilities.
177  *
178  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
179  * session capability lists.
180  */
181 struct ceph_cap {
182 	struct ceph_inode_info *ci;
183 	struct rb_node ci_node;          /* per-ci cap tree */
184 	struct ceph_mds_session *session;
185 	struct list_head session_caps;   /* per-session caplist */
186 	int mds;
187 	u64 cap_id;       /* unique cap id (mds provided) */
188 	int issued;       /* latest, from the mds */
189 	int implemented;  /* implemented superset of issued (for revocation) */
190 	int mds_wanted;
191 	u32 seq, issue_seq, mseq;
192 	u32 cap_gen;      /* active/stale cycle */
193 	unsigned long last_used;
194 	struct list_head caps_item;
195 };
196 
197 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
198 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
199 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
200 
201 /*
202  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
203  * we first complete any in-process sync writes and writeback any dirty
204  * data before flushing the snapped state (tracked here) back to the MDS.
205  */
206 struct ceph_cap_snap {
207 	atomic_t nref;
208 	struct ceph_inode_info *ci;
209 	struct list_head ci_item, flushing_item;
210 
211 	u64 follows, flush_tid;
212 	int issued, dirty;
213 	struct ceph_snap_context *context;
214 
215 	mode_t mode;
216 	uid_t uid;
217 	gid_t gid;
218 
219 	struct ceph_buffer *xattr_blob;
220 	u64 xattr_version;
221 
222 	u64 size;
223 	struct timespec mtime, atime, ctime;
224 	u64 time_warp_seq;
225 	int writing;   /* a sync write is still in progress */
226 	int dirty_pages;     /* dirty pages awaiting writeback */
227 };
228 
229 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
230 {
231 	if (atomic_dec_and_test(&capsnap->nref)) {
232 		if (capsnap->xattr_blob)
233 			ceph_buffer_put(capsnap->xattr_blob);
234 		kfree(capsnap);
235 	}
236 }
237 
238 /*
239  * The frag tree describes how a directory is fragmented, potentially across
240  * multiple metadata servers.  It is also used to indicate points where
241  * metadata authority is delegated, and whether/where metadata is replicated.
242  *
243  * A _leaf_ frag will be present in the i_fragtree IFF there is
244  * delegation info.  That is, if mds >= 0 || ndist > 0.
245  */
246 #define CEPH_MAX_DIRFRAG_REP 4
247 
248 struct ceph_inode_frag {
249 	struct rb_node node;
250 
251 	/* fragtree state */
252 	u32 frag;
253 	int split_by;         /* i.e. 2^(split_by) children */
254 
255 	/* delegation and replication info */
256 	int mds;              /* -1 if same authority as parent */
257 	int ndist;            /* >0 if replicated */
258 	int dist[CEPH_MAX_DIRFRAG_REP];
259 };
260 
261 /*
262  * We cache inode xattrs as an encoded blob until they are first used,
263  * at which point we parse them into an rbtree.
264  */
265 struct ceph_inode_xattr {
266 	struct rb_node node;
267 
268 	const char *name;
269 	int name_len;
270 	const char *val;
271 	int val_len;
272 	int dirty;
273 
274 	int should_free_name;
275 	int should_free_val;
276 };
277 
278 struct ceph_inode_xattrs_info {
279 	/*
280 	 * (still encoded) xattr blob. we avoid the overhead of parsing
281 	 * this until someone actually calls getxattr, etc.
282 	 *
283 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
284 	 * NULL means we don't know.
285 	*/
286 	struct ceph_buffer *blob, *prealloc_blob;
287 
288 	struct rb_root index;
289 	bool dirty;
290 	int count;
291 	int names_size;
292 	int vals_size;
293 	u64 version, index_version;
294 };
295 
296 /*
297  * Ceph inode.
298  */
299 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
300 #define CEPH_I_NODELAY   4  /* do not delay cap release */
301 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
302 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
303 
304 struct ceph_inode_info {
305 	struct ceph_vino i_vino;   /* ceph ino + snap */
306 
307 	u64 i_version;
308 	u32 i_time_warp_seq;
309 
310 	unsigned i_ceph_flags;
311 	unsigned long i_release_count;
312 
313 	struct ceph_file_layout i_layout;
314 	char *i_symlink;
315 
316 	/* for dirs */
317 	struct timespec i_rctime;
318 	u64 i_rbytes, i_rfiles, i_rsubdirs;
319 	u64 i_files, i_subdirs;
320 	u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
321 
322 	struct rb_root i_fragtree;
323 	struct mutex i_fragtree_mutex;
324 
325 	struct ceph_inode_xattrs_info i_xattrs;
326 
327 	/* capabilities.  protected _both_ by i_lock and cap->session's
328 	 * s_mutex. */
329 	struct rb_root i_caps;           /* cap list */
330 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
331 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
332 	struct list_head i_dirty_item, i_flushing_item;
333 	u64 i_cap_flush_seq;
334 	/* we need to track cap writeback on a per-cap-bit basis, to allow
335 	 * overlapping, pipelined cap flushes to the mds.  we can probably
336 	 * reduce the tid to 8 bits if we're concerned about inode size. */
337 	u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
338 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
339 	unsigned long i_hold_caps_min; /* jiffies */
340 	unsigned long i_hold_caps_max; /* jiffies */
341 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
342 	int i_cap_exporting_mds;         /* to handle cap migration between */
343 	unsigned i_cap_exporting_mseq;   /*  mds's. */
344 	unsigned i_cap_exporting_issued;
345 	struct ceph_cap_reservation i_cap_migration_resv;
346 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
347 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
348 						    dirty|flushing caps */
349 	unsigned i_snap_caps;           /* cap bits for snapped files */
350 
351 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
352 
353 	u32 i_truncate_seq;        /* last truncate to smaller size */
354 	u64 i_truncate_size;       /*  and the size we last truncated down to */
355 	int i_truncate_pending;    /*  still need to call vmtruncate */
356 
357 	u64 i_max_size;            /* max file size authorized by mds */
358 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
359 	u64 i_wanted_max_size;     /* offset we'd like to write too */
360 	u64 i_requested_max_size;  /* max_size we've requested */
361 
362 	/* held references to caps */
363 	int i_pin_ref;
364 	int i_rd_ref, i_rdcache_ref, i_wr_ref;
365 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
366 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
367 	u32 i_rdcache_gen;      /* we increment this each time we get
368 				   FILE_CACHE.  If it's non-zero, we
369 				   _may_ have cached pages. */
370 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
371 
372 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
373 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
374 	spinlock_t i_unsafe_lock;
375 
376 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
377 	int i_snap_realm_counter; /* snap realm (if caps) */
378 	struct list_head i_snap_realm_item;
379 	struct list_head i_snap_flush_item;
380 
381 	struct work_struct i_wb_work;  /* writeback work */
382 	struct work_struct i_pg_inv_work;  /* page invalidation work */
383 
384 	struct work_struct i_vmtruncate_work;
385 
386 	struct inode vfs_inode; /* at end */
387 };
388 
389 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
390 {
391 	return container_of(inode, struct ceph_inode_info, vfs_inode);
392 }
393 
394 static inline void ceph_i_clear(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 void ceph_i_set(struct inode *inode, unsigned mask)
404 {
405 	struct ceph_inode_info *ci = ceph_inode(inode);
406 
407 	spin_lock(&inode->i_lock);
408 	ci->i_ceph_flags |= mask;
409 	spin_unlock(&inode->i_lock);
410 }
411 
412 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
413 {
414 	struct ceph_inode_info *ci = ceph_inode(inode);
415 	bool r;
416 
417 	smp_mb();
418 	r = (ci->i_ceph_flags & mask) == mask;
419 	return r;
420 }
421 
422 
423 /* find a specific frag @f */
424 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
425 						u32 f);
426 
427 /*
428  * choose fragment for value @v.  copy frag content to pfrag, if leaf
429  * exists
430  */
431 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
432 			    struct ceph_inode_frag *pfrag,
433 			    int *found);
434 
435 /*
436  * Ceph dentry state
437  */
438 struct ceph_dentry_info {
439 	struct ceph_mds_session *lease_session;
440 	u32 lease_gen, lease_shared_gen;
441 	u32 lease_seq;
442 	unsigned long lease_renew_after, lease_renew_from;
443 	struct list_head lru;
444 	struct dentry *dentry;
445 	u64 time;
446 	u64 offset;
447 };
448 
449 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
450 {
451 	return (struct ceph_dentry_info *)dentry->d_fsdata;
452 }
453 
454 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
455 {
456 	return ((loff_t)frag << 32) | (loff_t)off;
457 }
458 
459 /*
460  * ino_t is <64 bits on many architectures, blech.
461  *
462  * don't include snap in ino hash, at least for now.
463  */
464 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
465 {
466 	ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
467 #if BITS_PER_LONG == 32
468 	ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
469 	if (!ino)
470 		ino = 1;
471 #endif
472 	return ino;
473 }
474 
475 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
476 {
477 	ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
478 	inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
479 	return 0;
480 }
481 
482 static inline struct ceph_vino ceph_vino(struct inode *inode)
483 {
484 	return ceph_inode(inode)->i_vino;
485 }
486 
487 /* for printf-style formatting */
488 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
489 
490 static inline u64 ceph_ino(struct inode *inode)
491 {
492 	return ceph_inode(inode)->i_vino.ino;
493 }
494 static inline u64 ceph_snap(struct inode *inode)
495 {
496 	return ceph_inode(inode)->i_vino.snap;
497 }
498 
499 static inline int ceph_ino_compare(struct inode *inode, void *data)
500 {
501 	struct ceph_vino *pvino = (struct ceph_vino *)data;
502 	struct ceph_inode_info *ci = ceph_inode(inode);
503 	return ci->i_vino.ino == pvino->ino &&
504 		ci->i_vino.snap == pvino->snap;
505 }
506 
507 static inline struct inode *ceph_find_inode(struct super_block *sb,
508 					    struct ceph_vino vino)
509 {
510 	ino_t t = ceph_vino_to_ino(vino);
511 	return ilookup5(sb, t, ceph_ino_compare, &vino);
512 }
513 
514 
515 /*
516  * caps helpers
517  */
518 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
519 {
520 	return !RB_EMPTY_ROOT(&ci->i_caps);
521 }
522 
523 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
524 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
525 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
526 				    struct ceph_cap *cap);
527 
528 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
529 {
530 	int issued;
531 	spin_lock(&ci->vfs_inode.i_lock);
532 	issued = __ceph_caps_issued(ci, NULL);
533 	spin_unlock(&ci->vfs_inode.i_lock);
534 	return issued;
535 }
536 
537 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
538 					int touch)
539 {
540 	int r;
541 	spin_lock(&ci->vfs_inode.i_lock);
542 	r = __ceph_caps_issued_mask(ci, mask, touch);
543 	spin_unlock(&ci->vfs_inode.i_lock);
544 	return r;
545 }
546 
547 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
548 {
549 	return ci->i_dirty_caps | ci->i_flushing_caps;
550 }
551 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
552 
553 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
554 extern int __ceph_caps_used(struct ceph_inode_info *ci);
555 
556 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
557 
558 /*
559  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
560  */
561 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
562 {
563 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
564 	if (w & CEPH_CAP_FILE_BUFFER)
565 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
566 	return w;
567 }
568 
569 /* what the mds thinks we want */
570 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
571 
572 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
573 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
574 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
575 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
576 			     struct ceph_cap_reservation *ctx, int need);
577 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
578 			       struct ceph_cap_reservation *ctx);
579 extern void ceph_reservation_status(struct ceph_client *client,
580 				    int *total, int *avail, int *used,
581 				    int *reserved, int *min);
582 
583 static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
584 {
585 	return (struct ceph_client *)inode->i_sb->s_fs_info;
586 }
587 
588 static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
589 {
590 	return (struct ceph_client *)sb->s_fs_info;
591 }
592 
593 
594 /*
595  * we keep buffered readdir results attached to file->private_data
596  */
597 struct ceph_file_info {
598 	int fmode;     /* initialized on open */
599 
600 	/* readdir: position within the dir */
601 	u32 frag;
602 	struct ceph_mds_request *last_readdir;
603 	int at_end;
604 
605 	/* readdir: position within a frag */
606 	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
607 	u64 next_offset;       /* offset of next chunk (last_name's + 1) */
608 	char *last_name;       /* last entry in previous chunk */
609 	struct dentry *dentry; /* next dentry (for dcache readdir) */
610 	unsigned long dir_release_count;
611 
612 	/* used for -o dirstat read() on directory thing */
613 	char *dir_info;
614 	int dir_info_len;
615 };
616 
617 
618 
619 /*
620  * snapshots
621  */
622 
623 /*
624  * A "snap context" is the set of existing snapshots when we
625  * write data.  It is used by the OSD to guide its COW behavior.
626  *
627  * The ceph_snap_context is refcounted, and attached to each dirty
628  * page, indicating which context the dirty data belonged when it was
629  * dirtied.
630  */
631 struct ceph_snap_context {
632 	atomic_t nref;
633 	u64 seq;
634 	int num_snaps;
635 	u64 snaps[];
636 };
637 
638 static inline struct ceph_snap_context *
639 ceph_get_snap_context(struct ceph_snap_context *sc)
640 {
641 	/*
642 	printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
643 	       atomic_read(&sc->nref)+1);
644 	*/
645 	if (sc)
646 		atomic_inc(&sc->nref);
647 	return sc;
648 }
649 
650 static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
651 {
652 	if (!sc)
653 		return;
654 	/*
655 	printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
656 	       atomic_read(&sc->nref)-1);
657 	*/
658 	if (atomic_dec_and_test(&sc->nref)) {
659 		/*printk(" deleting snap_context %p\n", sc);*/
660 		kfree(sc);
661 	}
662 }
663 
664 /*
665  * A "snap realm" describes a subset of the file hierarchy sharing
666  * the same set of snapshots that apply to it.  The realms themselves
667  * are organized into a hierarchy, such that children inherit (some of)
668  * the snapshots of their parents.
669  *
670  * All inodes within the realm that have capabilities are linked into a
671  * per-realm list.
672  */
673 struct ceph_snap_realm {
674 	u64 ino;
675 	atomic_t nref;
676 	struct rb_node node;
677 
678 	u64 created, seq;
679 	u64 parent_ino;
680 	u64 parent_since;   /* snapid when our current parent became so */
681 
682 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
683 	int num_prior_parent_snaps;   /*  had prior to parent_since */
684 	u64 *snaps;                   /* snaps specific to this realm */
685 	int num_snaps;
686 
687 	struct ceph_snap_realm *parent;
688 	struct list_head children;       /* list of child realms */
689 	struct list_head child_item;
690 
691 	struct list_head empty_item;     /* if i have ref==0 */
692 
693 	struct list_head dirty_item;     /* if realm needs new context */
694 
695 	/* the current set of snaps for this realm */
696 	struct ceph_snap_context *cached_context;
697 
698 	struct list_head inodes_with_caps;
699 	spinlock_t inodes_with_caps_lock;
700 };
701 
702 
703 
704 /*
705  * calculate the number of pages a given length and offset map onto,
706  * if we align the data.
707  */
708 static inline int calc_pages_for(u64 off, u64 len)
709 {
710 	return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
711 		(off >> PAGE_CACHE_SHIFT);
712 }
713 
714 
715 
716 /* snap.c */
717 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
718 					       u64 ino);
719 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
720 				struct ceph_snap_realm *realm);
721 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
722 				struct ceph_snap_realm *realm);
723 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
724 				  void *p, void *e, bool deletion);
725 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
726 			     struct ceph_mds_session *session,
727 			     struct ceph_msg *msg);
728 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
729 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
730 				  struct ceph_cap_snap *capsnap);
731 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
732 
733 /*
734  * a cap_snap is "pending" if it is still awaiting an in-progress
735  * sync write (that may/may not still update size, mtime, etc.).
736  */
737 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
738 {
739 	return !list_empty(&ci->i_cap_snaps) &&
740 		list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
741 			   ci_item)->writing;
742 }
743 
744 
745 /* super.c */
746 extern struct kmem_cache *ceph_inode_cachep;
747 extern struct kmem_cache *ceph_cap_cachep;
748 extern struct kmem_cache *ceph_dentry_cachep;
749 extern struct kmem_cache *ceph_file_cachep;
750 
751 extern const char *ceph_msg_type_name(int type);
752 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid);
753 
754 /* inode.c */
755 extern const struct inode_operations ceph_file_iops;
756 
757 extern struct inode *ceph_alloc_inode(struct super_block *sb);
758 extern void ceph_destroy_inode(struct inode *inode);
759 
760 extern struct inode *ceph_get_inode(struct super_block *sb,
761 				    struct ceph_vino vino);
762 extern struct inode *ceph_get_snapdir(struct inode *parent);
763 extern int ceph_fill_file_size(struct inode *inode, int issued,
764 			       u32 truncate_seq, u64 truncate_size, u64 size);
765 extern void ceph_fill_file_time(struct inode *inode, int issued,
766 				u64 time_warp_seq, struct timespec *ctime,
767 				struct timespec *mtime, struct timespec *atime);
768 extern int ceph_fill_trace(struct super_block *sb,
769 			   struct ceph_mds_request *req,
770 			   struct ceph_mds_session *session);
771 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
772 				    struct ceph_mds_session *session);
773 
774 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
775 
776 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
777 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
778 extern void ceph_queue_vmtruncate(struct inode *inode);
779 
780 extern void ceph_queue_invalidate(struct inode *inode);
781 extern void ceph_queue_writeback(struct inode *inode);
782 
783 extern int ceph_do_getattr(struct inode *inode, int mask);
784 extern int ceph_permission(struct inode *inode, int mask);
785 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
786 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
787 			struct kstat *stat);
788 
789 /* xattr.c */
790 extern int ceph_setxattr(struct dentry *, const char *, const void *,
791 			 size_t, int);
792 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
793 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
794 extern int ceph_removexattr(struct dentry *, const char *);
795 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
796 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
797 
798 /* caps.c */
799 extern const char *ceph_cap_string(int c);
800 extern void ceph_handle_caps(struct ceph_mds_session *session,
801 			     struct ceph_msg *msg);
802 extern int ceph_add_cap(struct inode *inode,
803 			struct ceph_mds_session *session, u64 cap_id,
804 			int fmode, unsigned issued, unsigned wanted,
805 			unsigned cap, unsigned seq, u64 realmino, int flags,
806 			struct ceph_cap_reservation *caps_reservation);
807 extern void __ceph_remove_cap(struct ceph_cap *cap);
808 static inline void ceph_remove_cap(struct ceph_cap *cap)
809 {
810 	struct inode *inode = &cap->ci->vfs_inode;
811 	spin_lock(&inode->i_lock);
812 	__ceph_remove_cap(cap);
813 	spin_unlock(&inode->i_lock);
814 }
815 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
816 			 struct ceph_cap *cap);
817 
818 extern void ceph_queue_caps_release(struct inode *inode);
819 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
820 extern int ceph_fsync(struct file *file, int datasync);
821 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
822 				    struct ceph_mds_session *session);
823 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
824 					     int mds);
825 extern int ceph_get_cap_mds(struct inode *inode);
826 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
827 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
828 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
829 				       struct ceph_snap_context *snapc);
830 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
831 			       struct ceph_mds_session **psession);
832 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
833 			    struct ceph_mds_session *session);
834 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
835 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
836 
837 extern int ceph_encode_inode_release(void **p, struct inode *inode,
838 				     int mds, int drop, int unless, int force);
839 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
840 				      int mds, int drop, int unless);
841 
842 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
843 			 int *got, loff_t endoff);
844 
845 /* for counting open files by mode */
846 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
847 {
848 	ci->i_nr_by_mode[mode]++;
849 }
850 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
851 
852 /* addr.c */
853 extern const struct address_space_operations ceph_aops;
854 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
855 
856 /* file.c */
857 extern const struct file_operations ceph_file_fops;
858 extern const struct address_space_operations ceph_aops;
859 extern int ceph_open(struct inode *inode, struct file *file);
860 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
861 				       struct nameidata *nd, int mode,
862 				       int locked_dir);
863 extern int ceph_release(struct inode *inode, struct file *filp);
864 extern void ceph_release_page_vector(struct page **pages, int num_pages);
865 
866 /* dir.c */
867 extern const struct file_operations ceph_dir_fops;
868 extern const struct inode_operations ceph_dir_iops;
869 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
870 	ceph_snapdir_dentry_ops;
871 
872 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
873 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
874 					 struct dentry *dentry, int err);
875 
876 extern void ceph_dentry_lru_add(struct dentry *dn);
877 extern void ceph_dentry_lru_touch(struct dentry *dn);
878 extern void ceph_dentry_lru_del(struct dentry *dn);
879 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
880 
881 /*
882  * our d_ops vary depending on whether the inode is live,
883  * snapshotted (read-only), or a virtual ".snap" directory.
884  */
885 int ceph_init_dentry(struct dentry *dentry);
886 
887 
888 /* ioctl.c */
889 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
890 
891 /* export.c */
892 extern const struct export_operations ceph_export_ops;
893 
894 /* debugfs.c */
895 extern int ceph_debugfs_init(void);
896 extern void ceph_debugfs_cleanup(void);
897 extern int ceph_debugfs_client_init(struct ceph_client *client);
898 extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
899 
900 /* locks.c */
901 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
902 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
903 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
904 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
905 			     int p_locks, int f_locks);
906 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
907 
908 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
909 {
910 	if (dentry && dentry->d_parent)
911 		return dentry->d_parent->d_inode;
912 
913 	return NULL;
914 }
915 
916 #endif /* _FS_CEPH_SUPER_H */
917