xref: /linux/fs/ceph/super.h (revision 6c363eafc4d637ac4bd83d4a7dd06dd3cfbe7c5f)
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 blocklisted */
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 	int mount_state;
110 
111 	bool blocklisted;
112 
113 	bool have_copy_from2;
114 
115 	u32 filp_gen;
116 	loff_t max_file_size;
117 
118 	struct ceph_mds_client *mdsc;
119 
120 	atomic_long_t writeback_count;
121 
122 	struct workqueue_struct *inode_wq;
123 	struct workqueue_struct *cap_wq;
124 
125 #ifdef CONFIG_DEBUG_FS
126 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
127 	struct dentry *debugfs_congestion_kb;
128 	struct dentry *debugfs_bdi;
129 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
130 	struct dentry *debugfs_metric;
131 	struct dentry *debugfs_status;
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 index for this cap */
164 			int mds_wanted;   /* caps wanted from this mds */
165 		};
166 		/* caps to release */
167 		struct {
168 			u64 cap_ino;
169 			int queue_release;
170 		};
171 	};
172 	u32 seq, issue_seq, mseq;
173 	u32 cap_gen;      /* active/stale cycle */
174 	unsigned long last_used;
175 	struct list_head caps_item;
176 };
177 
178 #define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
179 #define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
180 #define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
181 
182 struct ceph_cap_flush {
183 	u64 tid;
184 	int caps; /* 0 means capsnap */
185 	bool wake; /* wake up flush waiters when finish ? */
186 	struct list_head g_list; // global
187 	struct list_head i_list; // per inode
188 };
189 
190 /*
191  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
192  * we first complete any in-process sync writes and writeback any dirty
193  * data before flushing the snapped state (tracked here) back to the MDS.
194  */
195 struct ceph_cap_snap {
196 	refcount_t nref;
197 	struct list_head ci_item;
198 
199 	struct ceph_cap_flush cap_flush;
200 
201 	u64 follows;
202 	int issued, dirty;
203 	struct ceph_snap_context *context;
204 
205 	umode_t mode;
206 	kuid_t uid;
207 	kgid_t gid;
208 
209 	struct ceph_buffer *xattr_blob;
210 	u64 xattr_version;
211 
212 	u64 size;
213 	u64 change_attr;
214 	struct timespec64 mtime, atime, ctime, btime;
215 	u64 time_warp_seq;
216 	u64 truncate_size;
217 	u32 truncate_seq;
218 	int writing;   /* a sync write is still in progress */
219 	int dirty_pages;     /* dirty pages awaiting writeback */
220 	bool inline_data;
221 	bool need_flush;
222 };
223 
224 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
225 {
226 	if (refcount_dec_and_test(&capsnap->nref)) {
227 		if (capsnap->xattr_blob)
228 			ceph_buffer_put(capsnap->xattr_blob);
229 		kfree(capsnap);
230 	}
231 }
232 
233 /*
234  * The frag tree describes how a directory is fragmented, potentially across
235  * multiple metadata servers.  It is also used to indicate points where
236  * metadata authority is delegated, and whether/where metadata is replicated.
237  *
238  * A _leaf_ frag will be present in the i_fragtree IFF there is
239  * delegation info.  That is, if mds >= 0 || ndist > 0.
240  */
241 #define CEPH_MAX_DIRFRAG_REP 4
242 
243 struct ceph_inode_frag {
244 	struct rb_node node;
245 
246 	/* fragtree state */
247 	u32 frag;
248 	int split_by;         /* i.e. 2^(split_by) children */
249 
250 	/* delegation and replication info */
251 	int mds;              /* -1 if same authority as parent */
252 	int ndist;            /* >0 if replicated */
253 	int dist[CEPH_MAX_DIRFRAG_REP];
254 };
255 
256 /*
257  * We cache inode xattrs as an encoded blob until they are first used,
258  * at which point we parse them into an rbtree.
259  */
260 struct ceph_inode_xattr {
261 	struct rb_node node;
262 
263 	const char *name;
264 	int name_len;
265 	const char *val;
266 	int val_len;
267 	int dirty;
268 
269 	int should_free_name;
270 	int should_free_val;
271 };
272 
273 /*
274  * Ceph dentry state
275  */
276 struct ceph_dentry_info {
277 	struct dentry *dentry;
278 	struct ceph_mds_session *lease_session;
279 	struct list_head lease_list;
280 	unsigned flags;
281 	int lease_shared_gen;
282 	u32 lease_gen;
283 	u32 lease_seq;
284 	unsigned long lease_renew_after, lease_renew_from;
285 	unsigned long time;
286 	u64 offset;
287 };
288 
289 #define CEPH_DENTRY_REFERENCED		1
290 #define CEPH_DENTRY_LEASE_LIST		2
291 #define CEPH_DENTRY_SHRINK_LIST		4
292 #define CEPH_DENTRY_PRIMARY_LINK	8
293 
294 struct ceph_inode_xattrs_info {
295 	/*
296 	 * (still encoded) xattr blob. we avoid the overhead of parsing
297 	 * this until someone actually calls getxattr, etc.
298 	 *
299 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
300 	 * NULL means we don't know.
301 	*/
302 	struct ceph_buffer *blob, *prealloc_blob;
303 
304 	struct rb_root index;
305 	bool dirty;
306 	int count;
307 	int names_size;
308 	int vals_size;
309 	u64 version, index_version;
310 };
311 
312 /*
313  * Ceph inode.
314  */
315 struct ceph_inode_info {
316 	struct ceph_vino i_vino;   /* ceph ino + snap */
317 
318 	spinlock_t i_ceph_lock;
319 
320 	u64 i_version;
321 	u64 i_inline_version;
322 	u32 i_time_warp_seq;
323 
324 	unsigned long i_ceph_flags;
325 	atomic64_t i_release_count;
326 	atomic64_t i_ordered_count;
327 	atomic64_t i_complete_seq[2];
328 
329 	struct ceph_dir_layout i_dir_layout;
330 	struct ceph_file_layout i_layout;
331 	struct ceph_file_layout i_cached_layout;	// for async creates
332 	char *i_symlink;
333 
334 	/* for dirs */
335 	struct timespec64 i_rctime;
336 	u64 i_rbytes, i_rfiles, i_rsubdirs;
337 	u64 i_files, i_subdirs;
338 
339 	/* quotas */
340 	u64 i_max_bytes, i_max_files;
341 
342 	s32 i_dir_pin;
343 
344 	struct rb_root i_fragtree;
345 	int i_fragtree_nsplits;
346 	struct mutex i_fragtree_mutex;
347 
348 	struct ceph_inode_xattrs_info i_xattrs;
349 
350 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
351 	 * s_mutex. */
352 	struct rb_root i_caps;           /* cap list */
353 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
354 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
355 
356 	/*
357 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
358 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
359 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
360 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
361 	 * be tested under the i_ceph_lock. Changing anything requires both.
362 	 */
363 	struct list_head i_dirty_item;
364 
365 	/*
366 	 * Link to session's s_cap_flushing list. Protected in a similar
367 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
368 	 * s_cap_flushing list can be walked while holding either the s_mutex
369 	 * or msdc->cap_dirty_lock. List presence can also be checked while
370 	 * holding the i_ceph_lock for this inode.
371 	 */
372 	struct list_head i_flushing_item;
373 
374 	/* we need to track cap writeback on a per-cap-bit basis, to allow
375 	 * overlapping, pipelined cap flushes to the mds.  we can probably
376 	 * reduce the tid to 8 bits if we're concerned about inode size. */
377 	struct ceph_cap_flush *i_prealloc_cap_flush;
378 	struct list_head i_cap_flush_list;
379 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
380 	unsigned long i_hold_caps_max; /* jiffies */
381 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
382 	struct ceph_cap_reservation i_cap_migration_resv;
383 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
384 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
385 						    dirty|flushing caps */
386 	unsigned i_snap_caps;           /* cap bits for snapped files */
387 
388 	unsigned long i_last_rd;
389 	unsigned long i_last_wr;
390 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
391 
392 	struct mutex i_truncate_mutex;
393 	u32 i_truncate_seq;        /* last truncate to smaller size */
394 	u64 i_truncate_size;       /*  and the size we last truncated down to */
395 	int i_truncate_pending;    /*  still need to call vmtruncate */
396 
397 	u64 i_max_size;            /* max file size authorized by mds */
398 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
399 	u64 i_wanted_max_size;     /* offset we'd like to write too */
400 	u64 i_requested_max_size;  /* max_size we've requested */
401 
402 	/* held references to caps */
403 	int i_pin_ref;
404 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
405 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
406 	atomic_t i_filelock_ref;
407 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
408 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
409 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
410 
411 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
412 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
413 	spinlock_t i_unsafe_lock;
414 
415 	union {
416 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
417 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
418 	};
419 	int i_snap_realm_counter; /* snap realm (if caps) */
420 	struct list_head i_snap_realm_item;
421 	struct list_head i_snap_flush_item;
422 	struct timespec64 i_btime;
423 	struct timespec64 i_snap_btime;
424 
425 	struct work_struct i_work;
426 	unsigned long  i_work_mask;
427 
428 #ifdef CONFIG_CEPH_FSCACHE
429 	struct fscache_cookie *fscache;
430 	u32 i_fscache_gen;
431 #endif
432 	errseq_t i_meta_err;
433 
434 	struct inode vfs_inode; /* at end */
435 };
436 
437 static inline struct ceph_inode_info *
438 ceph_inode(const struct inode *inode)
439 {
440 	return container_of(inode, struct ceph_inode_info, vfs_inode);
441 }
442 
443 static inline struct ceph_fs_client *
444 ceph_inode_to_client(const struct inode *inode)
445 {
446 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
447 }
448 
449 static inline struct ceph_fs_client *
450 ceph_sb_to_client(const struct super_block *sb)
451 {
452 	return (struct ceph_fs_client *)sb->s_fs_info;
453 }
454 
455 static inline struct ceph_mds_client *
456 ceph_sb_to_mdsc(const struct super_block *sb)
457 {
458 	return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
459 }
460 
461 static inline struct ceph_vino
462 ceph_vino(const struct inode *inode)
463 {
464 	return ceph_inode(inode)->i_vino;
465 }
466 
467 static inline u32 ceph_ino_to_ino32(u64 vino)
468 {
469 	u32 ino = vino & 0xffffffff;
470 	ino ^= vino >> 32;
471 	if (!ino)
472 		ino = 2;
473 	return ino;
474 }
475 
476 /*
477  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
478  * some arches. We generally do not use this value inside the ceph driver, but
479  * we do want to set it to something, so that generic vfs code has an
480  * appropriate value for tracepoints and the like.
481  */
482 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
483 {
484 	if (sizeof(ino_t) == sizeof(u32))
485 		return ceph_ino_to_ino32(vino.ino);
486 	return (ino_t)vino.ino;
487 }
488 
489 /* for printf-style formatting */
490 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
491 
492 static inline u64 ceph_ino(struct inode *inode)
493 {
494 	return ceph_inode(inode)->i_vino.ino;
495 }
496 
497 static inline u64 ceph_snap(struct inode *inode)
498 {
499 	return ceph_inode(inode)->i_vino.snap;
500 }
501 
502 /**
503  * ceph_present_ino - format an inode number for presentation to userland
504  * @sb: superblock where the inode lives
505  * @ino: inode number to (possibly) convert
506  *
507  * If the user mounted with the ino32 option, then the 64-bit value needs
508  * to be converted to something that can fit inside 32 bits. Note that
509  * internal kernel code never uses this value, so this is entirely for
510  * userland consumption.
511  */
512 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
513 {
514 	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
515 		return ceph_ino_to_ino32(ino);
516 	return ino;
517 }
518 
519 static inline u64 ceph_present_inode(struct inode *inode)
520 {
521 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
522 }
523 
524 static inline int ceph_ino_compare(struct inode *inode, void *data)
525 {
526 	struct ceph_vino *pvino = (struct ceph_vino *)data;
527 	struct ceph_inode_info *ci = ceph_inode(inode);
528 	return ci->i_vino.ino == pvino->ino &&
529 		ci->i_vino.snap == pvino->snap;
530 }
531 
532 
533 static inline struct inode *ceph_find_inode(struct super_block *sb,
534 					    struct ceph_vino vino)
535 {
536 	/*
537 	 * NB: The hashval will be run through the fs/inode.c hash function
538 	 * anyway, so there is no need to squash the inode number down to
539 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
540 	 */
541 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
542 }
543 
544 
545 /*
546  * Ceph inode.
547  */
548 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
549 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
550 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
551 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
552 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
553 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
554 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
555 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
556 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
557 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
558 #define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
559 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
560 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
561 
562 /*
563  * Masks of ceph inode work.
564  */
565 #define CEPH_I_WORK_WRITEBACK		0
566 #define CEPH_I_WORK_INVALIDATE_PAGES	1
567 #define CEPH_I_WORK_VMTRUNCATE		2
568 #define CEPH_I_WORK_CHECK_CAPS		3
569 #define CEPH_I_WORK_FLUSH_SNAPS		4
570 
571 /*
572  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
573  * and then clear it when they start succeeding. Note that we do a lockless
574  * check first, and only take the lock if it looks like it needs to be changed.
575  * The write submission code just takes this as a hint, so we're not too
576  * worried if a few slip through in either direction.
577  */
578 static inline void ceph_set_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_clear_error_write(struct ceph_inode_info *ci)
588 {
589 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
590 		spin_lock(&ci->i_ceph_lock);
591 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
592 		spin_unlock(&ci->i_ceph_lock);
593 	}
594 }
595 
596 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
597 					   long long release_count,
598 					   long long ordered_count)
599 {
600 	/*
601 	 * Makes sure operations that setup readdir cache (update page
602 	 * cache and i_size) are strongly ordered w.r.t. the following
603 	 * atomic64_set() operations.
604 	 */
605 	smp_mb();
606 	atomic64_set(&ci->i_complete_seq[0], release_count);
607 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
608 }
609 
610 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
611 {
612 	atomic64_inc(&ci->i_release_count);
613 }
614 
615 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
616 {
617 	atomic64_inc(&ci->i_ordered_count);
618 }
619 
620 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
621 {
622 	return atomic64_read(&ci->i_complete_seq[0]) ==
623 		atomic64_read(&ci->i_release_count);
624 }
625 
626 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
627 {
628 	return  atomic64_read(&ci->i_complete_seq[0]) ==
629 		atomic64_read(&ci->i_release_count) &&
630 		atomic64_read(&ci->i_complete_seq[1]) ==
631 		atomic64_read(&ci->i_ordered_count);
632 }
633 
634 static inline void ceph_dir_clear_complete(struct inode *inode)
635 {
636 	__ceph_dir_clear_complete(ceph_inode(inode));
637 }
638 
639 static inline void ceph_dir_clear_ordered(struct inode *inode)
640 {
641 	__ceph_dir_clear_ordered(ceph_inode(inode));
642 }
643 
644 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
645 {
646 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
647 	smp_rmb();
648 	return ret;
649 }
650 
651 /* find a specific frag @f */
652 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
653 						u32 f);
654 
655 /*
656  * choose fragment for value @v.  copy frag content to pfrag, if leaf
657  * exists
658  */
659 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
660 			    struct ceph_inode_frag *pfrag,
661 			    int *found);
662 
663 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
664 {
665 	return (struct ceph_dentry_info *)dentry->d_fsdata;
666 }
667 
668 /*
669  * caps helpers
670  */
671 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
672 {
673 	return !RB_EMPTY_ROOT(&ci->i_caps);
674 }
675 
676 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
677 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
678 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
679 					  int t);
680 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
681 				    struct ceph_cap *cap);
682 
683 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
684 {
685 	int issued;
686 	spin_lock(&ci->i_ceph_lock);
687 	issued = __ceph_caps_issued(ci, NULL);
688 	spin_unlock(&ci->i_ceph_lock);
689 	return issued;
690 }
691 
692 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
693 					       int mask, int touch)
694 {
695 	int r;
696 	spin_lock(&ci->i_ceph_lock);
697 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
698 	spin_unlock(&ci->i_ceph_lock);
699 	return r;
700 }
701 
702 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
703 {
704 	return ci->i_dirty_caps | ci->i_flushing_caps;
705 }
706 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
707 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
708 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
709 				  struct ceph_cap_flush **pcf);
710 
711 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
712 				      struct ceph_cap *ocap, int mask);
713 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
714 extern int __ceph_caps_used(struct ceph_inode_info *ci);
715 
716 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
717 {
718 	return ci->i_nr_by_mode[0];
719 }
720 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
721 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
722 
723 /* what the mds thinks we want */
724 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
725 
726 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
727 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
728 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
729 				     struct ceph_mount_options *fsopt);
730 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
731 			     struct ceph_cap_reservation *ctx, int need);
732 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
733 			       struct ceph_cap_reservation *ctx);
734 extern void ceph_reservation_status(struct ceph_fs_client *client,
735 				    int *total, int *avail, int *used,
736 				    int *reserved, int *min);
737 
738 
739 
740 /*
741  * we keep buffered readdir results attached to file->private_data
742  */
743 #define CEPH_F_SYNC     1
744 #define CEPH_F_ATEND    2
745 
746 struct ceph_file_info {
747 	short fmode;     /* initialized on open */
748 	short flags;     /* CEPH_F_* */
749 
750 	spinlock_t rw_contexts_lock;
751 	struct list_head rw_contexts;
752 
753 	errseq_t meta_err;
754 	u32 filp_gen;
755 	atomic_t num_locks;
756 };
757 
758 struct ceph_dir_file_info {
759 	struct ceph_file_info file_info;
760 
761 	/* readdir: position within the dir */
762 	u32 frag;
763 	struct ceph_mds_request *last_readdir;
764 
765 	/* readdir: position within a frag */
766 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
767 	char *last_name;       /* last entry in previous chunk */
768 	long long dir_release_count;
769 	long long dir_ordered_count;
770 	int readdir_cache_idx;
771 
772 	/* used for -o dirstat read() on directory thing */
773 	char *dir_info;
774 	int dir_info_len;
775 };
776 
777 struct ceph_rw_context {
778 	struct list_head list;
779 	struct task_struct *thread;
780 	int caps;
781 };
782 
783 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
784 	struct ceph_rw_context _name = {	\
785 		.thread = current,		\
786 		.caps = _caps,			\
787 	}
788 
789 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
790 				       struct ceph_rw_context *ctx)
791 {
792 	spin_lock(&cf->rw_contexts_lock);
793 	list_add(&ctx->list, &cf->rw_contexts);
794 	spin_unlock(&cf->rw_contexts_lock);
795 }
796 
797 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
798 				       struct ceph_rw_context *ctx)
799 {
800 	spin_lock(&cf->rw_contexts_lock);
801 	list_del(&ctx->list);
802 	spin_unlock(&cf->rw_contexts_lock);
803 }
804 
805 static inline struct ceph_rw_context*
806 ceph_find_rw_context(struct ceph_file_info *cf)
807 {
808 	struct ceph_rw_context *ctx, *found = NULL;
809 	spin_lock(&cf->rw_contexts_lock);
810 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
811 		if (ctx->thread == current) {
812 			found = ctx;
813 			break;
814 		}
815 	}
816 	spin_unlock(&cf->rw_contexts_lock);
817 	return found;
818 }
819 
820 struct ceph_readdir_cache_control {
821 	struct page  *page;
822 	struct dentry **dentries;
823 	int index;
824 };
825 
826 /*
827  * A "snap realm" describes a subset of the file hierarchy sharing
828  * the same set of snapshots that apply to it.  The realms themselves
829  * are organized into a hierarchy, such that children inherit (some of)
830  * the snapshots of their parents.
831  *
832  * All inodes within the realm that have capabilities are linked into a
833  * per-realm list.
834  */
835 struct ceph_snap_realm {
836 	u64 ino;
837 	struct inode *inode;
838 	atomic_t nref;
839 	struct rb_node node;
840 
841 	u64 created, seq;
842 	u64 parent_ino;
843 	u64 parent_since;   /* snapid when our current parent became so */
844 
845 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
846 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
847 	u64 *snaps;                   /* snaps specific to this realm */
848 	u32 num_snaps;
849 
850 	struct ceph_snap_realm *parent;
851 	struct list_head children;       /* list of child realms */
852 	struct list_head child_item;
853 
854 	struct list_head empty_item;     /* if i have ref==0 */
855 
856 	struct list_head dirty_item;     /* if realm needs new context */
857 
858 	/* the current set of snaps for this realm */
859 	struct ceph_snap_context *cached_context;
860 
861 	struct list_head inodes_with_caps;
862 	spinlock_t inodes_with_caps_lock;
863 };
864 
865 static inline int default_congestion_kb(void)
866 {
867 	int congestion_kb;
868 
869 	/*
870 	 * Copied from NFS
871 	 *
872 	 * congestion size, scale with available memory.
873 	 *
874 	 *  64MB:    8192k
875 	 * 128MB:   11585k
876 	 * 256MB:   16384k
877 	 * 512MB:   23170k
878 	 *   1GB:   32768k
879 	 *   2GB:   46340k
880 	 *   4GB:   65536k
881 	 *   8GB:   92681k
882 	 *  16GB:  131072k
883 	 *
884 	 * This allows larger machines to have larger/more transfers.
885 	 * Limit the default to 256M
886 	 */
887 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
888 	if (congestion_kb > 256*1024)
889 		congestion_kb = 256*1024;
890 
891 	return congestion_kb;
892 }
893 
894 
895 /* super.c */
896 extern int ceph_force_reconnect(struct super_block *sb);
897 /* snap.c */
898 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
899 					       u64 ino);
900 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
901 				struct ceph_snap_realm *realm);
902 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
903 				struct ceph_snap_realm *realm);
904 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
905 				  void *p, void *e, bool deletion,
906 				  struct ceph_snap_realm **realm_ret);
907 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
908 			     struct ceph_mds_session *session,
909 			     struct ceph_msg *msg);
910 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
911 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
912 				  struct ceph_cap_snap *capsnap);
913 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
914 
915 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
916 						   u64 snap);
917 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
918 				struct ceph_snapid_map *sm);
919 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
920 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
921 
922 
923 /*
924  * a cap_snap is "pending" if it is still awaiting an in-progress
925  * sync write (that may/may not still update size, mtime, etc.).
926  */
927 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
928 {
929 	return !list_empty(&ci->i_cap_snaps) &&
930 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
931 			       ci_item)->writing;
932 }
933 
934 /* inode.c */
935 struct ceph_mds_reply_info_in;
936 struct ceph_mds_reply_dirfrag;
937 
938 extern const struct inode_operations ceph_file_iops;
939 
940 extern struct inode *ceph_alloc_inode(struct super_block *sb);
941 extern void ceph_evict_inode(struct inode *inode);
942 extern void ceph_free_inode(struct inode *inode);
943 
944 extern struct inode *ceph_get_inode(struct super_block *sb,
945 				    struct ceph_vino vino);
946 extern struct inode *ceph_get_snapdir(struct inode *parent);
947 extern int ceph_fill_file_size(struct inode *inode, int issued,
948 			       u32 truncate_seq, u64 truncate_size, u64 size);
949 extern void ceph_fill_file_time(struct inode *inode, int issued,
950 				u64 time_warp_seq, struct timespec64 *ctime,
951 				struct timespec64 *mtime,
952 				struct timespec64 *atime);
953 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
954 		    struct ceph_mds_reply_info_in *iinfo,
955 		    struct ceph_mds_reply_dirfrag *dirinfo,
956 		    struct ceph_mds_session *session, int cap_fmode,
957 		    struct ceph_cap_reservation *caps_reservation);
958 extern int ceph_fill_trace(struct super_block *sb,
959 			   struct ceph_mds_request *req);
960 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
961 				    struct ceph_mds_session *session);
962 
963 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
964 
965 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
966 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
967 
968 extern void ceph_async_iput(struct inode *inode);
969 
970 void ceph_queue_inode_work(struct inode *inode, int work_bit);
971 
972 static inline void ceph_queue_vmtruncate(struct inode *inode)
973 {
974 	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
975 }
976 
977 static inline void ceph_queue_invalidate(struct inode *inode)
978 {
979 	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
980 }
981 
982 static inline void ceph_queue_writeback(struct inode *inode)
983 {
984 	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
985 }
986 
987 static inline void ceph_queue_check_caps(struct inode *inode)
988 {
989 	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
990 }
991 
992 static inline void ceph_queue_flush_snaps(struct inode *inode)
993 {
994 	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
995 }
996 
997 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
998 			     int mask, bool force);
999 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1000 {
1001 	return __ceph_do_getattr(inode, NULL, mask, force);
1002 }
1003 extern int ceph_permission(struct user_namespace *mnt_userns,
1004 			   struct inode *inode, int mask);
1005 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1006 extern int ceph_setattr(struct user_namespace *mnt_userns,
1007 			struct dentry *dentry, struct iattr *attr);
1008 extern int ceph_getattr(struct user_namespace *mnt_userns,
1009 			const struct path *path, struct kstat *stat,
1010 			u32 request_mask, unsigned int flags);
1011 
1012 /* xattr.c */
1013 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1014 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1015 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1016 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1017 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1018 extern const struct xattr_handler *ceph_xattr_handlers[];
1019 
1020 struct ceph_acl_sec_ctx {
1021 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1022 	void *default_acl;
1023 	void *acl;
1024 #endif
1025 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1026 	void *sec_ctx;
1027 	u32 sec_ctxlen;
1028 #endif
1029 	struct ceph_pagelist *pagelist;
1030 };
1031 
1032 #ifdef CONFIG_SECURITY
1033 extern bool ceph_security_xattr_deadlock(struct inode *in);
1034 extern bool ceph_security_xattr_wanted(struct inode *in);
1035 #else
1036 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1037 {
1038 	return false;
1039 }
1040 static inline bool ceph_security_xattr_wanted(struct inode *in)
1041 {
1042 	return false;
1043 }
1044 #endif
1045 
1046 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1047 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1048 				     struct ceph_acl_sec_ctx *ctx);
1049 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1050 {
1051 	security_inode_invalidate_secctx(inode);
1052 }
1053 #else
1054 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1055 					    struct ceph_acl_sec_ctx *ctx)
1056 {
1057 	return 0;
1058 }
1059 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1060 {
1061 }
1062 #endif
1063 
1064 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1065 
1066 /* acl.c */
1067 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1068 
1069 struct posix_acl *ceph_get_acl(struct inode *, int);
1070 int ceph_set_acl(struct user_namespace *mnt_userns,
1071 		 struct inode *inode, struct posix_acl *acl, int type);
1072 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1073 		       struct ceph_acl_sec_ctx *as_ctx);
1074 void ceph_init_inode_acls(struct inode *inode,
1075 			  struct ceph_acl_sec_ctx *as_ctx);
1076 
1077 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1078 {
1079        forget_all_cached_acls(inode);
1080 }
1081 
1082 #else
1083 
1084 #define ceph_get_acl NULL
1085 #define ceph_set_acl NULL
1086 
1087 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1088 				     struct ceph_acl_sec_ctx *as_ctx)
1089 {
1090 	return 0;
1091 }
1092 static inline void ceph_init_inode_acls(struct inode *inode,
1093 					struct ceph_acl_sec_ctx *as_ctx)
1094 {
1095 }
1096 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1097 {
1098 	return 0;
1099 }
1100 
1101 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1102 {
1103 }
1104 
1105 #endif
1106 
1107 /* caps.c */
1108 extern const char *ceph_cap_string(int c);
1109 extern void ceph_handle_caps(struct ceph_mds_session *session,
1110 			     struct ceph_msg *msg);
1111 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1112 				     struct ceph_cap_reservation *ctx);
1113 extern void ceph_add_cap(struct inode *inode,
1114 			 struct ceph_mds_session *session, u64 cap_id,
1115 			 unsigned issued, unsigned wanted,
1116 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1117 			 struct ceph_cap **new_cap);
1118 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1119 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1120 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1121 			 struct ceph_cap *cap);
1122 extern int ceph_is_any_caps(struct inode *inode);
1123 
1124 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1125 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1126 		      int datasync);
1127 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1128 					  struct ceph_mds_session *session);
1129 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1130 				    struct ceph_mds_session *session);
1131 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1132 				   struct ceph_inode_info *ci);
1133 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1134 					     int mds);
1135 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1136 				bool snap_rwsem_locked);
1137 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1138 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1139 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1140 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1141 					    int had);
1142 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1143 				       struct ceph_snap_context *snapc);
1144 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1145 			     struct ceph_mds_session **psession);
1146 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1147 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1148 			    struct ceph_mds_session *session);
1149 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1150 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1151 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1152 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1153 				     int mds, int drop, int unless, int force);
1154 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1155 				      struct inode *dir,
1156 				      int mds, int drop, int unless);
1157 
1158 extern int ceph_get_caps(struct file *filp, int need, int want,
1159 			 loff_t endoff, int *got, struct page **pinned_page);
1160 extern int ceph_try_get_caps(struct inode *inode,
1161 			     int need, int want, bool nonblock, int *got);
1162 
1163 /* for counting open files by mode */
1164 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1165 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1166 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1167 			       struct ceph_mds_client *mdsc, int fmode);
1168 
1169 /* addr.c */
1170 extern const struct address_space_operations ceph_aops;
1171 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1172 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1173 extern int ceph_pool_perm_check(struct inode *inode, int need);
1174 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1175 
1176 /* file.c */
1177 extern const struct file_operations ceph_file_fops;
1178 
1179 extern int ceph_renew_caps(struct inode *inode, int fmode);
1180 extern int ceph_open(struct inode *inode, struct file *file);
1181 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1182 			    struct file *file, unsigned flags, umode_t mode);
1183 extern int ceph_release(struct inode *inode, struct file *filp);
1184 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1185 				  char *data, size_t len);
1186 
1187 /* dir.c */
1188 extern const struct file_operations ceph_dir_fops;
1189 extern const struct file_operations ceph_snapdir_fops;
1190 extern const struct inode_operations ceph_dir_iops;
1191 extern const struct inode_operations ceph_snapdir_iops;
1192 extern const struct dentry_operations ceph_dentry_ops;
1193 
1194 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1195 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1196 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1197 			       struct dentry *dentry, int err);
1198 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1199 					 struct dentry *dentry, int err);
1200 
1201 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1202 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1203 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1204 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1205 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1206 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1207 
1208 /* ioctl.c */
1209 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1210 
1211 /* export.c */
1212 extern const struct export_operations ceph_export_ops;
1213 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1214 
1215 /* locks.c */
1216 extern __init void ceph_flock_init(void);
1217 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1218 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1219 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1220 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1221 				       struct ceph_filelock *flocks,
1222 				       int num_fcntl_locks,
1223 				       int num_flock_locks);
1224 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1225 				  struct ceph_pagelist *pagelist,
1226 				  int num_fcntl_locks, int num_flock_locks);
1227 
1228 /* debugfs.c */
1229 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1230 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1231 
1232 /* quota.c */
1233 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1234 {
1235 	return ci->i_max_files || ci->i_max_bytes;
1236 }
1237 
1238 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1239 
1240 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1241 				       u64 max_bytes, u64 max_files)
1242 {
1243 	bool had_quota, has_quota;
1244 	had_quota = __ceph_has_any_quota(ci);
1245 	ci->i_max_bytes = max_bytes;
1246 	ci->i_max_files = max_files;
1247 	has_quota = __ceph_has_any_quota(ci);
1248 
1249 	if (had_quota != has_quota)
1250 		ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1251 }
1252 
1253 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1254 			      struct ceph_mds_session *session,
1255 			      struct ceph_msg *msg);
1256 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1257 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1258 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1259 					     loff_t newlen);
1260 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1261 						loff_t newlen);
1262 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1263 				     struct kstatfs *buf);
1264 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1265 
1266 #endif /* _FS_CEPH_SUPER_H */
1267