xref: /openbmc/linux/fs/ceph/super.h (revision dbf563ee)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_SUPER_H
3 #define _FS_CEPH_SUPER_H
4 
5 #include <linux/ceph/ceph_debug.h>
6 
7 #include <asm/unaligned.h>
8 #include <linux/backing-dev.h>
9 #include <linux/completion.h>
10 #include <linux/exportfs.h>
11 #include <linux/fs.h>
12 #include <linux/mempool.h>
13 #include <linux/pagemap.h>
14 #include <linux/wait.h>
15 #include <linux/writeback.h>
16 #include <linux/slab.h>
17 #include <linux/posix_acl.h>
18 #include <linux/refcount.h>
19 #include <linux/security.h>
20 
21 #include <linux/ceph/libceph.h>
22 
23 #ifdef CONFIG_CEPH_FSCACHE
24 #include <linux/fscache.h>
25 #endif
26 
27 /* f_type in struct statfs */
28 #define CEPH_SUPER_MAGIC 0x00c36400
29 
30 /* large granularity for statfs utilization stats to facilitate
31  * large volume sizes on 32-bit machines. */
32 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
33 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
34 
35 #define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blacklisted */
36 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
37 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
38 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
39 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
40 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
41 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
42 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
43 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
44 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
45 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
46 #define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
47 
48 #define CEPH_MOUNT_OPT_DEFAULT			\
49 	(CEPH_MOUNT_OPT_DCACHE |		\
50 	 CEPH_MOUNT_OPT_NOCOPYFROM)
51 
52 #define ceph_set_mount_opt(fsc, opt) \
53 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
54 #define ceph_clear_mount_opt(fsc, opt) \
55 	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
56 #define ceph_test_mount_opt(fsc, opt) \
57 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
58 
59 /* max size of osd read request, limited by libceph */
60 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
61 /* osd has a configurable limitaion of max write size.
62  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
63 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
64 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
65 #define CEPH_MAX_READDIR_DEFAULT        1024
66 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
67 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
68 
69 /*
70  * Delay telling the MDS we no longer want caps, in case we reopen
71  * the file.  Delay a minimum amount of time, even if we send a cap
72  * message for some other reason.  Otherwise, take the oppotunity to
73  * update the mds to avoid sending another message later.
74  */
75 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
76 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
77 
78 struct ceph_mount_options {
79 	unsigned int flags;
80 
81 	unsigned int wsize;            /* max write size */
82 	unsigned int rsize;            /* max read size */
83 	unsigned int rasize;           /* max readahead */
84 	unsigned int congestion_kb;    /* max writeback in flight */
85 	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
86 	int caps_max;
87 	unsigned int max_readdir;       /* max readdir result (entries) */
88 	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
89 
90 	/*
91 	 * everything above this point can be memcmp'd; everything below
92 	 * is handled in compare_mount_options()
93 	 */
94 
95 	char *snapdir_name;   /* default ".snap" */
96 	char *mds_namespace;  /* default NULL */
97 	char *server_path;    /* default NULL (means "/") */
98 	char *fscache_uniq;   /* default NULL */
99 };
100 
101 struct ceph_fs_client {
102 	struct super_block *sb;
103 
104 	struct list_head metric_wakeup;
105 
106 	struct ceph_mount_options *mount_options;
107 	struct ceph_client *client;
108 
109 	unsigned long mount_state;
110 
111 	unsigned long last_auto_reconnect;
112 	bool blacklisted;
113 
114 	bool have_copy_from2;
115 
116 	u32 filp_gen;
117 	loff_t max_file_size;
118 
119 	struct ceph_mds_client *mdsc;
120 
121 	atomic_long_t writeback_count;
122 
123 	struct workqueue_struct *inode_wq;
124 	struct workqueue_struct *cap_wq;
125 
126 #ifdef CONFIG_DEBUG_FS
127 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
128 	struct dentry *debugfs_congestion_kb;
129 	struct dentry *debugfs_bdi;
130 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
131 	struct dentry *debugfs_metric;
132 	struct dentry *debugfs_mds_sessions;
133 #endif
134 
135 #ifdef CONFIG_CEPH_FSCACHE
136 	struct fscache_cookie *fscache;
137 #endif
138 };
139 
140 
141 /*
142  * File i/o capability.  This tracks shared state with the metadata
143  * server that allows us to cache or writeback attributes or to read
144  * and write data.  For any given inode, we should have one or more
145  * capabilities, one issued by each metadata server, and our
146  * cumulative access is the OR of all issued capabilities.
147  *
148  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
149  * session capability lists.
150  */
151 struct ceph_cap {
152 	struct ceph_inode_info *ci;
153 	struct rb_node ci_node;          /* per-ci cap tree */
154 	struct ceph_mds_session *session;
155 	struct list_head session_caps;   /* per-session caplist */
156 	u64 cap_id;       /* unique cap id (mds provided) */
157 	union {
158 		/* in-use caps */
159 		struct {
160 			int issued;       /* latest, from the mds */
161 			int implemented;  /* implemented superset of
162 					     issued (for revocation) */
163 			int mds, mds_wanted;
164 		};
165 		/* caps to release */
166 		struct {
167 			u64 cap_ino;
168 			int queue_release;
169 		};
170 	};
171 	u32 seq, issue_seq, mseq;
172 	u32 cap_gen;      /* active/stale cycle */
173 	unsigned long last_used;
174 	struct list_head caps_item;
175 };
176 
177 #define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
178 #define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
179 #define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
180 
181 struct ceph_cap_flush {
182 	u64 tid;
183 	int caps; /* 0 means capsnap */
184 	bool wake; /* wake up flush waiters when finish ? */
185 	struct list_head g_list; // global
186 	struct list_head i_list; // per inode
187 };
188 
189 /*
190  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
191  * we first complete any in-process sync writes and writeback any dirty
192  * data before flushing the snapped state (tracked here) back to the MDS.
193  */
194 struct ceph_cap_snap {
195 	refcount_t nref;
196 	struct list_head ci_item;
197 
198 	struct ceph_cap_flush cap_flush;
199 
200 	u64 follows;
201 	int issued, dirty;
202 	struct ceph_snap_context *context;
203 
204 	umode_t mode;
205 	kuid_t uid;
206 	kgid_t gid;
207 
208 	struct ceph_buffer *xattr_blob;
209 	u64 xattr_version;
210 
211 	u64 size;
212 	u64 change_attr;
213 	struct timespec64 mtime, atime, ctime, btime;
214 	u64 time_warp_seq;
215 	u64 truncate_size;
216 	u32 truncate_seq;
217 	int writing;   /* a sync write is still in progress */
218 	int dirty_pages;     /* dirty pages awaiting writeback */
219 	bool inline_data;
220 	bool need_flush;
221 };
222 
223 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
224 {
225 	if (refcount_dec_and_test(&capsnap->nref)) {
226 		if (capsnap->xattr_blob)
227 			ceph_buffer_put(capsnap->xattr_blob);
228 		kfree(capsnap);
229 	}
230 }
231 
232 /*
233  * The frag tree describes how a directory is fragmented, potentially across
234  * multiple metadata servers.  It is also used to indicate points where
235  * metadata authority is delegated, and whether/where metadata is replicated.
236  *
237  * A _leaf_ frag will be present in the i_fragtree IFF there is
238  * delegation info.  That is, if mds >= 0 || ndist > 0.
239  */
240 #define CEPH_MAX_DIRFRAG_REP 4
241 
242 struct ceph_inode_frag {
243 	struct rb_node node;
244 
245 	/* fragtree state */
246 	u32 frag;
247 	int split_by;         /* i.e. 2^(split_by) children */
248 
249 	/* delegation and replication info */
250 	int mds;              /* -1 if same authority as parent */
251 	int ndist;            /* >0 if replicated */
252 	int dist[CEPH_MAX_DIRFRAG_REP];
253 };
254 
255 /*
256  * We cache inode xattrs as an encoded blob until they are first used,
257  * at which point we parse them into an rbtree.
258  */
259 struct ceph_inode_xattr {
260 	struct rb_node node;
261 
262 	const char *name;
263 	int name_len;
264 	const char *val;
265 	int val_len;
266 	int dirty;
267 
268 	int should_free_name;
269 	int should_free_val;
270 };
271 
272 /*
273  * Ceph dentry state
274  */
275 struct ceph_dentry_info {
276 	struct dentry *dentry;
277 	struct ceph_mds_session *lease_session;
278 	struct list_head lease_list;
279 	unsigned flags;
280 	int lease_shared_gen;
281 	u32 lease_gen;
282 	u32 lease_seq;
283 	unsigned long lease_renew_after, lease_renew_from;
284 	unsigned long time;
285 	u64 offset;
286 };
287 
288 #define CEPH_DENTRY_REFERENCED		1
289 #define CEPH_DENTRY_LEASE_LIST		2
290 #define CEPH_DENTRY_SHRINK_LIST		4
291 #define CEPH_DENTRY_PRIMARY_LINK	8
292 
293 struct ceph_inode_xattrs_info {
294 	/*
295 	 * (still encoded) xattr blob. we avoid the overhead of parsing
296 	 * this until someone actually calls getxattr, etc.
297 	 *
298 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
299 	 * NULL means we don't know.
300 	*/
301 	struct ceph_buffer *blob, *prealloc_blob;
302 
303 	struct rb_root index;
304 	bool dirty;
305 	int count;
306 	int names_size;
307 	int vals_size;
308 	u64 version, index_version;
309 };
310 
311 /*
312  * Ceph inode.
313  */
314 struct ceph_inode_info {
315 	struct ceph_vino i_vino;   /* ceph ino + snap */
316 
317 	spinlock_t i_ceph_lock;
318 
319 	u64 i_version;
320 	u64 i_inline_version;
321 	u32 i_time_warp_seq;
322 
323 	unsigned long i_ceph_flags;
324 	atomic64_t i_release_count;
325 	atomic64_t i_ordered_count;
326 	atomic64_t i_complete_seq[2];
327 
328 	struct ceph_dir_layout i_dir_layout;
329 	struct ceph_file_layout i_layout;
330 	struct ceph_file_layout i_cached_layout;	// for async creates
331 	char *i_symlink;
332 
333 	/* for dirs */
334 	struct timespec64 i_rctime;
335 	u64 i_rbytes, i_rfiles, i_rsubdirs;
336 	u64 i_files, i_subdirs;
337 
338 	/* quotas */
339 	u64 i_max_bytes, i_max_files;
340 
341 	s32 i_dir_pin;
342 
343 	struct rb_root i_fragtree;
344 	int i_fragtree_nsplits;
345 	struct mutex i_fragtree_mutex;
346 
347 	struct ceph_inode_xattrs_info i_xattrs;
348 
349 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
350 	 * s_mutex. */
351 	struct rb_root i_caps;           /* cap list */
352 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
353 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
354 
355 	/*
356 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
357 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
358 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
359 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
360 	 * be tested under the i_ceph_lock. Changing anything requires both.
361 	 */
362 	struct list_head i_dirty_item;
363 
364 	/*
365 	 * Link to session's s_cap_flushing list. Protected in a similar
366 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
367 	 * s_cap_flushing list can be walked while holding either the s_mutex
368 	 * or msdc->cap_dirty_lock. List presence can also be checked while
369 	 * holding the i_ceph_lock for this inode.
370 	 */
371 	struct list_head i_flushing_item;
372 
373 	/* we need to track cap writeback on a per-cap-bit basis, to allow
374 	 * overlapping, pipelined cap flushes to the mds.  we can probably
375 	 * reduce the tid to 8 bits if we're concerned about inode size. */
376 	struct ceph_cap_flush *i_prealloc_cap_flush;
377 	struct list_head i_cap_flush_list;
378 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
379 	unsigned long i_hold_caps_max; /* jiffies */
380 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
381 	struct ceph_cap_reservation i_cap_migration_resv;
382 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
383 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
384 						    dirty|flushing caps */
385 	unsigned i_snap_caps;           /* cap bits for snapped files */
386 
387 	unsigned long i_last_rd;
388 	unsigned long i_last_wr;
389 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
390 
391 	struct mutex i_truncate_mutex;
392 	u32 i_truncate_seq;        /* last truncate to smaller size */
393 	u64 i_truncate_size;       /*  and the size we last truncated down to */
394 	int i_truncate_pending;    /*  still need to call vmtruncate */
395 
396 	u64 i_max_size;            /* max file size authorized by mds */
397 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
398 	u64 i_wanted_max_size;     /* offset we'd like to write too */
399 	u64 i_requested_max_size;  /* max_size we've requested */
400 
401 	/* held references to caps */
402 	int i_pin_ref;
403 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
404 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
405 	atomic_t i_filelock_ref;
406 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
407 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
408 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
409 
410 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
411 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
412 	spinlock_t i_unsafe_lock;
413 
414 	union {
415 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
416 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
417 	};
418 	int i_snap_realm_counter; /* snap realm (if caps) */
419 	struct list_head i_snap_realm_item;
420 	struct list_head i_snap_flush_item;
421 	struct timespec64 i_btime;
422 	struct timespec64 i_snap_btime;
423 
424 	struct work_struct i_work;
425 	unsigned long  i_work_mask;
426 
427 #ifdef CONFIG_CEPH_FSCACHE
428 	struct fscache_cookie *fscache;
429 	u32 i_fscache_gen;
430 #endif
431 	errseq_t i_meta_err;
432 
433 	struct inode vfs_inode; /* at end */
434 };
435 
436 static inline struct ceph_inode_info *
437 ceph_inode(const struct inode *inode)
438 {
439 	return container_of(inode, struct ceph_inode_info, vfs_inode);
440 }
441 
442 static inline struct ceph_fs_client *
443 ceph_inode_to_client(const struct inode *inode)
444 {
445 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
446 }
447 
448 static inline struct ceph_fs_client *
449 ceph_sb_to_client(const struct super_block *sb)
450 {
451 	return (struct ceph_fs_client *)sb->s_fs_info;
452 }
453 
454 static inline struct ceph_vino
455 ceph_vino(const struct inode *inode)
456 {
457 	return ceph_inode(inode)->i_vino;
458 }
459 
460 static inline u32 ceph_ino_to_ino32(u64 vino)
461 {
462 	u32 ino = vino & 0xffffffff;
463 	ino ^= vino >> 32;
464 	if (!ino)
465 		ino = 2;
466 	return ino;
467 }
468 
469 /*
470  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
471  * some arches. We generally do not use this value inside the ceph driver, but
472  * we do want to set it to something, so that generic vfs code has an
473  * appropriate value for tracepoints and the like.
474  */
475 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
476 {
477 	if (sizeof(ino_t) == sizeof(u32))
478 		return ceph_ino_to_ino32(vino.ino);
479 	return (ino_t)vino.ino;
480 }
481 
482 /* for printf-style formatting */
483 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
484 
485 static inline u64 ceph_ino(struct inode *inode)
486 {
487 	return ceph_inode(inode)->i_vino.ino;
488 }
489 
490 static inline u64 ceph_snap(struct inode *inode)
491 {
492 	return ceph_inode(inode)->i_vino.snap;
493 }
494 
495 /**
496  * ceph_present_ino - format an inode number for presentation to userland
497  * @sb: superblock where the inode lives
498  * @ino: inode number to (possibly) convert
499  *
500  * If the user mounted with the ino32 option, then the 64-bit value needs
501  * to be converted to something that can fit inside 32 bits. Note that
502  * internal kernel code never uses this value, so this is entirely for
503  * userland consumption.
504  */
505 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
506 {
507 	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
508 		return ceph_ino_to_ino32(ino);
509 	return ino;
510 }
511 
512 static inline u64 ceph_present_inode(struct inode *inode)
513 {
514 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
515 }
516 
517 static inline int ceph_ino_compare(struct inode *inode, void *data)
518 {
519 	struct ceph_vino *pvino = (struct ceph_vino *)data;
520 	struct ceph_inode_info *ci = ceph_inode(inode);
521 	return ci->i_vino.ino == pvino->ino &&
522 		ci->i_vino.snap == pvino->snap;
523 }
524 
525 
526 static inline struct inode *ceph_find_inode(struct super_block *sb,
527 					    struct ceph_vino vino)
528 {
529 	/*
530 	 * NB: The hashval will be run through the fs/inode.c hash function
531 	 * anyway, so there is no need to squash the inode number down to
532 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
533 	 */
534 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
535 }
536 
537 
538 /*
539  * Ceph inode.
540  */
541 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
542 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
543 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
544 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
545 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
546 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
547 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
548 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
549 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
550 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
551 #define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
552 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
553 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
554 
555 /*
556  * Masks of ceph inode work.
557  */
558 #define CEPH_I_WORK_WRITEBACK		0 /* writeback */
559 #define CEPH_I_WORK_INVALIDATE_PAGES	1 /* invalidate pages */
560 #define CEPH_I_WORK_VMTRUNCATE		2 /* vmtruncate */
561 
562 /*
563  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
564  * and then clear it when they start succeeding. Note that we do a lockless
565  * check first, and only take the lock if it looks like it needs to be changed.
566  * The write submission code just takes this as a hint, so we're not too
567  * worried if a few slip through in either direction.
568  */
569 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
570 {
571 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
572 		spin_lock(&ci->i_ceph_lock);
573 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
574 		spin_unlock(&ci->i_ceph_lock);
575 	}
576 }
577 
578 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
579 {
580 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
581 		spin_lock(&ci->i_ceph_lock);
582 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
583 		spin_unlock(&ci->i_ceph_lock);
584 	}
585 }
586 
587 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
588 					   long long release_count,
589 					   long long ordered_count)
590 {
591 	/*
592 	 * Makes sure operations that setup readdir cache (update page
593 	 * cache and i_size) are strongly ordered w.r.t. the following
594 	 * atomic64_set() operations.
595 	 */
596 	smp_mb();
597 	atomic64_set(&ci->i_complete_seq[0], release_count);
598 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
599 }
600 
601 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
602 {
603 	atomic64_inc(&ci->i_release_count);
604 }
605 
606 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
607 {
608 	atomic64_inc(&ci->i_ordered_count);
609 }
610 
611 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
612 {
613 	return atomic64_read(&ci->i_complete_seq[0]) ==
614 		atomic64_read(&ci->i_release_count);
615 }
616 
617 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
618 {
619 	return  atomic64_read(&ci->i_complete_seq[0]) ==
620 		atomic64_read(&ci->i_release_count) &&
621 		atomic64_read(&ci->i_complete_seq[1]) ==
622 		atomic64_read(&ci->i_ordered_count);
623 }
624 
625 static inline void ceph_dir_clear_complete(struct inode *inode)
626 {
627 	__ceph_dir_clear_complete(ceph_inode(inode));
628 }
629 
630 static inline void ceph_dir_clear_ordered(struct inode *inode)
631 {
632 	__ceph_dir_clear_ordered(ceph_inode(inode));
633 }
634 
635 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
636 {
637 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
638 	smp_rmb();
639 	return ret;
640 }
641 
642 /* find a specific frag @f */
643 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
644 						u32 f);
645 
646 /*
647  * choose fragment for value @v.  copy frag content to pfrag, if leaf
648  * exists
649  */
650 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
651 			    struct ceph_inode_frag *pfrag,
652 			    int *found);
653 
654 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
655 {
656 	return (struct ceph_dentry_info *)dentry->d_fsdata;
657 }
658 
659 /*
660  * caps helpers
661  */
662 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
663 {
664 	return !RB_EMPTY_ROOT(&ci->i_caps);
665 }
666 
667 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
668 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
669 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
670 					  int t);
671 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
672 				    struct ceph_cap *cap);
673 
674 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
675 {
676 	int issued;
677 	spin_lock(&ci->i_ceph_lock);
678 	issued = __ceph_caps_issued(ci, NULL);
679 	spin_unlock(&ci->i_ceph_lock);
680 	return issued;
681 }
682 
683 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
684 					       int mask, int touch)
685 {
686 	int r;
687 	spin_lock(&ci->i_ceph_lock);
688 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
689 	spin_unlock(&ci->i_ceph_lock);
690 	return r;
691 }
692 
693 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
694 {
695 	return ci->i_dirty_caps | ci->i_flushing_caps;
696 }
697 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
698 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
699 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
700 				  struct ceph_cap_flush **pcf);
701 
702 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
703 				      struct ceph_cap *ocap, int mask);
704 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
705 extern int __ceph_caps_used(struct ceph_inode_info *ci);
706 
707 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
708 {
709 	return ci->i_nr_by_mode[0];
710 }
711 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
712 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
713 
714 /* what the mds thinks we want */
715 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
716 
717 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
718 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
719 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
720 				     struct ceph_mount_options *fsopt);
721 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
722 			     struct ceph_cap_reservation *ctx, int need);
723 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
724 			       struct ceph_cap_reservation *ctx);
725 extern void ceph_reservation_status(struct ceph_fs_client *client,
726 				    int *total, int *avail, int *used,
727 				    int *reserved, int *min);
728 
729 
730 
731 /*
732  * we keep buffered readdir results attached to file->private_data
733  */
734 #define CEPH_F_SYNC     1
735 #define CEPH_F_ATEND    2
736 
737 struct ceph_file_info {
738 	short fmode;     /* initialized on open */
739 	short flags;     /* CEPH_F_* */
740 
741 	spinlock_t rw_contexts_lock;
742 	struct list_head rw_contexts;
743 
744 	errseq_t meta_err;
745 	u32 filp_gen;
746 	atomic_t num_locks;
747 };
748 
749 struct ceph_dir_file_info {
750 	struct ceph_file_info file_info;
751 
752 	/* readdir: position within the dir */
753 	u32 frag;
754 	struct ceph_mds_request *last_readdir;
755 
756 	/* readdir: position within a frag */
757 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
758 	char *last_name;       /* last entry in previous chunk */
759 	long long dir_release_count;
760 	long long dir_ordered_count;
761 	int readdir_cache_idx;
762 
763 	/* used for -o dirstat read() on directory thing */
764 	char *dir_info;
765 	int dir_info_len;
766 };
767 
768 struct ceph_rw_context {
769 	struct list_head list;
770 	struct task_struct *thread;
771 	int caps;
772 };
773 
774 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
775 	struct ceph_rw_context _name = {	\
776 		.thread = current,		\
777 		.caps = _caps,			\
778 	}
779 
780 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
781 				       struct ceph_rw_context *ctx)
782 {
783 	spin_lock(&cf->rw_contexts_lock);
784 	list_add(&ctx->list, &cf->rw_contexts);
785 	spin_unlock(&cf->rw_contexts_lock);
786 }
787 
788 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
789 				       struct ceph_rw_context *ctx)
790 {
791 	spin_lock(&cf->rw_contexts_lock);
792 	list_del(&ctx->list);
793 	spin_unlock(&cf->rw_contexts_lock);
794 }
795 
796 static inline struct ceph_rw_context*
797 ceph_find_rw_context(struct ceph_file_info *cf)
798 {
799 	struct ceph_rw_context *ctx, *found = NULL;
800 	spin_lock(&cf->rw_contexts_lock);
801 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
802 		if (ctx->thread == current) {
803 			found = ctx;
804 			break;
805 		}
806 	}
807 	spin_unlock(&cf->rw_contexts_lock);
808 	return found;
809 }
810 
811 struct ceph_readdir_cache_control {
812 	struct page  *page;
813 	struct dentry **dentries;
814 	int index;
815 };
816 
817 /*
818  * A "snap realm" describes a subset of the file hierarchy sharing
819  * the same set of snapshots that apply to it.  The realms themselves
820  * are organized into a hierarchy, such that children inherit (some of)
821  * the snapshots of their parents.
822  *
823  * All inodes within the realm that have capabilities are linked into a
824  * per-realm list.
825  */
826 struct ceph_snap_realm {
827 	u64 ino;
828 	struct inode *inode;
829 	atomic_t nref;
830 	struct rb_node node;
831 
832 	u64 created, seq;
833 	u64 parent_ino;
834 	u64 parent_since;   /* snapid when our current parent became so */
835 
836 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
837 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
838 	u64 *snaps;                   /* snaps specific to this realm */
839 	u32 num_snaps;
840 
841 	struct ceph_snap_realm *parent;
842 	struct list_head children;       /* list of child realms */
843 	struct list_head child_item;
844 
845 	struct list_head empty_item;     /* if i have ref==0 */
846 
847 	struct list_head dirty_item;     /* if realm needs new context */
848 
849 	/* the current set of snaps for this realm */
850 	struct ceph_snap_context *cached_context;
851 
852 	struct list_head inodes_with_caps;
853 	spinlock_t inodes_with_caps_lock;
854 };
855 
856 static inline int default_congestion_kb(void)
857 {
858 	int congestion_kb;
859 
860 	/*
861 	 * Copied from NFS
862 	 *
863 	 * congestion size, scale with available memory.
864 	 *
865 	 *  64MB:    8192k
866 	 * 128MB:   11585k
867 	 * 256MB:   16384k
868 	 * 512MB:   23170k
869 	 *   1GB:   32768k
870 	 *   2GB:   46340k
871 	 *   4GB:   65536k
872 	 *   8GB:   92681k
873 	 *  16GB:  131072k
874 	 *
875 	 * This allows larger machines to have larger/more transfers.
876 	 * Limit the default to 256M
877 	 */
878 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
879 	if (congestion_kb > 256*1024)
880 		congestion_kb = 256*1024;
881 
882 	return congestion_kb;
883 }
884 
885 
886 /* super.c */
887 extern int ceph_force_reconnect(struct super_block *sb);
888 /* snap.c */
889 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
890 					       u64 ino);
891 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
892 				struct ceph_snap_realm *realm);
893 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
894 				struct ceph_snap_realm *realm);
895 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
896 				  void *p, void *e, bool deletion,
897 				  struct ceph_snap_realm **realm_ret);
898 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
899 			     struct ceph_mds_session *session,
900 			     struct ceph_msg *msg);
901 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
902 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
903 				  struct ceph_cap_snap *capsnap);
904 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
905 
906 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
907 						   u64 snap);
908 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
909 				struct ceph_snapid_map *sm);
910 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
911 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
912 
913 
914 /*
915  * a cap_snap is "pending" if it is still awaiting an in-progress
916  * sync write (that may/may not still update size, mtime, etc.).
917  */
918 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
919 {
920 	return !list_empty(&ci->i_cap_snaps) &&
921 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
922 			       ci_item)->writing;
923 }
924 
925 /* inode.c */
926 struct ceph_mds_reply_info_in;
927 struct ceph_mds_reply_dirfrag;
928 
929 extern const struct inode_operations ceph_file_iops;
930 
931 extern struct inode *ceph_alloc_inode(struct super_block *sb);
932 extern void ceph_evict_inode(struct inode *inode);
933 extern void ceph_free_inode(struct inode *inode);
934 
935 extern struct inode *ceph_get_inode(struct super_block *sb,
936 				    struct ceph_vino vino);
937 extern struct inode *ceph_get_snapdir(struct inode *parent);
938 extern int ceph_fill_file_size(struct inode *inode, int issued,
939 			       u32 truncate_seq, u64 truncate_size, u64 size);
940 extern void ceph_fill_file_time(struct inode *inode, int issued,
941 				u64 time_warp_seq, struct timespec64 *ctime,
942 				struct timespec64 *mtime,
943 				struct timespec64 *atime);
944 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
945 		    struct ceph_mds_reply_info_in *iinfo,
946 		    struct ceph_mds_reply_dirfrag *dirinfo,
947 		    struct ceph_mds_session *session, int cap_fmode,
948 		    struct ceph_cap_reservation *caps_reservation);
949 extern int ceph_fill_trace(struct super_block *sb,
950 			   struct ceph_mds_request *req);
951 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
952 				    struct ceph_mds_session *session);
953 
954 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
955 
956 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
957 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
958 extern void ceph_queue_vmtruncate(struct inode *inode);
959 extern void ceph_queue_invalidate(struct inode *inode);
960 extern void ceph_queue_writeback(struct inode *inode);
961 extern void ceph_async_iput(struct inode *inode);
962 
963 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
964 			     int mask, bool force);
965 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
966 {
967 	return __ceph_do_getattr(inode, NULL, mask, force);
968 }
969 extern int ceph_permission(struct inode *inode, int mask);
970 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
971 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
972 extern int ceph_getattr(const struct path *path, struct kstat *stat,
973 			u32 request_mask, unsigned int flags);
974 
975 /* xattr.c */
976 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
977 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
978 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
979 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
980 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
981 extern const struct xattr_handler *ceph_xattr_handlers[];
982 
983 struct ceph_acl_sec_ctx {
984 #ifdef CONFIG_CEPH_FS_POSIX_ACL
985 	void *default_acl;
986 	void *acl;
987 #endif
988 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
989 	void *sec_ctx;
990 	u32 sec_ctxlen;
991 #endif
992 	struct ceph_pagelist *pagelist;
993 };
994 
995 #ifdef CONFIG_SECURITY
996 extern bool ceph_security_xattr_deadlock(struct inode *in);
997 extern bool ceph_security_xattr_wanted(struct inode *in);
998 #else
999 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1000 {
1001 	return false;
1002 }
1003 static inline bool ceph_security_xattr_wanted(struct inode *in)
1004 {
1005 	return false;
1006 }
1007 #endif
1008 
1009 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1010 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1011 				     struct ceph_acl_sec_ctx *ctx);
1012 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1013 {
1014 	security_inode_invalidate_secctx(inode);
1015 }
1016 #else
1017 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1018 					    struct ceph_acl_sec_ctx *ctx)
1019 {
1020 	return 0;
1021 }
1022 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1023 {
1024 }
1025 #endif
1026 
1027 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1028 
1029 /* acl.c */
1030 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1031 
1032 struct posix_acl *ceph_get_acl(struct inode *, int);
1033 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1034 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1035 		       struct ceph_acl_sec_ctx *as_ctx);
1036 void ceph_init_inode_acls(struct inode *inode,
1037 			  struct ceph_acl_sec_ctx *as_ctx);
1038 
1039 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1040 {
1041        forget_all_cached_acls(inode);
1042 }
1043 
1044 #else
1045 
1046 #define ceph_get_acl NULL
1047 #define ceph_set_acl NULL
1048 
1049 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1050 				     struct ceph_acl_sec_ctx *as_ctx)
1051 {
1052 	return 0;
1053 }
1054 static inline void ceph_init_inode_acls(struct inode *inode,
1055 					struct ceph_acl_sec_ctx *as_ctx)
1056 {
1057 }
1058 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1059 {
1060 	return 0;
1061 }
1062 
1063 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1064 {
1065 }
1066 
1067 #endif
1068 
1069 /* caps.c */
1070 extern const char *ceph_cap_string(int c);
1071 extern void ceph_handle_caps(struct ceph_mds_session *session,
1072 			     struct ceph_msg *msg);
1073 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1074 				     struct ceph_cap_reservation *ctx);
1075 extern void ceph_add_cap(struct inode *inode,
1076 			 struct ceph_mds_session *session, u64 cap_id,
1077 			 unsigned issued, unsigned wanted,
1078 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1079 			 struct ceph_cap **new_cap);
1080 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1081 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1082 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1083 			 struct ceph_cap *cap);
1084 extern int ceph_is_any_caps(struct inode *inode);
1085 
1086 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1087 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1088 		      int datasync);
1089 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1090 					  struct ceph_mds_session *session);
1091 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1092 				    struct ceph_mds_session *session);
1093 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1094 				   struct ceph_inode_info *ci);
1095 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1096 					     int mds);
1097 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1098 				bool snap_rwsem_locked);
1099 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1100 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1101 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1102 					    int had);
1103 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1104 				       struct ceph_snap_context *snapc);
1105 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1106 			     struct ceph_mds_session **psession);
1107 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1108 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1109 			    struct ceph_mds_session *session);
1110 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1111 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1112 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1113 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1114 				     int mds, int drop, int unless, int force);
1115 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1116 				      struct inode *dir,
1117 				      int mds, int drop, int unless);
1118 
1119 extern int ceph_get_caps(struct file *filp, int need, int want,
1120 			 loff_t endoff, int *got, struct page **pinned_page);
1121 extern int ceph_try_get_caps(struct inode *inode,
1122 			     int need, int want, bool nonblock, int *got);
1123 
1124 /* for counting open files by mode */
1125 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1126 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1127 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1128 			       struct ceph_mds_client *mdsc, int fmode);
1129 
1130 /* addr.c */
1131 extern const struct address_space_operations ceph_aops;
1132 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1133 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1134 extern int ceph_pool_perm_check(struct inode *inode, int need);
1135 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1136 
1137 /* file.c */
1138 extern const struct file_operations ceph_file_fops;
1139 
1140 extern int ceph_renew_caps(struct inode *inode, int fmode);
1141 extern int ceph_open(struct inode *inode, struct file *file);
1142 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1143 			    struct file *file, unsigned flags, umode_t mode);
1144 extern int ceph_release(struct inode *inode, struct file *filp);
1145 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1146 				  char *data, size_t len);
1147 
1148 /* dir.c */
1149 extern const struct file_operations ceph_dir_fops;
1150 extern const struct file_operations ceph_snapdir_fops;
1151 extern const struct inode_operations ceph_dir_iops;
1152 extern const struct inode_operations ceph_snapdir_iops;
1153 extern const struct dentry_operations ceph_dentry_ops;
1154 
1155 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1156 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1157 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1158 			       struct dentry *dentry, int err);
1159 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1160 					 struct dentry *dentry, int err);
1161 
1162 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1163 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1164 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1165 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1166 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1167 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1168 
1169 /* ioctl.c */
1170 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1171 
1172 /* export.c */
1173 extern const struct export_operations ceph_export_ops;
1174 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1175 
1176 /* locks.c */
1177 extern __init void ceph_flock_init(void);
1178 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1179 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1180 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1181 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1182 				       struct ceph_filelock *flocks,
1183 				       int num_fcntl_locks,
1184 				       int num_flock_locks);
1185 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1186 				  struct ceph_pagelist *pagelist,
1187 				  int num_fcntl_locks, int num_flock_locks);
1188 
1189 /* debugfs.c */
1190 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1191 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1192 
1193 /* quota.c */
1194 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1195 {
1196 	return ci->i_max_files || ci->i_max_bytes;
1197 }
1198 
1199 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1200 
1201 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1202 				       u64 max_bytes, u64 max_files)
1203 {
1204 	bool had_quota, has_quota;
1205 	had_quota = __ceph_has_any_quota(ci);
1206 	ci->i_max_bytes = max_bytes;
1207 	ci->i_max_files = max_files;
1208 	has_quota = __ceph_has_any_quota(ci);
1209 
1210 	if (had_quota != has_quota)
1211 		ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1212 }
1213 
1214 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1215 			      struct ceph_mds_session *session,
1216 			      struct ceph_msg *msg);
1217 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1218 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1219 					     loff_t newlen);
1220 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1221 						loff_t newlen);
1222 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1223 				     struct kstatfs *buf);
1224 extern int ceph_quota_check_rename(struct ceph_mds_client *mdsc,
1225 				   struct inode *old, struct inode *new);
1226 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1227 
1228 #endif /* _FS_CEPH_SUPER_H */
1229