xref: /openbmc/linux/fs/ceph/super.h (revision 270384cc)
1 #ifndef _FS_CEPH_SUPER_H
2 #define _FS_CEPH_SUPER_H
3 
4 #include <linux/ceph/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 #include <linux/posix_acl.h>
17 
18 #include <linux/ceph/libceph.h>
19 
20 #ifdef CONFIG_CEPH_FSCACHE
21 #include <linux/fscache.h>
22 #endif
23 
24 /* f_type in struct statfs */
25 #define CEPH_SUPER_MAGIC 0x00c36400
26 
27 /* large granularity for statfs utilization stats to facilitate
28  * large volume sizes on 32-bit machines. */
29 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
30 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
31 
32 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
33 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
34 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
35 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
36 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
37 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
38 
39 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES)
40 
41 #define ceph_set_mount_opt(fsc, opt) \
42 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
43 #define ceph_test_mount_opt(fsc, opt) \
44 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
45 
46 #define CEPH_RSIZE_DEFAULT             0           /* max read size */
47 #define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
48 #define CEPH_MAX_READDIR_DEFAULT        1024
49 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
50 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
51 
52 struct ceph_mount_options {
53 	int flags;
54 	int sb_flags;
55 
56 	int wsize;            /* max write size */
57 	int rsize;            /* max read size */
58 	int rasize;           /* max readahead */
59 	int congestion_kb;    /* max writeback in flight */
60 	int caps_wanted_delay_min, caps_wanted_delay_max;
61 	int cap_release_safety;
62 	int max_readdir;       /* max readdir result (entires) */
63 	int max_readdir_bytes; /* max readdir result (bytes) */
64 
65 	/*
66 	 * everything above this point can be memcmp'd; everything below
67 	 * is handled in compare_mount_options()
68 	 */
69 
70 	char *snapdir_name;   /* default ".snap" */
71 };
72 
73 struct ceph_fs_client {
74 	struct super_block *sb;
75 
76 	struct ceph_mount_options *mount_options;
77 	struct ceph_client *client;
78 
79 	unsigned long mount_state;
80 	int min_caps;                  /* min caps i added */
81 
82 	struct ceph_mds_client *mdsc;
83 
84 	/* writeback */
85 	mempool_t *wb_pagevec_pool;
86 	struct workqueue_struct *wb_wq;
87 	struct workqueue_struct *pg_inv_wq;
88 	struct workqueue_struct *trunc_wq;
89 	atomic_long_t writeback_count;
90 
91 	struct backing_dev_info backing_dev_info;
92 
93 #ifdef CONFIG_DEBUG_FS
94 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
95 	struct dentry *debugfs_congestion_kb;
96 	struct dentry *debugfs_bdi;
97 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
98 	struct dentry *debugfs_mds_sessions;
99 #endif
100 
101 #ifdef CONFIG_CEPH_FSCACHE
102 	struct fscache_cookie *fscache;
103 	struct workqueue_struct *revalidate_wq;
104 #endif
105 };
106 
107 
108 /*
109  * File i/o capability.  This tracks shared state with the metadata
110  * server that allows us to cache or writeback attributes or to read
111  * and write data.  For any given inode, we should have one or more
112  * capabilities, one issued by each metadata server, and our
113  * cumulative access is the OR of all issued capabilities.
114  *
115  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
116  * session capability lists.
117  */
118 struct ceph_cap {
119 	struct ceph_inode_info *ci;
120 	struct rb_node ci_node;          /* per-ci cap tree */
121 	struct ceph_mds_session *session;
122 	struct list_head session_caps;   /* per-session caplist */
123 	int mds;
124 	u64 cap_id;       /* unique cap id (mds provided) */
125 	int issued;       /* latest, from the mds */
126 	int implemented;  /* implemented superset of issued (for revocation) */
127 	int mds_wanted;
128 	u32 seq, issue_seq, mseq;
129 	u32 cap_gen;      /* active/stale cycle */
130 	unsigned long last_used;
131 	struct list_head caps_item;
132 };
133 
134 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
135 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
136 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
137 
138 /*
139  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
140  * we first complete any in-process sync writes and writeback any dirty
141  * data before flushing the snapped state (tracked here) back to the MDS.
142  */
143 struct ceph_cap_snap {
144 	atomic_t nref;
145 	struct ceph_inode_info *ci;
146 	struct list_head ci_item, flushing_item;
147 
148 	u64 follows, flush_tid;
149 	int issued, dirty;
150 	struct ceph_snap_context *context;
151 
152 	umode_t mode;
153 	kuid_t uid;
154 	kgid_t gid;
155 
156 	struct ceph_buffer *xattr_blob;
157 	u64 xattr_version;
158 
159 	u64 size;
160 	struct timespec mtime, atime, ctime;
161 	u64 time_warp_seq;
162 	int writing;   /* a sync write is still in progress */
163 	int dirty_pages;     /* dirty pages awaiting writeback */
164 	bool inline_data;
165 };
166 
167 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
168 {
169 	if (atomic_dec_and_test(&capsnap->nref)) {
170 		if (capsnap->xattr_blob)
171 			ceph_buffer_put(capsnap->xattr_blob);
172 		kfree(capsnap);
173 	}
174 }
175 
176 /*
177  * The frag tree describes how a directory is fragmented, potentially across
178  * multiple metadata servers.  It is also used to indicate points where
179  * metadata authority is delegated, and whether/where metadata is replicated.
180  *
181  * A _leaf_ frag will be present in the i_fragtree IFF there is
182  * delegation info.  That is, if mds >= 0 || ndist > 0.
183  */
184 #define CEPH_MAX_DIRFRAG_REP 4
185 
186 struct ceph_inode_frag {
187 	struct rb_node node;
188 
189 	/* fragtree state */
190 	u32 frag;
191 	int split_by;         /* i.e. 2^(split_by) children */
192 
193 	/* delegation and replication info */
194 	int mds;              /* -1 if same authority as parent */
195 	int ndist;            /* >0 if replicated */
196 	int dist[CEPH_MAX_DIRFRAG_REP];
197 };
198 
199 /*
200  * We cache inode xattrs as an encoded blob until they are first used,
201  * at which point we parse them into an rbtree.
202  */
203 struct ceph_inode_xattr {
204 	struct rb_node node;
205 
206 	const char *name;
207 	int name_len;
208 	const char *val;
209 	int val_len;
210 	int dirty;
211 
212 	int should_free_name;
213 	int should_free_val;
214 };
215 
216 /*
217  * Ceph dentry state
218  */
219 struct ceph_dentry_info {
220 	struct ceph_mds_session *lease_session;
221 	u32 lease_gen, lease_shared_gen;
222 	u32 lease_seq;
223 	unsigned long lease_renew_after, lease_renew_from;
224 	struct list_head lru;
225 	struct dentry *dentry;
226 	u64 time;
227 	u64 offset;
228 };
229 
230 struct ceph_inode_xattrs_info {
231 	/*
232 	 * (still encoded) xattr blob. we avoid the overhead of parsing
233 	 * this until someone actually calls getxattr, etc.
234 	 *
235 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
236 	 * NULL means we don't know.
237 	*/
238 	struct ceph_buffer *blob, *prealloc_blob;
239 
240 	struct rb_root index;
241 	bool dirty;
242 	int count;
243 	int names_size;
244 	int vals_size;
245 	u64 version, index_version;
246 };
247 
248 /*
249  * Ceph inode.
250  */
251 struct ceph_inode_info {
252 	struct ceph_vino i_vino;   /* ceph ino + snap */
253 
254 	spinlock_t i_ceph_lock;
255 
256 	u64 i_version;
257 	u64 i_inline_version;
258 	u32 i_time_warp_seq;
259 
260 	unsigned i_ceph_flags;
261 	int i_ordered_count;
262 	atomic_t i_release_count;
263 	atomic_t i_complete_count;
264 
265 	struct ceph_dir_layout i_dir_layout;
266 	struct ceph_file_layout i_layout;
267 	char *i_symlink;
268 
269 	/* for dirs */
270 	struct timespec i_rctime;
271 	u64 i_rbytes, i_rfiles, i_rsubdirs;
272 	u64 i_files, i_subdirs;
273 
274 	struct rb_root i_fragtree;
275 	struct mutex i_fragtree_mutex;
276 
277 	struct ceph_inode_xattrs_info i_xattrs;
278 
279 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
280 	 * s_mutex. */
281 	struct rb_root i_caps;           /* cap list */
282 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
283 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
284 	struct list_head i_dirty_item, i_flushing_item;
285 	u64 i_cap_flush_seq;
286 	/* we need to track cap writeback on a per-cap-bit basis, to allow
287 	 * overlapping, pipelined cap flushes to the mds.  we can probably
288 	 * reduce the tid to 8 bits if we're concerned about inode size. */
289 	u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
290 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
291 	unsigned long i_hold_caps_min; /* jiffies */
292 	unsigned long i_hold_caps_max; /* jiffies */
293 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
294 	struct ceph_cap_reservation i_cap_migration_resv;
295 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
296 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
297 						    dirty|flushing caps */
298 	unsigned i_snap_caps;           /* cap bits for snapped files */
299 
300 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
301 
302 	struct mutex i_truncate_mutex;
303 	u32 i_truncate_seq;        /* last truncate to smaller size */
304 	u64 i_truncate_size;       /*  and the size we last truncated down to */
305 	int i_truncate_pending;    /*  still need to call vmtruncate */
306 
307 	u64 i_max_size;            /* max file size authorized by mds */
308 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
309 	u64 i_wanted_max_size;     /* offset we'd like to write too */
310 	u64 i_requested_max_size;  /* max_size we've requested */
311 
312 	/* held references to caps */
313 	int i_pin_ref;
314 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
315 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
316 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
317 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
318 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
319 
320 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
321 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
322 	spinlock_t i_unsafe_lock;
323 
324 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
325 	int i_snap_realm_counter; /* snap realm (if caps) */
326 	struct list_head i_snap_realm_item;
327 	struct list_head i_snap_flush_item;
328 
329 	struct work_struct i_wb_work;  /* writeback work */
330 	struct work_struct i_pg_inv_work;  /* page invalidation work */
331 
332 	struct work_struct i_vmtruncate_work;
333 
334 #ifdef CONFIG_CEPH_FSCACHE
335 	struct fscache_cookie *fscache;
336 	u32 i_fscache_gen; /* sequence, for delayed fscache validate */
337 	struct work_struct i_revalidate_work;
338 #endif
339 	struct inode vfs_inode; /* at end */
340 };
341 
342 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
343 {
344 	return container_of(inode, struct ceph_inode_info, vfs_inode);
345 }
346 
347 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
348 {
349 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
350 }
351 
352 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
353 {
354 	return (struct ceph_fs_client *)sb->s_fs_info;
355 }
356 
357 static inline struct ceph_vino ceph_vino(struct inode *inode)
358 {
359 	return ceph_inode(inode)->i_vino;
360 }
361 
362 /*
363  * ino_t is <64 bits on many architectures, blech.
364  *
365  *               i_ino (kernel inode)   st_ino (userspace)
366  * i386          32                     32
367  * x86_64+ino32  64                     32
368  * x86_64        64                     64
369  */
370 static inline u32 ceph_ino_to_ino32(__u64 vino)
371 {
372 	u32 ino = vino & 0xffffffff;
373 	ino ^= vino >> 32;
374 	if (!ino)
375 		ino = 2;
376 	return ino;
377 }
378 
379 /*
380  * kernel i_ino value
381  */
382 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
383 {
384 #if BITS_PER_LONG == 32
385 	return ceph_ino_to_ino32(vino.ino);
386 #else
387 	return (ino_t)vino.ino;
388 #endif
389 }
390 
391 /*
392  * user-visible ino (stat, filldir)
393  */
394 #if BITS_PER_LONG == 32
395 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
396 {
397 	return ino;
398 }
399 #else
400 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
401 {
402 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
403 		ino = ceph_ino_to_ino32(ino);
404 	return ino;
405 }
406 #endif
407 
408 
409 /* for printf-style formatting */
410 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
411 
412 static inline u64 ceph_ino(struct inode *inode)
413 {
414 	return ceph_inode(inode)->i_vino.ino;
415 }
416 static inline u64 ceph_snap(struct inode *inode)
417 {
418 	return ceph_inode(inode)->i_vino.snap;
419 }
420 
421 static inline int ceph_ino_compare(struct inode *inode, void *data)
422 {
423 	struct ceph_vino *pvino = (struct ceph_vino *)data;
424 	struct ceph_inode_info *ci = ceph_inode(inode);
425 	return ci->i_vino.ino == pvino->ino &&
426 		ci->i_vino.snap == pvino->snap;
427 }
428 
429 static inline struct inode *ceph_find_inode(struct super_block *sb,
430 					    struct ceph_vino vino)
431 {
432 	ino_t t = ceph_vino_to_ino(vino);
433 	return ilookup5(sb, t, ceph_ino_compare, &vino);
434 }
435 
436 
437 /*
438  * Ceph inode.
439  */
440 #define CEPH_I_DIR_ORDERED	1  /* dentries in dir are ordered */
441 #define CEPH_I_NODELAY		4  /* do not delay cap release */
442 #define CEPH_I_FLUSH		8  /* do not delay flush of dirty metadata */
443 #define CEPH_I_NOFLUSH		16 /* do not flush dirty caps */
444 
445 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
446 					   int release_count, int ordered_count)
447 {
448 	atomic_set(&ci->i_complete_count, release_count);
449 	if (ci->i_ordered_count == ordered_count)
450 		ci->i_ceph_flags |= CEPH_I_DIR_ORDERED;
451 	else
452 		ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED;
453 }
454 
455 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
456 {
457 	atomic_inc(&ci->i_release_count);
458 }
459 
460 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
461 {
462 	return atomic_read(&ci->i_complete_count) ==
463 		atomic_read(&ci->i_release_count);
464 }
465 
466 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
467 {
468 	return __ceph_dir_is_complete(ci) &&
469 		(ci->i_ceph_flags & CEPH_I_DIR_ORDERED);
470 }
471 
472 static inline void ceph_dir_clear_complete(struct inode *inode)
473 {
474 	__ceph_dir_clear_complete(ceph_inode(inode));
475 }
476 
477 static inline void ceph_dir_clear_ordered(struct inode *inode)
478 {
479 	struct ceph_inode_info *ci = ceph_inode(inode);
480 	spin_lock(&ci->i_ceph_lock);
481 	ci->i_ordered_count++;
482 	ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED;
483 	spin_unlock(&ci->i_ceph_lock);
484 }
485 
486 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
487 {
488 	struct ceph_inode_info *ci = ceph_inode(inode);
489 	bool ret;
490 	spin_lock(&ci->i_ceph_lock);
491 	ret = __ceph_dir_is_complete_ordered(ci);
492 	spin_unlock(&ci->i_ceph_lock);
493 	return ret;
494 }
495 
496 /* find a specific frag @f */
497 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
498 						u32 f);
499 
500 /*
501  * choose fragment for value @v.  copy frag content to pfrag, if leaf
502  * exists
503  */
504 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
505 			    struct ceph_inode_frag *pfrag,
506 			    int *found);
507 
508 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
509 {
510 	return (struct ceph_dentry_info *)dentry->d_fsdata;
511 }
512 
513 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
514 {
515 	return ((loff_t)frag << 32) | (loff_t)off;
516 }
517 
518 /*
519  * caps helpers
520  */
521 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
522 {
523 	return !RB_EMPTY_ROOT(&ci->i_caps);
524 }
525 
526 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
527 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
528 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
529 				    struct ceph_cap *cap);
530 
531 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
532 {
533 	int issued;
534 	spin_lock(&ci->i_ceph_lock);
535 	issued = __ceph_caps_issued(ci, NULL);
536 	spin_unlock(&ci->i_ceph_lock);
537 	return issued;
538 }
539 
540 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
541 					int touch)
542 {
543 	int r;
544 	spin_lock(&ci->i_ceph_lock);
545 	r = __ceph_caps_issued_mask(ci, mask, touch);
546 	spin_unlock(&ci->i_ceph_lock);
547 	return r;
548 }
549 
550 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
551 {
552 	return ci->i_dirty_caps | ci->i_flushing_caps;
553 }
554 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
555 
556 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
557 				      struct ceph_cap *ocap, int mask);
558 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
559 extern int __ceph_caps_used(struct ceph_inode_info *ci);
560 
561 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
562 
563 /*
564  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
565  */
566 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
567 {
568 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
569 	if (w & CEPH_CAP_FILE_BUFFER)
570 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
571 	return w;
572 }
573 
574 /* what the mds thinks we want */
575 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
576 
577 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
578 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
579 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
580 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
581 			     struct ceph_cap_reservation *ctx, int need);
582 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
583 			       struct ceph_cap_reservation *ctx);
584 extern void ceph_reservation_status(struct ceph_fs_client *client,
585 				    int *total, int *avail, int *used,
586 				    int *reserved, int *min);
587 
588 
589 
590 /*
591  * we keep buffered readdir results attached to file->private_data
592  */
593 #define CEPH_F_SYNC     1
594 #define CEPH_F_ATEND    2
595 
596 struct ceph_file_info {
597 	short fmode;     /* initialized on open */
598 	short flags;     /* CEPH_F_* */
599 
600 	/* readdir: position within the dir */
601 	u32 frag;
602 	struct ceph_mds_request *last_readdir;
603 
604 	/* readdir: position within a frag */
605 	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
606 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
607 	char *last_name;       /* last entry in previous chunk */
608 	struct dentry *dentry; /* next dentry (for dcache readdir) */
609 	int dir_release_count;
610 	int dir_ordered_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  * A "snap realm" describes a subset of the file hierarchy sharing
621  * the same set of snapshots that apply to it.  The realms themselves
622  * are organized into a hierarchy, such that children inherit (some of)
623  * the snapshots of their parents.
624  *
625  * All inodes within the realm that have capabilities are linked into a
626  * per-realm list.
627  */
628 struct ceph_snap_realm {
629 	u64 ino;
630 	atomic_t nref;
631 	struct rb_node node;
632 
633 	u64 created, seq;
634 	u64 parent_ino;
635 	u64 parent_since;   /* snapid when our current parent became so */
636 
637 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
638 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
639 	u64 *snaps;                   /* snaps specific to this realm */
640 	u32 num_snaps;
641 
642 	struct ceph_snap_realm *parent;
643 	struct list_head children;       /* list of child realms */
644 	struct list_head child_item;
645 
646 	struct list_head empty_item;     /* if i have ref==0 */
647 
648 	struct list_head dirty_item;     /* if realm needs new context */
649 
650 	/* the current set of snaps for this realm */
651 	struct ceph_snap_context *cached_context;
652 
653 	struct list_head inodes_with_caps;
654 	spinlock_t inodes_with_caps_lock;
655 };
656 
657 static inline int default_congestion_kb(void)
658 {
659 	int congestion_kb;
660 
661 	/*
662 	 * Copied from NFS
663 	 *
664 	 * congestion size, scale with available memory.
665 	 *
666 	 *  64MB:    8192k
667 	 * 128MB:   11585k
668 	 * 256MB:   16384k
669 	 * 512MB:   23170k
670 	 *   1GB:   32768k
671 	 *   2GB:   46340k
672 	 *   4GB:   65536k
673 	 *   8GB:   92681k
674 	 *  16GB:  131072k
675 	 *
676 	 * This allows larger machines to have larger/more transfers.
677 	 * Limit the default to 256M
678 	 */
679 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
680 	if (congestion_kb > 256*1024)
681 		congestion_kb = 256*1024;
682 
683 	return congestion_kb;
684 }
685 
686 
687 
688 /* snap.c */
689 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
690 					       u64 ino);
691 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
692 				struct ceph_snap_realm *realm);
693 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
694 				struct ceph_snap_realm *realm);
695 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
696 				  void *p, void *e, bool deletion);
697 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
698 			     struct ceph_mds_session *session,
699 			     struct ceph_msg *msg);
700 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
701 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
702 				  struct ceph_cap_snap *capsnap);
703 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
704 extern int ceph_snap_init(void);
705 extern void ceph_snap_exit(void);
706 
707 /*
708  * a cap_snap is "pending" if it is still awaiting an in-progress
709  * sync write (that may/may not still update size, mtime, etc.).
710  */
711 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
712 {
713 	return !list_empty(&ci->i_cap_snaps) &&
714 		list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
715 			   ci_item)->writing;
716 }
717 
718 /* inode.c */
719 extern const struct inode_operations ceph_file_iops;
720 
721 extern struct inode *ceph_alloc_inode(struct super_block *sb);
722 extern void ceph_destroy_inode(struct inode *inode);
723 extern int ceph_drop_inode(struct inode *inode);
724 
725 extern struct inode *ceph_get_inode(struct super_block *sb,
726 				    struct ceph_vino vino);
727 extern struct inode *ceph_get_snapdir(struct inode *parent);
728 extern int ceph_fill_file_size(struct inode *inode, int issued,
729 			       u32 truncate_seq, u64 truncate_size, u64 size);
730 extern void ceph_fill_file_time(struct inode *inode, int issued,
731 				u64 time_warp_seq, struct timespec *ctime,
732 				struct timespec *mtime, struct timespec *atime);
733 extern int ceph_fill_trace(struct super_block *sb,
734 			   struct ceph_mds_request *req,
735 			   struct ceph_mds_session *session);
736 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
737 				    struct ceph_mds_session *session);
738 
739 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
740 
741 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
742 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
743 extern void ceph_queue_vmtruncate(struct inode *inode);
744 
745 extern void ceph_queue_invalidate(struct inode *inode);
746 extern void ceph_queue_writeback(struct inode *inode);
747 
748 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
749 			     int mask, bool force);
750 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
751 {
752 	return __ceph_do_getattr(inode, NULL, mask, force);
753 }
754 extern int ceph_permission(struct inode *inode, int mask);
755 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
756 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
757 			struct kstat *stat);
758 
759 /* xattr.c */
760 extern int ceph_setxattr(struct dentry *, const char *, const void *,
761 			 size_t, int);
762 int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int);
763 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
764 int __ceph_removexattr(struct dentry *, const char *);
765 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
766 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
767 extern int ceph_removexattr(struct dentry *, const char *);
768 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
769 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
770 extern void __init ceph_xattr_init(void);
771 extern void ceph_xattr_exit(void);
772 extern const struct xattr_handler *ceph_xattr_handlers[];
773 
774 /* acl.c */
775 struct ceph_acls_info {
776 	void *default_acl;
777 	void *acl;
778 	struct ceph_pagelist *pagelist;
779 };
780 
781 #ifdef CONFIG_CEPH_FS_POSIX_ACL
782 
783 struct posix_acl *ceph_get_acl(struct inode *, int);
784 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
785 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
786 		       struct ceph_acls_info *info);
787 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
788 void ceph_release_acls_info(struct ceph_acls_info *info);
789 
790 static inline void ceph_forget_all_cached_acls(struct inode *inode)
791 {
792        forget_all_cached_acls(inode);
793 }
794 
795 #else
796 
797 #define ceph_get_acl NULL
798 #define ceph_set_acl NULL
799 
800 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
801 				     struct ceph_acls_info *info)
802 {
803 	return 0;
804 }
805 static inline void ceph_init_inode_acls(struct inode *inode,
806 					struct ceph_acls_info *info)
807 {
808 }
809 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
810 {
811 }
812 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
813 {
814 	return 0;
815 }
816 
817 static inline void ceph_forget_all_cached_acls(struct inode *inode)
818 {
819 }
820 
821 #endif
822 
823 /* caps.c */
824 extern const char *ceph_cap_string(int c);
825 extern void ceph_handle_caps(struct ceph_mds_session *session,
826 			     struct ceph_msg *msg);
827 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
828 				     struct ceph_cap_reservation *ctx);
829 extern void ceph_add_cap(struct inode *inode,
830 			 struct ceph_mds_session *session, u64 cap_id,
831 			 int fmode, unsigned issued, unsigned wanted,
832 			 unsigned cap, unsigned seq, u64 realmino, int flags,
833 			 struct ceph_cap **new_cap);
834 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
835 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
836 			 struct ceph_cap *cap);
837 extern int ceph_is_any_caps(struct inode *inode);
838 
839 extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino,
840 				u64 cap_id, u32 migrate_seq, u32 issue_seq);
841 extern void ceph_queue_caps_release(struct inode *inode);
842 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
843 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
844 		      int datasync);
845 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
846 				    struct ceph_mds_session *session);
847 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
848 					     int mds);
849 extern int ceph_get_cap_mds(struct inode *inode);
850 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
851 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
852 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
853 				       struct ceph_snap_context *snapc);
854 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
855 			       struct ceph_mds_session **psession,
856 			       int again);
857 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
858 			    struct ceph_mds_session *session);
859 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
860 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
861 
862 extern int ceph_encode_inode_release(void **p, struct inode *inode,
863 				     int mds, int drop, int unless, int force);
864 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
865 				      int mds, int drop, int unless);
866 
867 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
868 			 loff_t endoff, int *got, struct page **pinned_page);
869 
870 /* for counting open files by mode */
871 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
872 {
873 	ci->i_nr_by_mode[mode]++;
874 }
875 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
876 
877 /* addr.c */
878 extern const struct address_space_operations ceph_aops;
879 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
880 
881 /* file.c */
882 extern const struct file_operations ceph_file_fops;
883 extern const struct address_space_operations ceph_aops;
884 
885 extern int ceph_open(struct inode *inode, struct file *file);
886 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
887 			    struct file *file, unsigned flags, umode_t mode,
888 			    int *opened);
889 extern int ceph_release(struct inode *inode, struct file *filp);
890 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
891 				  char *data, size_t len);
892 int ceph_uninline_data(struct file *filp, struct page *locked_page);
893 /* dir.c */
894 extern const struct file_operations ceph_dir_fops;
895 extern const struct inode_operations ceph_dir_iops;
896 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
897 	ceph_snapdir_dentry_ops;
898 
899 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
900 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
901 			       struct dentry *dentry, int err);
902 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
903 					 struct dentry *dentry, int err);
904 
905 extern void ceph_dentry_lru_add(struct dentry *dn);
906 extern void ceph_dentry_lru_touch(struct dentry *dn);
907 extern void ceph_dentry_lru_del(struct dentry *dn);
908 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
909 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
910 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
911 
912 /*
913  * our d_ops vary depending on whether the inode is live,
914  * snapshotted (read-only), or a virtual ".snap" directory.
915  */
916 int ceph_init_dentry(struct dentry *dentry);
917 
918 
919 /* ioctl.c */
920 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
921 
922 /* export.c */
923 extern const struct export_operations ceph_export_ops;
924 
925 /* locks.c */
926 extern __init void ceph_flock_init(void);
927 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
928 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
929 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
930 extern int ceph_encode_locks_to_buffer(struct inode *inode,
931 				       struct ceph_filelock *flocks,
932 				       int num_fcntl_locks,
933 				       int num_flock_locks);
934 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
935 				  struct ceph_pagelist *pagelist,
936 				  int num_fcntl_locks, int num_flock_locks);
937 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
938 
939 /* debugfs.c */
940 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
941 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
942 
943 #endif /* _FS_CEPH_SUPER_H */
944