xref: /linux/fs/ceph/super.h (revision d4f6b31d721779d91b5e2f8072478af73b196c34)
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 #define FSCACHE_USE_NEW_IO_API
25 #include <linux/fscache.h>
26 #endif
27 
28 /* f_type in struct statfs */
29 #define CEPH_SUPER_MAGIC 0x00c36400
30 
31 /* large granularity for statfs utilization stats to facilitate
32  * large volume sizes on 32-bit machines. */
33 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
34 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
35 
36 #define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
37 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
38 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
39 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
40 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
41 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
42 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
43 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
44 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
45 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
46 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
47 #define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
48 
49 #define CEPH_MOUNT_OPT_DEFAULT			\
50 	(CEPH_MOUNT_OPT_DCACHE |		\
51 	 CEPH_MOUNT_OPT_NOCOPYFROM)
52 
53 #define ceph_set_mount_opt(fsc, opt) \
54 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
55 #define ceph_clear_mount_opt(fsc, opt) \
56 	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
57 #define ceph_test_mount_opt(fsc, opt) \
58 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
59 
60 /* max size of osd read request, limited by libceph */
61 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
62 /* osd has a configurable limitaion of max write size.
63  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
64 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
65 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
66 #define CEPH_MAX_READDIR_DEFAULT        1024
67 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
68 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
69 
70 /*
71  * Delay telling the MDS we no longer want caps, in case we reopen
72  * the file.  Delay a minimum amount of time, even if we send a cap
73  * message for some other reason.  Otherwise, take the oppotunity to
74  * update the mds to avoid sending another message later.
75  */
76 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
77 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
78 
79 struct ceph_mount_options {
80 	unsigned int flags;
81 
82 	unsigned int wsize;            /* max write size */
83 	unsigned int rsize;            /* max read size */
84 	unsigned int rasize;           /* max readahead */
85 	unsigned int congestion_kb;    /* max writeback in flight */
86 	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
87 	int caps_max;
88 	unsigned int max_readdir;       /* max readdir result (entries) */
89 	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
90 
91 	/*
92 	 * everything above this point can be memcmp'd; everything below
93 	 * is handled in compare_mount_options()
94 	 */
95 
96 	char *snapdir_name;   /* default ".snap" */
97 	char *mds_namespace;  /* default NULL */
98 	char *server_path;    /* default NULL (means "/") */
99 	char *fscache_uniq;   /* default NULL */
100 };
101 
102 struct ceph_fs_client {
103 	struct super_block *sb;
104 
105 	struct list_head metric_wakeup;
106 
107 	struct ceph_mount_options *mount_options;
108 	struct ceph_client *client;
109 
110 	int mount_state;
111 
112 	bool blocklisted;
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_status;
133 	struct dentry *debugfs_mds_sessions;
134 #endif
135 
136 #ifdef CONFIG_CEPH_FSCACHE
137 	struct fscache_cookie *fscache;
138 #endif
139 };
140 
141 
142 /*
143  * File i/o capability.  This tracks shared state with the metadata
144  * server that allows us to cache or writeback attributes or to read
145  * and write data.  For any given inode, we should have one or more
146  * capabilities, one issued by each metadata server, and our
147  * cumulative access is the OR of all issued capabilities.
148  *
149  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
150  * session capability lists.
151  */
152 struct ceph_cap {
153 	struct ceph_inode_info *ci;
154 	struct rb_node ci_node;          /* per-ci cap tree */
155 	struct ceph_mds_session *session;
156 	struct list_head session_caps;   /* per-session caplist */
157 	u64 cap_id;       /* unique cap id (mds provided) */
158 	union {
159 		/* in-use caps */
160 		struct {
161 			int issued;       /* latest, from the mds */
162 			int implemented;  /* implemented superset of
163 					     issued (for revocation) */
164 			int mds;	  /* mds index for this cap */
165 			int mds_wanted;   /* caps wanted from this mds */
166 		};
167 		/* caps to release */
168 		struct {
169 			u64 cap_ino;
170 			int queue_release;
171 		};
172 	};
173 	u32 seq, issue_seq, mseq;
174 	u32 cap_gen;      /* active/stale cycle */
175 	unsigned long last_used;
176 	struct list_head caps_item;
177 };
178 
179 #define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
180 #define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
181 #define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
182 
183 struct ceph_cap_flush {
184 	u64 tid;
185 	int caps; /* 0 means capsnap */
186 	bool wake; /* wake up flush waiters when finish ? */
187 	struct list_head g_list; // global
188 	struct list_head i_list; // per inode
189 };
190 
191 /*
192  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
193  * we first complete any in-process sync writes and writeback any dirty
194  * data before flushing the snapped state (tracked here) back to the MDS.
195  */
196 struct ceph_cap_snap {
197 	refcount_t nref;
198 	struct list_head ci_item;
199 
200 	struct ceph_cap_flush cap_flush;
201 
202 	u64 follows;
203 	int issued, dirty;
204 	struct ceph_snap_context *context;
205 
206 	umode_t mode;
207 	kuid_t uid;
208 	kgid_t gid;
209 
210 	struct ceph_buffer *xattr_blob;
211 	u64 xattr_version;
212 
213 	u64 size;
214 	u64 change_attr;
215 	struct timespec64 mtime, atime, ctime, btime;
216 	u64 time_warp_seq;
217 	u64 truncate_size;
218 	u32 truncate_seq;
219 	int writing;   /* a sync write is still in progress */
220 	int dirty_pages;     /* dirty pages awaiting writeback */
221 	bool inline_data;
222 	bool need_flush;
223 };
224 
225 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
226 {
227 	if (refcount_dec_and_test(&capsnap->nref)) {
228 		if (capsnap->xattr_blob)
229 			ceph_buffer_put(capsnap->xattr_blob);
230 		kfree(capsnap);
231 	}
232 }
233 
234 /*
235  * The frag tree describes how a directory is fragmented, potentially across
236  * multiple metadata servers.  It is also used to indicate points where
237  * metadata authority is delegated, and whether/where metadata is replicated.
238  *
239  * A _leaf_ frag will be present in the i_fragtree IFF there is
240  * delegation info.  That is, if mds >= 0 || ndist > 0.
241  */
242 #define CEPH_MAX_DIRFRAG_REP 4
243 
244 struct ceph_inode_frag {
245 	struct rb_node node;
246 
247 	/* fragtree state */
248 	u32 frag;
249 	int split_by;         /* i.e. 2^(split_by) children */
250 
251 	/* delegation and replication info */
252 	int mds;              /* -1 if same authority as parent */
253 	int ndist;            /* >0 if replicated */
254 	int dist[CEPH_MAX_DIRFRAG_REP];
255 };
256 
257 /*
258  * We cache inode xattrs as an encoded blob until they are first used,
259  * at which point we parse them into an rbtree.
260  */
261 struct ceph_inode_xattr {
262 	struct rb_node node;
263 
264 	const char *name;
265 	int name_len;
266 	const char *val;
267 	int val_len;
268 	int dirty;
269 
270 	int should_free_name;
271 	int should_free_val;
272 };
273 
274 /*
275  * Ceph dentry state
276  */
277 struct ceph_dentry_info {
278 	struct dentry *dentry;
279 	struct ceph_mds_session *lease_session;
280 	struct list_head lease_list;
281 	unsigned flags;
282 	int lease_shared_gen;
283 	u32 lease_gen;
284 	u32 lease_seq;
285 	unsigned long lease_renew_after, lease_renew_from;
286 	unsigned long time;
287 	u64 offset;
288 };
289 
290 #define CEPH_DENTRY_REFERENCED		1
291 #define CEPH_DENTRY_LEASE_LIST		2
292 #define CEPH_DENTRY_SHRINK_LIST		4
293 #define CEPH_DENTRY_PRIMARY_LINK	8
294 
295 struct ceph_inode_xattrs_info {
296 	/*
297 	 * (still encoded) xattr blob. we avoid the overhead of parsing
298 	 * this until someone actually calls getxattr, etc.
299 	 *
300 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
301 	 * NULL means we don't know.
302 	*/
303 	struct ceph_buffer *blob, *prealloc_blob;
304 
305 	struct rb_root index;
306 	bool dirty;
307 	int count;
308 	int names_size;
309 	int vals_size;
310 	u64 version, index_version;
311 };
312 
313 /*
314  * Ceph inode.
315  */
316 struct ceph_inode_info {
317 	struct ceph_vino i_vino;   /* ceph ino + snap */
318 
319 	spinlock_t i_ceph_lock;
320 
321 	u64 i_version;
322 	u64 i_inline_version;
323 	u32 i_time_warp_seq;
324 
325 	unsigned long i_ceph_flags;
326 	atomic64_t i_release_count;
327 	atomic64_t i_ordered_count;
328 	atomic64_t i_complete_seq[2];
329 
330 	struct ceph_dir_layout i_dir_layout;
331 	struct ceph_file_layout i_layout;
332 	struct ceph_file_layout i_cached_layout;	// for async creates
333 	char *i_symlink;
334 
335 	/* for dirs */
336 	struct timespec64 i_rctime;
337 	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
338 	u64 i_files, i_subdirs;
339 
340 	/* quotas */
341 	u64 i_max_bytes, i_max_files;
342 
343 	s32 i_dir_pin;
344 
345 	struct rb_root i_fragtree;
346 	int i_fragtree_nsplits;
347 	struct mutex i_fragtree_mutex;
348 
349 	struct ceph_inode_xattrs_info i_xattrs;
350 
351 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
352 	 * s_mutex. */
353 	struct rb_root i_caps;           /* cap list */
354 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
355 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
356 
357 	/*
358 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
359 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
360 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
361 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
362 	 * be tested under the i_ceph_lock. Changing anything requires both.
363 	 */
364 	struct list_head i_dirty_item;
365 
366 	/*
367 	 * Link to session's s_cap_flushing list. Protected in a similar
368 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
369 	 * s_cap_flushing list can be walked while holding either the s_mutex
370 	 * or msdc->cap_dirty_lock. List presence can also be checked while
371 	 * holding the i_ceph_lock for this inode.
372 	 */
373 	struct list_head i_flushing_item;
374 
375 	/* we need to track cap writeback on a per-cap-bit basis, to allow
376 	 * overlapping, pipelined cap flushes to the mds.  we can probably
377 	 * reduce the tid to 8 bits if we're concerned about inode size. */
378 	struct ceph_cap_flush *i_prealloc_cap_flush;
379 	struct list_head i_cap_flush_list;
380 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
381 	unsigned long i_hold_caps_max; /* jiffies */
382 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
383 	struct ceph_cap_reservation i_cap_migration_resv;
384 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
385 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
386 						    dirty|flushing caps */
387 	unsigned i_snap_caps;           /* cap bits for snapped files */
388 
389 	unsigned long i_last_rd;
390 	unsigned long i_last_wr;
391 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
392 
393 	struct mutex i_truncate_mutex;
394 	u32 i_truncate_seq;        /* last truncate to smaller size */
395 	u64 i_truncate_size;       /*  and the size we last truncated down to */
396 	int i_truncate_pending;    /*  still need to call vmtruncate */
397 
398 	u64 i_max_size;            /* max file size authorized by mds */
399 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
400 	u64 i_wanted_max_size;     /* offset we'd like to write too */
401 	u64 i_requested_max_size;  /* max_size we've requested */
402 
403 	/* held references to caps */
404 	int i_pin_ref;
405 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
406 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
407 	atomic_t i_filelock_ref;
408 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
409 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
410 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
411 
412 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
413 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
414 	spinlock_t i_unsafe_lock;
415 
416 	union {
417 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
418 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
419 	};
420 	int i_snap_realm_counter; /* snap realm (if caps) */
421 	struct list_head i_snap_realm_item;
422 	struct list_head i_snap_flush_item;
423 	struct timespec64 i_btime;
424 	struct timespec64 i_snap_btime;
425 
426 	struct work_struct i_work;
427 	unsigned long  i_work_mask;
428 
429 #ifdef CONFIG_CEPH_FSCACHE
430 	struct fscache_cookie *fscache;
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  * The MDS reserves a set of inodes for its own usage. These should never
534  * be accessible by clients, and so the MDS has no reason to ever hand these
535  * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
536  *
537  * These come from src/mds/mdstypes.h in the ceph sources.
538  */
539 #define CEPH_MAX_MDS		0x100
540 #define CEPH_NUM_STRAY		10
541 #define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
542 #define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
543 
544 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
545 {
546 	if (vino.ino < CEPH_INO_SYSTEM_BASE &&
547 	    vino.ino >= CEPH_MDS_INO_MDSDIR_OFFSET) {
548 		WARN_RATELIMIT(1, "Attempt to access reserved inode number 0x%llx", vino.ino);
549 		return true;
550 	}
551 	return false;
552 }
553 
554 static inline struct inode *ceph_find_inode(struct super_block *sb,
555 					    struct ceph_vino vino)
556 {
557 	if (ceph_vino_is_reserved(vino))
558 		return NULL;
559 
560 	/*
561 	 * NB: The hashval will be run through the fs/inode.c hash function
562 	 * anyway, so there is no need to squash the inode number down to
563 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
564 	 */
565 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
566 }
567 
568 
569 /*
570  * Ceph inode.
571  */
572 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
573 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
574 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
575 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
576 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
577 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
578 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
579 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
580 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
581 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
582 #define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
583 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
584 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
585 
586 /*
587  * Masks of ceph inode work.
588  */
589 #define CEPH_I_WORK_WRITEBACK		0
590 #define CEPH_I_WORK_INVALIDATE_PAGES	1
591 #define CEPH_I_WORK_VMTRUNCATE		2
592 #define CEPH_I_WORK_CHECK_CAPS		3
593 #define CEPH_I_WORK_FLUSH_SNAPS		4
594 
595 /*
596  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
597  * and then clear it when they start succeeding. Note that we do a lockless
598  * check first, and only take the lock if it looks like it needs to be changed.
599  * The write submission code just takes this as a hint, so we're not too
600  * worried if a few slip through in either direction.
601  */
602 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
603 {
604 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
605 		spin_lock(&ci->i_ceph_lock);
606 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
607 		spin_unlock(&ci->i_ceph_lock);
608 	}
609 }
610 
611 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
612 {
613 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
614 		spin_lock(&ci->i_ceph_lock);
615 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
616 		spin_unlock(&ci->i_ceph_lock);
617 	}
618 }
619 
620 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
621 					   long long release_count,
622 					   long long ordered_count)
623 {
624 	/*
625 	 * Makes sure operations that setup readdir cache (update page
626 	 * cache and i_size) are strongly ordered w.r.t. the following
627 	 * atomic64_set() operations.
628 	 */
629 	smp_mb();
630 	atomic64_set(&ci->i_complete_seq[0], release_count);
631 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
632 }
633 
634 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
635 {
636 	atomic64_inc(&ci->i_release_count);
637 }
638 
639 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
640 {
641 	atomic64_inc(&ci->i_ordered_count);
642 }
643 
644 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
645 {
646 	return atomic64_read(&ci->i_complete_seq[0]) ==
647 		atomic64_read(&ci->i_release_count);
648 }
649 
650 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
651 {
652 	return  atomic64_read(&ci->i_complete_seq[0]) ==
653 		atomic64_read(&ci->i_release_count) &&
654 		atomic64_read(&ci->i_complete_seq[1]) ==
655 		atomic64_read(&ci->i_ordered_count);
656 }
657 
658 static inline void ceph_dir_clear_complete(struct inode *inode)
659 {
660 	__ceph_dir_clear_complete(ceph_inode(inode));
661 }
662 
663 static inline void ceph_dir_clear_ordered(struct inode *inode)
664 {
665 	__ceph_dir_clear_ordered(ceph_inode(inode));
666 }
667 
668 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
669 {
670 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
671 	smp_rmb();
672 	return ret;
673 }
674 
675 /* find a specific frag @f */
676 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
677 						u32 f);
678 
679 /*
680  * choose fragment for value @v.  copy frag content to pfrag, if leaf
681  * exists
682  */
683 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
684 			    struct ceph_inode_frag *pfrag,
685 			    int *found);
686 
687 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
688 {
689 	return (struct ceph_dentry_info *)dentry->d_fsdata;
690 }
691 
692 /*
693  * caps helpers
694  */
695 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
696 {
697 	return !RB_EMPTY_ROOT(&ci->i_caps);
698 }
699 
700 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
701 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
702 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
703 					  int t);
704 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
705 				    struct ceph_cap *cap);
706 
707 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
708 {
709 	int issued;
710 	spin_lock(&ci->i_ceph_lock);
711 	issued = __ceph_caps_issued(ci, NULL);
712 	spin_unlock(&ci->i_ceph_lock);
713 	return issued;
714 }
715 
716 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
717 					       int mask, int touch)
718 {
719 	int r;
720 	spin_lock(&ci->i_ceph_lock);
721 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
722 	spin_unlock(&ci->i_ceph_lock);
723 	return r;
724 }
725 
726 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
727 {
728 	return ci->i_dirty_caps | ci->i_flushing_caps;
729 }
730 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
731 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
732 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
733 				  struct ceph_cap_flush **pcf);
734 
735 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
736 				      struct ceph_cap *ocap, int mask);
737 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
738 extern int __ceph_caps_used(struct ceph_inode_info *ci);
739 
740 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
741 {
742 	return ci->i_nr_by_mode[0];
743 }
744 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
745 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
746 
747 /* what the mds thinks we want */
748 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
749 
750 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
751 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
752 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
753 				     struct ceph_mount_options *fsopt);
754 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
755 			     struct ceph_cap_reservation *ctx, int need);
756 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
757 			       struct ceph_cap_reservation *ctx);
758 extern void ceph_reservation_status(struct ceph_fs_client *client,
759 				    int *total, int *avail, int *used,
760 				    int *reserved, int *min);
761 
762 
763 
764 /*
765  * we keep buffered readdir results attached to file->private_data
766  */
767 #define CEPH_F_SYNC     1
768 #define CEPH_F_ATEND    2
769 
770 struct ceph_file_info {
771 	short fmode;     /* initialized on open */
772 	short flags;     /* CEPH_F_* */
773 
774 	spinlock_t rw_contexts_lock;
775 	struct list_head rw_contexts;
776 
777 	errseq_t meta_err;
778 	u32 filp_gen;
779 	atomic_t num_locks;
780 };
781 
782 struct ceph_dir_file_info {
783 	struct ceph_file_info file_info;
784 
785 	/* readdir: position within the dir */
786 	u32 frag;
787 	struct ceph_mds_request *last_readdir;
788 
789 	/* readdir: position within a frag */
790 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
791 	char *last_name;       /* last entry in previous chunk */
792 	long long dir_release_count;
793 	long long dir_ordered_count;
794 	int readdir_cache_idx;
795 
796 	/* used for -o dirstat read() on directory thing */
797 	char *dir_info;
798 	int dir_info_len;
799 };
800 
801 struct ceph_rw_context {
802 	struct list_head list;
803 	struct task_struct *thread;
804 	int caps;
805 };
806 
807 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
808 	struct ceph_rw_context _name = {	\
809 		.thread = current,		\
810 		.caps = _caps,			\
811 	}
812 
813 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
814 				       struct ceph_rw_context *ctx)
815 {
816 	spin_lock(&cf->rw_contexts_lock);
817 	list_add(&ctx->list, &cf->rw_contexts);
818 	spin_unlock(&cf->rw_contexts_lock);
819 }
820 
821 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
822 				       struct ceph_rw_context *ctx)
823 {
824 	spin_lock(&cf->rw_contexts_lock);
825 	list_del(&ctx->list);
826 	spin_unlock(&cf->rw_contexts_lock);
827 }
828 
829 static inline struct ceph_rw_context*
830 ceph_find_rw_context(struct ceph_file_info *cf)
831 {
832 	struct ceph_rw_context *ctx, *found = NULL;
833 	spin_lock(&cf->rw_contexts_lock);
834 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
835 		if (ctx->thread == current) {
836 			found = ctx;
837 			break;
838 		}
839 	}
840 	spin_unlock(&cf->rw_contexts_lock);
841 	return found;
842 }
843 
844 struct ceph_readdir_cache_control {
845 	struct page  *page;
846 	struct dentry **dentries;
847 	int index;
848 };
849 
850 /*
851  * A "snap realm" describes a subset of the file hierarchy sharing
852  * the same set of snapshots that apply to it.  The realms themselves
853  * are organized into a hierarchy, such that children inherit (some of)
854  * the snapshots of their parents.
855  *
856  * All inodes within the realm that have capabilities are linked into a
857  * per-realm list.
858  */
859 struct ceph_snap_realm {
860 	u64 ino;
861 	struct inode *inode;
862 	atomic_t nref;
863 	struct rb_node node;
864 
865 	u64 created, seq;
866 	u64 parent_ino;
867 	u64 parent_since;   /* snapid when our current parent became so */
868 
869 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
870 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
871 	u64 *snaps;                   /* snaps specific to this realm */
872 	u32 num_snaps;
873 
874 	struct ceph_snap_realm *parent;
875 	struct list_head children;       /* list of child realms */
876 	struct list_head child_item;
877 
878 	struct list_head empty_item;     /* if i have ref==0 */
879 
880 	struct list_head dirty_item;     /* if realm needs new context */
881 
882 	/* the current set of snaps for this realm */
883 	struct ceph_snap_context *cached_context;
884 
885 	struct list_head inodes_with_caps;
886 	spinlock_t inodes_with_caps_lock;
887 };
888 
889 static inline int default_congestion_kb(void)
890 {
891 	int congestion_kb;
892 
893 	/*
894 	 * Copied from NFS
895 	 *
896 	 * congestion size, scale with available memory.
897 	 *
898 	 *  64MB:    8192k
899 	 * 128MB:   11585k
900 	 * 256MB:   16384k
901 	 * 512MB:   23170k
902 	 *   1GB:   32768k
903 	 *   2GB:   46340k
904 	 *   4GB:   65536k
905 	 *   8GB:   92681k
906 	 *  16GB:  131072k
907 	 *
908 	 * This allows larger machines to have larger/more transfers.
909 	 * Limit the default to 256M
910 	 */
911 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
912 	if (congestion_kb > 256*1024)
913 		congestion_kb = 256*1024;
914 
915 	return congestion_kb;
916 }
917 
918 
919 /* super.c */
920 extern int ceph_force_reconnect(struct super_block *sb);
921 /* snap.c */
922 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
923 					       u64 ino);
924 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
925 				struct ceph_snap_realm *realm);
926 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
927 				struct ceph_snap_realm *realm);
928 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
929 				  void *p, void *e, bool deletion,
930 				  struct ceph_snap_realm **realm_ret);
931 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
932 			     struct ceph_mds_session *session,
933 			     struct ceph_msg *msg);
934 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
935 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
936 				  struct ceph_cap_snap *capsnap);
937 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
938 
939 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
940 						   u64 snap);
941 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
942 				struct ceph_snapid_map *sm);
943 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
944 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
945 
946 
947 /*
948  * a cap_snap is "pending" if it is still awaiting an in-progress
949  * sync write (that may/may not still update size, mtime, etc.).
950  */
951 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
952 {
953 	return !list_empty(&ci->i_cap_snaps) &&
954 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
955 			       ci_item)->writing;
956 }
957 
958 /* inode.c */
959 struct ceph_mds_reply_info_in;
960 struct ceph_mds_reply_dirfrag;
961 
962 extern const struct inode_operations ceph_file_iops;
963 
964 extern struct inode *ceph_alloc_inode(struct super_block *sb);
965 extern void ceph_evict_inode(struct inode *inode);
966 extern void ceph_free_inode(struct inode *inode);
967 
968 extern struct inode *ceph_get_inode(struct super_block *sb,
969 				    struct ceph_vino vino);
970 extern struct inode *ceph_get_snapdir(struct inode *parent);
971 extern int ceph_fill_file_size(struct inode *inode, int issued,
972 			       u32 truncate_seq, u64 truncate_size, u64 size);
973 extern void ceph_fill_file_time(struct inode *inode, int issued,
974 				u64 time_warp_seq, struct timespec64 *ctime,
975 				struct timespec64 *mtime,
976 				struct timespec64 *atime);
977 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
978 		    struct ceph_mds_reply_info_in *iinfo,
979 		    struct ceph_mds_reply_dirfrag *dirinfo,
980 		    struct ceph_mds_session *session, int cap_fmode,
981 		    struct ceph_cap_reservation *caps_reservation);
982 extern int ceph_fill_trace(struct super_block *sb,
983 			   struct ceph_mds_request *req);
984 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
985 				    struct ceph_mds_session *session);
986 
987 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
988 
989 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
990 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
991 
992 extern void ceph_async_iput(struct inode *inode);
993 
994 void ceph_queue_inode_work(struct inode *inode, int work_bit);
995 
996 static inline void ceph_queue_vmtruncate(struct inode *inode)
997 {
998 	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
999 }
1000 
1001 static inline void ceph_queue_invalidate(struct inode *inode)
1002 {
1003 	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1004 }
1005 
1006 static inline void ceph_queue_writeback(struct inode *inode)
1007 {
1008 	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1009 }
1010 
1011 static inline void ceph_queue_check_caps(struct inode *inode)
1012 {
1013 	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1014 }
1015 
1016 static inline void ceph_queue_flush_snaps(struct inode *inode)
1017 {
1018 	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1019 }
1020 
1021 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1022 			     int mask, bool force);
1023 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1024 {
1025 	return __ceph_do_getattr(inode, NULL, mask, force);
1026 }
1027 extern int ceph_permission(struct user_namespace *mnt_userns,
1028 			   struct inode *inode, int mask);
1029 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1030 extern int ceph_setattr(struct user_namespace *mnt_userns,
1031 			struct dentry *dentry, struct iattr *attr);
1032 extern int ceph_getattr(struct user_namespace *mnt_userns,
1033 			const struct path *path, struct kstat *stat,
1034 			u32 request_mask, unsigned int flags);
1035 
1036 /* xattr.c */
1037 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1038 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1039 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1040 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1041 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1042 extern const struct xattr_handler *ceph_xattr_handlers[];
1043 
1044 struct ceph_acl_sec_ctx {
1045 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1046 	void *default_acl;
1047 	void *acl;
1048 #endif
1049 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1050 	void *sec_ctx;
1051 	u32 sec_ctxlen;
1052 #endif
1053 	struct ceph_pagelist *pagelist;
1054 };
1055 
1056 #ifdef CONFIG_SECURITY
1057 extern bool ceph_security_xattr_deadlock(struct inode *in);
1058 extern bool ceph_security_xattr_wanted(struct inode *in);
1059 #else
1060 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1061 {
1062 	return false;
1063 }
1064 static inline bool ceph_security_xattr_wanted(struct inode *in)
1065 {
1066 	return false;
1067 }
1068 #endif
1069 
1070 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1071 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1072 				     struct ceph_acl_sec_ctx *ctx);
1073 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1074 {
1075 	security_inode_invalidate_secctx(inode);
1076 }
1077 #else
1078 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1079 					    struct ceph_acl_sec_ctx *ctx)
1080 {
1081 	return 0;
1082 }
1083 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1084 {
1085 }
1086 #endif
1087 
1088 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1089 
1090 /* acl.c */
1091 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1092 
1093 struct posix_acl *ceph_get_acl(struct inode *, int);
1094 int ceph_set_acl(struct user_namespace *mnt_userns,
1095 		 struct inode *inode, struct posix_acl *acl, int type);
1096 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1097 		       struct ceph_acl_sec_ctx *as_ctx);
1098 void ceph_init_inode_acls(struct inode *inode,
1099 			  struct ceph_acl_sec_ctx *as_ctx);
1100 
1101 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1102 {
1103        forget_all_cached_acls(inode);
1104 }
1105 
1106 #else
1107 
1108 #define ceph_get_acl NULL
1109 #define ceph_set_acl NULL
1110 
1111 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1112 				     struct ceph_acl_sec_ctx *as_ctx)
1113 {
1114 	return 0;
1115 }
1116 static inline void ceph_init_inode_acls(struct inode *inode,
1117 					struct ceph_acl_sec_ctx *as_ctx)
1118 {
1119 }
1120 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1121 {
1122 	return 0;
1123 }
1124 
1125 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1126 {
1127 }
1128 
1129 #endif
1130 
1131 /* caps.c */
1132 extern const char *ceph_cap_string(int c);
1133 extern void ceph_handle_caps(struct ceph_mds_session *session,
1134 			     struct ceph_msg *msg);
1135 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1136 				     struct ceph_cap_reservation *ctx);
1137 extern void ceph_add_cap(struct inode *inode,
1138 			 struct ceph_mds_session *session, u64 cap_id,
1139 			 unsigned issued, unsigned wanted,
1140 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1141 			 struct ceph_cap **new_cap);
1142 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1143 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1144 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1145 			 struct ceph_cap *cap);
1146 extern int ceph_is_any_caps(struct inode *inode);
1147 
1148 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1149 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1150 		      int datasync);
1151 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1152 					  struct ceph_mds_session *session);
1153 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1154 				    struct ceph_mds_session *session);
1155 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1156 				   struct ceph_inode_info *ci);
1157 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1158 					     int mds);
1159 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1160 				bool snap_rwsem_locked);
1161 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1162 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1163 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1164 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1165 					    int had);
1166 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1167 				       struct ceph_snap_context *snapc);
1168 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1169 			     struct ceph_mds_session **psession);
1170 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1171 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1172 			    struct ceph_mds_session *session);
1173 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1174 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1175 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1176 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1177 				     int mds, int drop, int unless, int force);
1178 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1179 				      struct inode *dir,
1180 				      int mds, int drop, int unless);
1181 
1182 extern int ceph_get_caps(struct file *filp, int need, int want,
1183 			 loff_t endoff, int *got);
1184 extern int ceph_try_get_caps(struct inode *inode,
1185 			     int need, int want, bool nonblock, int *got);
1186 
1187 /* for counting open files by mode */
1188 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1189 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1190 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1191 			       struct ceph_mds_client *mdsc, int fmode);
1192 
1193 /* addr.c */
1194 extern const struct address_space_operations ceph_aops;
1195 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1196 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1197 extern int ceph_pool_perm_check(struct inode *inode, int need);
1198 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1199 
1200 /* file.c */
1201 extern const struct file_operations ceph_file_fops;
1202 
1203 extern int ceph_renew_caps(struct inode *inode, int fmode);
1204 extern int ceph_open(struct inode *inode, struct file *file);
1205 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1206 			    struct file *file, unsigned flags, umode_t mode);
1207 extern int ceph_release(struct inode *inode, struct file *filp);
1208 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1209 				  char *data, size_t len);
1210 
1211 /* dir.c */
1212 extern const struct file_operations ceph_dir_fops;
1213 extern const struct file_operations ceph_snapdir_fops;
1214 extern const struct inode_operations ceph_dir_iops;
1215 extern const struct inode_operations ceph_snapdir_iops;
1216 extern const struct dentry_operations ceph_dentry_ops;
1217 
1218 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1219 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1220 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1221 			       struct dentry *dentry, int err);
1222 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1223 					 struct dentry *dentry, int err);
1224 
1225 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1226 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1227 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1228 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1229 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1230 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1231 
1232 /* ioctl.c */
1233 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1234 
1235 /* export.c */
1236 extern const struct export_operations ceph_export_ops;
1237 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1238 
1239 /* locks.c */
1240 extern __init void ceph_flock_init(void);
1241 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1242 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1243 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1244 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1245 				       struct ceph_filelock *flocks,
1246 				       int num_fcntl_locks,
1247 				       int num_flock_locks);
1248 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1249 				  struct ceph_pagelist *pagelist,
1250 				  int num_fcntl_locks, int num_flock_locks);
1251 
1252 /* debugfs.c */
1253 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1254 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1255 
1256 /* quota.c */
1257 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1258 {
1259 	return ci->i_max_files || ci->i_max_bytes;
1260 }
1261 
1262 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1263 
1264 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1265 				       u64 max_bytes, u64 max_files)
1266 {
1267 	bool had_quota, has_quota;
1268 	had_quota = __ceph_has_any_quota(ci);
1269 	ci->i_max_bytes = max_bytes;
1270 	ci->i_max_files = max_files;
1271 	has_quota = __ceph_has_any_quota(ci);
1272 
1273 	if (had_quota != has_quota)
1274 		ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1275 }
1276 
1277 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1278 			      struct ceph_mds_session *session,
1279 			      struct ceph_msg *msg);
1280 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1281 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1282 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1283 					     loff_t newlen);
1284 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1285 						loff_t newlen);
1286 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1287 				     struct kstatfs *buf);
1288 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1289 
1290 #endif /* _FS_CEPH_SUPER_H */
1291