xref: /linux/fs/ceph/super.h (revision e5c86679d5e864947a52fb31e45a425dea3e7fa9)
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 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
39 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
40 
41 #define CEPH_MOUNT_OPT_DEFAULT    CEPH_MOUNT_OPT_DCACHE
42 
43 #define ceph_set_mount_opt(fsc, opt) \
44 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
45 #define ceph_test_mount_opt(fsc, opt) \
46 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
47 
48 #define CEPH_RSIZE_DEFAULT              (64*1024*1024) /* max read size */
49 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
50 #define CEPH_MAX_READDIR_DEFAULT        1024
51 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
52 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
53 
54 struct ceph_mount_options {
55 	int flags;
56 	int sb_flags;
57 
58 	int wsize;            /* max write size */
59 	int rsize;            /* max read size */
60 	int rasize;           /* max readahead */
61 	int congestion_kb;    /* max writeback in flight */
62 	int caps_wanted_delay_min, caps_wanted_delay_max;
63 	int cap_release_safety;
64 	int max_readdir;       /* max readdir result (entires) */
65 	int max_readdir_bytes; /* max readdir result (bytes) */
66 
67 	/*
68 	 * everything above this point can be memcmp'd; everything below
69 	 * is handled in compare_mount_options()
70 	 */
71 
72 	char *snapdir_name;   /* default ".snap" */
73 	char *mds_namespace;  /* default NULL */
74 	char *server_path;    /* default  "/" */
75 };
76 
77 struct ceph_fs_client {
78 	struct super_block *sb;
79 
80 	struct ceph_mount_options *mount_options;
81 	struct ceph_client *client;
82 
83 	unsigned long mount_state;
84 	int min_caps;                  /* min caps i added */
85 
86 	struct ceph_mds_client *mdsc;
87 
88 	/* writeback */
89 	mempool_t *wb_pagevec_pool;
90 	struct workqueue_struct *wb_wq;
91 	struct workqueue_struct *pg_inv_wq;
92 	struct workqueue_struct *trunc_wq;
93 	atomic_long_t writeback_count;
94 
95 	struct backing_dev_info backing_dev_info;
96 
97 #ifdef CONFIG_DEBUG_FS
98 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
99 	struct dentry *debugfs_congestion_kb;
100 	struct dentry *debugfs_bdi;
101 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
102 	struct dentry *debugfs_mds_sessions;
103 #endif
104 
105 #ifdef CONFIG_CEPH_FSCACHE
106 	struct fscache_cookie *fscache;
107 #endif
108 };
109 
110 
111 /*
112  * File i/o capability.  This tracks shared state with the metadata
113  * server that allows us to cache or writeback attributes or to read
114  * and write data.  For any given inode, we should have one or more
115  * capabilities, one issued by each metadata server, and our
116  * cumulative access is the OR of all issued capabilities.
117  *
118  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
119  * session capability lists.
120  */
121 struct ceph_cap {
122 	struct ceph_inode_info *ci;
123 	struct rb_node ci_node;          /* per-ci cap tree */
124 	struct ceph_mds_session *session;
125 	struct list_head session_caps;   /* per-session caplist */
126 	u64 cap_id;       /* unique cap id (mds provided) */
127 	union {
128 		/* in-use caps */
129 		struct {
130 			int issued;       /* latest, from the mds */
131 			int implemented;  /* implemented superset of
132 					     issued (for revocation) */
133 			int mds, mds_wanted;
134 		};
135 		/* caps to release */
136 		struct {
137 			u64 cap_ino;
138 			int queue_release;
139 		};
140 	};
141 	u32 seq, issue_seq, mseq;
142 	u32 cap_gen;      /* active/stale cycle */
143 	unsigned long last_used;
144 	struct list_head caps_item;
145 };
146 
147 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
148 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
149 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
150 
151 struct ceph_cap_flush {
152 	u64 tid;
153 	int caps; /* 0 means capsnap */
154 	bool wake; /* wake up flush waiters when finish ? */
155 	struct list_head g_list; // global
156 	struct list_head i_list; // per inode
157 };
158 
159 /*
160  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
161  * we first complete any in-process sync writes and writeback any dirty
162  * data before flushing the snapped state (tracked here) back to the MDS.
163  */
164 struct ceph_cap_snap {
165 	atomic_t nref;
166 	struct list_head ci_item;
167 
168 	struct ceph_cap_flush cap_flush;
169 
170 	u64 follows;
171 	int issued, dirty;
172 	struct ceph_snap_context *context;
173 
174 	umode_t mode;
175 	kuid_t uid;
176 	kgid_t gid;
177 
178 	struct ceph_buffer *xattr_blob;
179 	u64 xattr_version;
180 
181 	u64 size;
182 	struct timespec mtime, atime, ctime;
183 	u64 time_warp_seq;
184 	u64 truncate_size;
185 	u32 truncate_seq;
186 	int writing;   /* a sync write is still in progress */
187 	int dirty_pages;     /* dirty pages awaiting writeback */
188 	bool inline_data;
189 	bool need_flush;
190 };
191 
192 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
193 {
194 	if (atomic_dec_and_test(&capsnap->nref)) {
195 		if (capsnap->xattr_blob)
196 			ceph_buffer_put(capsnap->xattr_blob);
197 		kfree(capsnap);
198 	}
199 }
200 
201 /*
202  * The frag tree describes how a directory is fragmented, potentially across
203  * multiple metadata servers.  It is also used to indicate points where
204  * metadata authority is delegated, and whether/where metadata is replicated.
205  *
206  * A _leaf_ frag will be present in the i_fragtree IFF there is
207  * delegation info.  That is, if mds >= 0 || ndist > 0.
208  */
209 #define CEPH_MAX_DIRFRAG_REP 4
210 
211 struct ceph_inode_frag {
212 	struct rb_node node;
213 
214 	/* fragtree state */
215 	u32 frag;
216 	int split_by;         /* i.e. 2^(split_by) children */
217 
218 	/* delegation and replication info */
219 	int mds;              /* -1 if same authority as parent */
220 	int ndist;            /* >0 if replicated */
221 	int dist[CEPH_MAX_DIRFRAG_REP];
222 };
223 
224 /*
225  * We cache inode xattrs as an encoded blob until they are first used,
226  * at which point we parse them into an rbtree.
227  */
228 struct ceph_inode_xattr {
229 	struct rb_node node;
230 
231 	const char *name;
232 	int name_len;
233 	const char *val;
234 	int val_len;
235 	int dirty;
236 
237 	int should_free_name;
238 	int should_free_val;
239 };
240 
241 /*
242  * Ceph dentry state
243  */
244 struct ceph_dentry_info {
245 	struct ceph_mds_session *lease_session;
246 	u32 lease_gen, lease_shared_gen;
247 	u32 lease_seq;
248 	unsigned long lease_renew_after, lease_renew_from;
249 	struct list_head lru;
250 	struct dentry *dentry;
251 	unsigned long time;
252 	u64 offset;
253 };
254 
255 struct ceph_inode_xattrs_info {
256 	/*
257 	 * (still encoded) xattr blob. we avoid the overhead of parsing
258 	 * this until someone actually calls getxattr, etc.
259 	 *
260 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
261 	 * NULL means we don't know.
262 	*/
263 	struct ceph_buffer *blob, *prealloc_blob;
264 
265 	struct rb_root index;
266 	bool dirty;
267 	int count;
268 	int names_size;
269 	int vals_size;
270 	u64 version, index_version;
271 };
272 
273 /*
274  * Ceph inode.
275  */
276 struct ceph_inode_info {
277 	struct ceph_vino i_vino;   /* ceph ino + snap */
278 
279 	spinlock_t i_ceph_lock;
280 
281 	u64 i_version;
282 	u64 i_inline_version;
283 	u32 i_time_warp_seq;
284 
285 	unsigned i_ceph_flags;
286 	atomic64_t i_release_count;
287 	atomic64_t i_ordered_count;
288 	atomic64_t i_complete_seq[2];
289 
290 	struct ceph_dir_layout i_dir_layout;
291 	struct ceph_file_layout i_layout;
292 	char *i_symlink;
293 
294 	/* for dirs */
295 	struct timespec i_rctime;
296 	u64 i_rbytes, i_rfiles, i_rsubdirs;
297 	u64 i_files, i_subdirs;
298 
299 	struct rb_root i_fragtree;
300 	int i_fragtree_nsplits;
301 	struct mutex i_fragtree_mutex;
302 
303 	struct ceph_inode_xattrs_info i_xattrs;
304 
305 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
306 	 * s_mutex. */
307 	struct rb_root i_caps;           /* cap list */
308 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
309 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
310 	struct list_head i_dirty_item, i_flushing_item;
311 	/* we need to track cap writeback on a per-cap-bit basis, to allow
312 	 * overlapping, pipelined cap flushes to the mds.  we can probably
313 	 * reduce the tid to 8 bits if we're concerned about inode size. */
314 	struct ceph_cap_flush *i_prealloc_cap_flush;
315 	struct list_head i_cap_flush_list;
316 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
317 	unsigned long i_hold_caps_min; /* jiffies */
318 	unsigned long i_hold_caps_max; /* jiffies */
319 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
320 	struct ceph_cap_reservation i_cap_migration_resv;
321 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
322 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
323 						    dirty|flushing caps */
324 	unsigned i_snap_caps;           /* cap bits for snapped files */
325 
326 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
327 
328 	struct mutex i_truncate_mutex;
329 	u32 i_truncate_seq;        /* last truncate to smaller size */
330 	u64 i_truncate_size;       /*  and the size we last truncated down to */
331 	int i_truncate_pending;    /*  still need to call vmtruncate */
332 
333 	u64 i_max_size;            /* max file size authorized by mds */
334 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
335 	u64 i_wanted_max_size;     /* offset we'd like to write too */
336 	u64 i_requested_max_size;  /* max_size we've requested */
337 
338 	/* held references to caps */
339 	int i_pin_ref;
340 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
341 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
342 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
343 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
344 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
345 
346 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
347 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
348 	spinlock_t i_unsafe_lock;
349 
350 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
351 	int i_snap_realm_counter; /* snap realm (if caps) */
352 	struct list_head i_snap_realm_item;
353 	struct list_head i_snap_flush_item;
354 
355 	struct work_struct i_wb_work;  /* writeback work */
356 	struct work_struct i_pg_inv_work;  /* page invalidation work */
357 
358 	struct work_struct i_vmtruncate_work;
359 
360 #ifdef CONFIG_CEPH_FSCACHE
361 	struct fscache_cookie *fscache;
362 	u32 i_fscache_gen;
363 #endif
364 	struct inode vfs_inode; /* at end */
365 };
366 
367 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
368 {
369 	return container_of(inode, struct ceph_inode_info, vfs_inode);
370 }
371 
372 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
373 {
374 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
375 }
376 
377 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
378 {
379 	return (struct ceph_fs_client *)sb->s_fs_info;
380 }
381 
382 static inline struct ceph_vino ceph_vino(struct inode *inode)
383 {
384 	return ceph_inode(inode)->i_vino;
385 }
386 
387 /*
388  * ino_t is <64 bits on many architectures, blech.
389  *
390  *               i_ino (kernel inode)   st_ino (userspace)
391  * i386          32                     32
392  * x86_64+ino32  64                     32
393  * x86_64        64                     64
394  */
395 static inline u32 ceph_ino_to_ino32(__u64 vino)
396 {
397 	u32 ino = vino & 0xffffffff;
398 	ino ^= vino >> 32;
399 	if (!ino)
400 		ino = 2;
401 	return ino;
402 }
403 
404 /*
405  * kernel i_ino value
406  */
407 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
408 {
409 #if BITS_PER_LONG == 32
410 	return ceph_ino_to_ino32(vino.ino);
411 #else
412 	return (ino_t)vino.ino;
413 #endif
414 }
415 
416 /*
417  * user-visible ino (stat, filldir)
418  */
419 #if BITS_PER_LONG == 32
420 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
421 {
422 	return ino;
423 }
424 #else
425 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
426 {
427 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
428 		ino = ceph_ino_to_ino32(ino);
429 	return ino;
430 }
431 #endif
432 
433 
434 /* for printf-style formatting */
435 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
436 
437 static inline u64 ceph_ino(struct inode *inode)
438 {
439 	return ceph_inode(inode)->i_vino.ino;
440 }
441 static inline u64 ceph_snap(struct inode *inode)
442 {
443 	return ceph_inode(inode)->i_vino.snap;
444 }
445 
446 static inline int ceph_ino_compare(struct inode *inode, void *data)
447 {
448 	struct ceph_vino *pvino = (struct ceph_vino *)data;
449 	struct ceph_inode_info *ci = ceph_inode(inode);
450 	return ci->i_vino.ino == pvino->ino &&
451 		ci->i_vino.snap == pvino->snap;
452 }
453 
454 static inline struct inode *ceph_find_inode(struct super_block *sb,
455 					    struct ceph_vino vino)
456 {
457 	ino_t t = ceph_vino_to_ino(vino);
458 	return ilookup5(sb, t, ceph_ino_compare, &vino);
459 }
460 
461 
462 /*
463  * Ceph inode.
464  */
465 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
466 #define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
467 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
468 #define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
469 #define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
470 #define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
471 #define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
472 #define CEPH_I_SEC_INITED	(1 << 7)  /* security initialized */
473 #define CEPH_I_CAP_DROPPED	(1 << 8)  /* caps were forcibly dropped */
474 #define CEPH_I_KICK_FLUSH	(1 << 9)  /* kick flushing caps */
475 #define CEPH_I_FLUSH_SNAPS	(1 << 10) /* need flush snapss */
476 
477 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
478 					   long long release_count,
479 					   long long ordered_count)
480 {
481 	smp_mb__before_atomic();
482 	atomic64_set(&ci->i_complete_seq[0], release_count);
483 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
484 }
485 
486 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
487 {
488 	atomic64_inc(&ci->i_release_count);
489 }
490 
491 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
492 {
493 	atomic64_inc(&ci->i_ordered_count);
494 }
495 
496 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
497 {
498 	return atomic64_read(&ci->i_complete_seq[0]) ==
499 		atomic64_read(&ci->i_release_count);
500 }
501 
502 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
503 {
504 	return  atomic64_read(&ci->i_complete_seq[0]) ==
505 		atomic64_read(&ci->i_release_count) &&
506 		atomic64_read(&ci->i_complete_seq[1]) ==
507 		atomic64_read(&ci->i_ordered_count);
508 }
509 
510 static inline void ceph_dir_clear_complete(struct inode *inode)
511 {
512 	__ceph_dir_clear_complete(ceph_inode(inode));
513 }
514 
515 static inline void ceph_dir_clear_ordered(struct inode *inode)
516 {
517 	__ceph_dir_clear_ordered(ceph_inode(inode));
518 }
519 
520 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
521 {
522 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
523 	smp_rmb();
524 	return ret;
525 }
526 
527 /* find a specific frag @f */
528 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
529 						u32 f);
530 
531 /*
532  * choose fragment for value @v.  copy frag content to pfrag, if leaf
533  * exists
534  */
535 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
536 			    struct ceph_inode_frag *pfrag,
537 			    int *found);
538 
539 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
540 {
541 	return (struct ceph_dentry_info *)dentry->d_fsdata;
542 }
543 
544 /*
545  * caps helpers
546  */
547 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
548 {
549 	return !RB_EMPTY_ROOT(&ci->i_caps);
550 }
551 
552 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
553 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
554 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
555 				    struct ceph_cap *cap);
556 
557 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
558 {
559 	int issued;
560 	spin_lock(&ci->i_ceph_lock);
561 	issued = __ceph_caps_issued(ci, NULL);
562 	spin_unlock(&ci->i_ceph_lock);
563 	return issued;
564 }
565 
566 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
567 					int touch)
568 {
569 	int r;
570 	spin_lock(&ci->i_ceph_lock);
571 	r = __ceph_caps_issued_mask(ci, mask, touch);
572 	spin_unlock(&ci->i_ceph_lock);
573 	return r;
574 }
575 
576 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
577 {
578 	return ci->i_dirty_caps | ci->i_flushing_caps;
579 }
580 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
581 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
582 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
583 				  struct ceph_cap_flush **pcf);
584 
585 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
586 				      struct ceph_cap *ocap, int mask);
587 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
588 extern int __ceph_caps_used(struct ceph_inode_info *ci);
589 
590 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
591 
592 /*
593  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
594  */
595 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
596 {
597 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
598 	if (w & CEPH_CAP_FILE_BUFFER)
599 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
600 	return w;
601 }
602 
603 /* what the mds thinks we want */
604 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
605 
606 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
607 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
608 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
609 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
610 			     struct ceph_cap_reservation *ctx, int need);
611 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
612 			       struct ceph_cap_reservation *ctx);
613 extern void ceph_reservation_status(struct ceph_fs_client *client,
614 				    int *total, int *avail, int *used,
615 				    int *reserved, int *min);
616 
617 
618 
619 /*
620  * we keep buffered readdir results attached to file->private_data
621  */
622 #define CEPH_F_SYNC     1
623 #define CEPH_F_ATEND    2
624 
625 struct ceph_file_info {
626 	short fmode;     /* initialized on open */
627 	short flags;     /* CEPH_F_* */
628 
629 	/* readdir: position within the dir */
630 	u32 frag;
631 	struct ceph_mds_request *last_readdir;
632 
633 	/* readdir: position within a frag */
634 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
635 	char *last_name;       /* last entry in previous chunk */
636 	long long dir_release_count;
637 	long long dir_ordered_count;
638 	int readdir_cache_idx;
639 
640 	/* used for -o dirstat read() on directory thing */
641 	char *dir_info;
642 	int dir_info_len;
643 };
644 
645 struct ceph_readdir_cache_control {
646 	struct page  *page;
647 	struct dentry **dentries;
648 	int index;
649 };
650 
651 /*
652  * A "snap realm" describes a subset of the file hierarchy sharing
653  * the same set of snapshots that apply to it.  The realms themselves
654  * are organized into a hierarchy, such that children inherit (some of)
655  * the snapshots of their parents.
656  *
657  * All inodes within the realm that have capabilities are linked into a
658  * per-realm list.
659  */
660 struct ceph_snap_realm {
661 	u64 ino;
662 	atomic_t nref;
663 	struct rb_node node;
664 
665 	u64 created, seq;
666 	u64 parent_ino;
667 	u64 parent_since;   /* snapid when our current parent became so */
668 
669 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
670 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
671 	u64 *snaps;                   /* snaps specific to this realm */
672 	u32 num_snaps;
673 
674 	struct ceph_snap_realm *parent;
675 	struct list_head children;       /* list of child realms */
676 	struct list_head child_item;
677 
678 	struct list_head empty_item;     /* if i have ref==0 */
679 
680 	struct list_head dirty_item;     /* if realm needs new context */
681 
682 	/* the current set of snaps for this realm */
683 	struct ceph_snap_context *cached_context;
684 
685 	struct list_head inodes_with_caps;
686 	spinlock_t inodes_with_caps_lock;
687 };
688 
689 static inline int default_congestion_kb(void)
690 {
691 	int congestion_kb;
692 
693 	/*
694 	 * Copied from NFS
695 	 *
696 	 * congestion size, scale with available memory.
697 	 *
698 	 *  64MB:    8192k
699 	 * 128MB:   11585k
700 	 * 256MB:   16384k
701 	 * 512MB:   23170k
702 	 *   1GB:   32768k
703 	 *   2GB:   46340k
704 	 *   4GB:   65536k
705 	 *   8GB:   92681k
706 	 *  16GB:  131072k
707 	 *
708 	 * This allows larger machines to have larger/more transfers.
709 	 * Limit the default to 256M
710 	 */
711 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
712 	if (congestion_kb > 256*1024)
713 		congestion_kb = 256*1024;
714 
715 	return congestion_kb;
716 }
717 
718 
719 
720 /* snap.c */
721 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
722 					       u64 ino);
723 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
724 				struct ceph_snap_realm *realm);
725 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
726 				struct ceph_snap_realm *realm);
727 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
728 				  void *p, void *e, bool deletion,
729 				  struct ceph_snap_realm **realm_ret);
730 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
731 			     struct ceph_mds_session *session,
732 			     struct ceph_msg *msg);
733 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
734 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
735 				  struct ceph_cap_snap *capsnap);
736 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
737 
738 /*
739  * a cap_snap is "pending" if it is still awaiting an in-progress
740  * sync write (that may/may not still update size, mtime, etc.).
741  */
742 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
743 {
744 	return !list_empty(&ci->i_cap_snaps) &&
745 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
746 			       ci_item)->writing;
747 }
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 extern int ceph_drop_inode(struct inode *inode);
755 
756 extern struct inode *ceph_get_inode(struct super_block *sb,
757 				    struct ceph_vino vino);
758 extern struct inode *ceph_get_snapdir(struct inode *parent);
759 extern int ceph_fill_file_size(struct inode *inode, int issued,
760 			       u32 truncate_seq, u64 truncate_size, u64 size);
761 extern void ceph_fill_file_time(struct inode *inode, int issued,
762 				u64 time_warp_seq, struct timespec *ctime,
763 				struct timespec *mtime, struct timespec *atime);
764 extern int ceph_fill_trace(struct super_block *sb,
765 			   struct ceph_mds_request *req);
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, struct page *locked_page,
779 			     int mask, bool force);
780 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
781 {
782 	return __ceph_do_getattr(inode, NULL, mask, force);
783 }
784 extern int ceph_permission(struct inode *inode, int mask);
785 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
786 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
787 extern int ceph_getattr(const struct path *path, struct kstat *stat,
788 			u32 request_mask, unsigned int flags);
789 
790 /* xattr.c */
791 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
792 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
793 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
794 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
795 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
796 extern void __init ceph_xattr_init(void);
797 extern void ceph_xattr_exit(void);
798 extern const struct xattr_handler *ceph_xattr_handlers[];
799 
800 #ifdef CONFIG_SECURITY
801 extern bool ceph_security_xattr_deadlock(struct inode *in);
802 extern bool ceph_security_xattr_wanted(struct inode *in);
803 #else
804 static inline bool ceph_security_xattr_deadlock(struct inode *in)
805 {
806 	return false;
807 }
808 static inline bool ceph_security_xattr_wanted(struct inode *in)
809 {
810 	return false;
811 }
812 #endif
813 
814 /* acl.c */
815 struct ceph_acls_info {
816 	void *default_acl;
817 	void *acl;
818 	struct ceph_pagelist *pagelist;
819 };
820 
821 #ifdef CONFIG_CEPH_FS_POSIX_ACL
822 
823 struct posix_acl *ceph_get_acl(struct inode *, int);
824 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
825 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
826 		       struct ceph_acls_info *info);
827 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
828 void ceph_release_acls_info(struct ceph_acls_info *info);
829 
830 static inline void ceph_forget_all_cached_acls(struct inode *inode)
831 {
832        forget_all_cached_acls(inode);
833 }
834 
835 #else
836 
837 #define ceph_get_acl NULL
838 #define ceph_set_acl NULL
839 
840 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
841 				     struct ceph_acls_info *info)
842 {
843 	return 0;
844 }
845 static inline void ceph_init_inode_acls(struct inode *inode,
846 					struct ceph_acls_info *info)
847 {
848 }
849 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
850 {
851 }
852 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
853 {
854 	return 0;
855 }
856 
857 static inline void ceph_forget_all_cached_acls(struct inode *inode)
858 {
859 }
860 
861 #endif
862 
863 /* caps.c */
864 extern const char *ceph_cap_string(int c);
865 extern void ceph_handle_caps(struct ceph_mds_session *session,
866 			     struct ceph_msg *msg);
867 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
868 				     struct ceph_cap_reservation *ctx);
869 extern void ceph_add_cap(struct inode *inode,
870 			 struct ceph_mds_session *session, u64 cap_id,
871 			 int fmode, unsigned issued, unsigned wanted,
872 			 unsigned cap, unsigned seq, u64 realmino, int flags,
873 			 struct ceph_cap **new_cap);
874 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
875 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
876 			 struct ceph_cap *cap);
877 extern int ceph_is_any_caps(struct inode *inode);
878 
879 extern void ceph_queue_caps_release(struct inode *inode);
880 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
881 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
882 		      int datasync);
883 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
884 					  struct ceph_mds_session *session);
885 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
886 				    struct ceph_mds_session *session);
887 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
888 					     int mds);
889 extern int ceph_get_cap_mds(struct inode *inode);
890 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
891 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
892 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
893 				       struct ceph_snap_context *snapc);
894 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
895 			     struct ceph_mds_session **psession);
896 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
897 			    struct ceph_mds_session *session);
898 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
899 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
900 
901 extern int ceph_encode_inode_release(void **p, struct inode *inode,
902 				     int mds, int drop, int unless, int force);
903 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
904 				      struct inode *dir,
905 				      int mds, int drop, int unless);
906 
907 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
908 			 loff_t endoff, int *got, struct page **pinned_page);
909 extern int ceph_try_get_caps(struct ceph_inode_info *ci,
910 			     int need, int want, int *got);
911 
912 /* for counting open files by mode */
913 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode);
914 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
915 
916 /* addr.c */
917 extern const struct address_space_operations ceph_aops;
918 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
919 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
920 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
921 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
922 
923 /* file.c */
924 extern const struct file_operations ceph_file_fops;
925 
926 extern int ceph_renew_caps(struct inode *inode);
927 extern int ceph_open(struct inode *inode, struct file *file);
928 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
929 			    struct file *file, unsigned flags, umode_t mode,
930 			    int *opened);
931 extern int ceph_release(struct inode *inode, struct file *filp);
932 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
933 				  char *data, size_t len);
934 
935 /* dir.c */
936 extern const struct file_operations ceph_dir_fops;
937 extern const struct file_operations ceph_snapdir_fops;
938 extern const struct inode_operations ceph_dir_iops;
939 extern const struct inode_operations ceph_snapdir_iops;
940 extern const struct dentry_operations ceph_dentry_ops;
941 
942 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
943 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
944 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
945 			       struct dentry *dentry, int err);
946 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
947 					 struct dentry *dentry, int err);
948 
949 extern void ceph_dentry_lru_add(struct dentry *dn);
950 extern void ceph_dentry_lru_touch(struct dentry *dn);
951 extern void ceph_dentry_lru_del(struct dentry *dn);
952 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
953 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
954 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
955 
956 /* ioctl.c */
957 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
958 
959 /* export.c */
960 extern const struct export_operations ceph_export_ops;
961 
962 /* locks.c */
963 extern __init void ceph_flock_init(void);
964 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
965 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
966 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
967 extern int ceph_encode_locks_to_buffer(struct inode *inode,
968 				       struct ceph_filelock *flocks,
969 				       int num_fcntl_locks,
970 				       int num_flock_locks);
971 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
972 				  struct ceph_pagelist *pagelist,
973 				  int num_fcntl_locks, int num_flock_locks);
974 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
975 
976 /* debugfs.c */
977 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
978 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
979 
980 #endif /* _FS_CEPH_SUPER_H */
981