xref: /linux/fs/ceph/super.h (revision 5a48b7433a5aee719ab242d2feadaf4c9e065989)
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 /* large granularity for statfs utilization stats to facilitate
28  * large volume sizes on 32-bit machines. */
29 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
30 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
31 #define CEPH_4K_BLOCK_SHIFT 12  /* 4 KB */
32 
33 #define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
34 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
35 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
36 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
37 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
38 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
39 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
40 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
41 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
42 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
43 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
44 #define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
45 #define CEPH_MOUNT_OPT_NOPAGECACHE     (1<<16) /* bypass pagecache altogether */
46 
47 #define CEPH_MOUNT_OPT_DEFAULT			\
48 	(CEPH_MOUNT_OPT_DCACHE |		\
49 	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
50 	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
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 	bool new_dev_syntax;
91 
92 	/*
93 	 * everything above this point can be memcmp'd; everything below
94 	 * is handled in compare_mount_options()
95 	 */
96 
97 	char *snapdir_name;   /* default ".snap" */
98 	char *mds_namespace;  /* default NULL */
99 	char *server_path;    /* default NULL (means "/") */
100 	char *fscache_uniq;   /* default NULL */
101 	char *mon_addr;
102 };
103 
104 struct ceph_fs_client {
105 	struct super_block *sb;
106 
107 	struct list_head metric_wakeup;
108 
109 	struct ceph_mount_options *mount_options;
110 	struct ceph_client *client;
111 
112 	int mount_state;
113 
114 	bool blocklisted;
115 
116 	bool have_copy_from2;
117 
118 	u32 filp_gen;
119 	loff_t max_file_size;
120 
121 	struct ceph_mds_client *mdsc;
122 
123 	atomic_long_t writeback_count;
124 	bool write_congested;
125 
126 	struct workqueue_struct *inode_wq;
127 	struct workqueue_struct *cap_wq;
128 
129 #ifdef CONFIG_DEBUG_FS
130 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
131 	struct dentry *debugfs_congestion_kb;
132 	struct dentry *debugfs_bdi;
133 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
134 	struct dentry *debugfs_status;
135 	struct dentry *debugfs_mds_sessions;
136 	struct dentry *debugfs_metrics_dir;
137 #endif
138 
139 #ifdef CONFIG_CEPH_FSCACHE
140 	struct fscache_volume *fscache;
141 #endif
142 };
143 
144 
145 /*
146  * File i/o capability.  This tracks shared state with the metadata
147  * server that allows us to cache or writeback attributes or to read
148  * and write data.  For any given inode, we should have one or more
149  * capabilities, one issued by each metadata server, and our
150  * cumulative access is the OR of all issued capabilities.
151  *
152  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
153  * session capability lists.
154  */
155 struct ceph_cap {
156 	struct ceph_inode_info *ci;
157 	struct rb_node ci_node;          /* per-ci cap tree */
158 	struct ceph_mds_session *session;
159 	struct list_head session_caps;   /* per-session caplist */
160 	u64 cap_id;       /* unique cap id (mds provided) */
161 	union {
162 		/* in-use caps */
163 		struct {
164 			int issued;       /* latest, from the mds */
165 			int implemented;  /* implemented superset of
166 					     issued (for revocation) */
167 			int mds;	  /* mds index for this cap */
168 			int mds_wanted;   /* caps wanted from this mds */
169 		};
170 		/* caps to release */
171 		struct {
172 			u64 cap_ino;
173 			int queue_release;
174 		};
175 	};
176 	u32 seq, issue_seq, mseq;
177 	u32 cap_gen;      /* active/stale cycle */
178 	unsigned long last_used;
179 	struct list_head caps_item;
180 };
181 
182 #define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
183 #define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
184 #define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
185 
186 struct ceph_cap_flush {
187 	u64 tid;
188 	int caps;
189 	bool wake; /* wake up flush waiters when finish ? */
190 	bool is_capsnap; /* true means capsnap */
191 	struct list_head g_list; // global
192 	struct list_head i_list; // per inode
193 };
194 
195 /*
196  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
197  * we first complete any in-process sync writes and writeback any dirty
198  * data before flushing the snapped state (tracked here) back to the MDS.
199  */
200 struct ceph_cap_snap {
201 	refcount_t nref;
202 	struct list_head ci_item;
203 
204 	struct ceph_cap_flush cap_flush;
205 
206 	u64 follows;
207 	int issued, dirty;
208 	struct ceph_snap_context *context;
209 
210 	umode_t mode;
211 	kuid_t uid;
212 	kgid_t gid;
213 
214 	struct ceph_buffer *xattr_blob;
215 	u64 xattr_version;
216 
217 	u64 size;
218 	u64 change_attr;
219 	struct timespec64 mtime, atime, ctime, btime;
220 	u64 time_warp_seq;
221 	u64 truncate_size;
222 	u32 truncate_seq;
223 	int writing;   /* a sync write is still in progress */
224 	int dirty_pages;     /* dirty pages awaiting writeback */
225 	bool inline_data;
226 	bool need_flush;
227 };
228 
229 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
230 {
231 	if (refcount_dec_and_test(&capsnap->nref)) {
232 		if (capsnap->xattr_blob)
233 			ceph_buffer_put(capsnap->xattr_blob);
234 		kmem_cache_free(ceph_cap_snap_cachep, capsnap);
235 	}
236 }
237 
238 /*
239  * The frag tree describes how a directory is fragmented, potentially across
240  * multiple metadata servers.  It is also used to indicate points where
241  * metadata authority is delegated, and whether/where metadata is replicated.
242  *
243  * A _leaf_ frag will be present in the i_fragtree IFF there is
244  * delegation info.  That is, if mds >= 0 || ndist > 0.
245  */
246 #define CEPH_MAX_DIRFRAG_REP 4
247 
248 struct ceph_inode_frag {
249 	struct rb_node node;
250 
251 	/* fragtree state */
252 	u32 frag;
253 	int split_by;         /* i.e. 2^(split_by) children */
254 
255 	/* delegation and replication info */
256 	int mds;              /* -1 if same authority as parent */
257 	int ndist;            /* >0 if replicated */
258 	int dist[CEPH_MAX_DIRFRAG_REP];
259 };
260 
261 /*
262  * We cache inode xattrs as an encoded blob until they are first used,
263  * at which point we parse them into an rbtree.
264  */
265 struct ceph_inode_xattr {
266 	struct rb_node node;
267 
268 	const char *name;
269 	int name_len;
270 	const char *val;
271 	int val_len;
272 	int dirty;
273 
274 	int should_free_name;
275 	int should_free_val;
276 };
277 
278 /*
279  * Ceph dentry state
280  */
281 struct ceph_dentry_info {
282 	struct dentry *dentry;
283 	struct ceph_mds_session *lease_session;
284 	struct list_head lease_list;
285 	unsigned flags;
286 	int lease_shared_gen;
287 	u32 lease_gen;
288 	u32 lease_seq;
289 	unsigned long lease_renew_after, lease_renew_from;
290 	unsigned long time;
291 	u64 offset;
292 };
293 
294 #define CEPH_DENTRY_REFERENCED		1
295 #define CEPH_DENTRY_LEASE_LIST		2
296 #define CEPH_DENTRY_SHRINK_LIST		4
297 #define CEPH_DENTRY_PRIMARY_LINK	8
298 
299 struct ceph_inode_xattrs_info {
300 	/*
301 	 * (still encoded) xattr blob. we avoid the overhead of parsing
302 	 * this until someone actually calls getxattr, etc.
303 	 *
304 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
305 	 * NULL means we don't know.
306 	*/
307 	struct ceph_buffer *blob, *prealloc_blob;
308 
309 	struct rb_root index;
310 	bool dirty;
311 	int count;
312 	int names_size;
313 	int vals_size;
314 	u64 version, index_version;
315 };
316 
317 /*
318  * Ceph inode.
319  */
320 struct ceph_inode_info {
321 	struct ceph_vino i_vino;   /* ceph ino + snap */
322 
323 	spinlock_t i_ceph_lock;
324 
325 	u64 i_version;
326 	u64 i_inline_version;
327 	u32 i_time_warp_seq;
328 
329 	unsigned long i_ceph_flags;
330 	atomic64_t i_release_count;
331 	atomic64_t i_ordered_count;
332 	atomic64_t i_complete_seq[2];
333 
334 	struct ceph_dir_layout i_dir_layout;
335 	struct ceph_file_layout i_layout;
336 	struct ceph_file_layout i_cached_layout;	// for async creates
337 	char *i_symlink;
338 
339 	/* for dirs */
340 	struct timespec64 i_rctime;
341 	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
342 	u64 i_files, i_subdirs;
343 
344 	/* quotas */
345 	u64 i_max_bytes, i_max_files;
346 
347 	s32 i_dir_pin;
348 
349 	struct rb_root i_fragtree;
350 	int i_fragtree_nsplits;
351 	struct mutex i_fragtree_mutex;
352 
353 	struct ceph_inode_xattrs_info i_xattrs;
354 
355 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
356 	 * s_mutex. */
357 	struct rb_root i_caps;           /* cap list */
358 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
359 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
360 
361 	/*
362 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
363 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
364 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
365 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
366 	 * be tested under the i_ceph_lock. Changing anything requires both.
367 	 */
368 	struct list_head i_dirty_item;
369 
370 	/*
371 	 * Link to session's s_cap_flushing list. Protected in a similar
372 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
373 	 * s_cap_flushing list can be walked while holding either the s_mutex
374 	 * or msdc->cap_dirty_lock. List presence can also be checked while
375 	 * holding the i_ceph_lock for this inode.
376 	 */
377 	struct list_head i_flushing_item;
378 
379 	/* we need to track cap writeback on a per-cap-bit basis, to allow
380 	 * overlapping, pipelined cap flushes to the mds.  we can probably
381 	 * reduce the tid to 8 bits if we're concerned about inode size. */
382 	struct ceph_cap_flush *i_prealloc_cap_flush;
383 	struct list_head i_cap_flush_list;
384 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
385 	unsigned long i_hold_caps_max; /* jiffies */
386 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
387 	struct ceph_cap_reservation i_cap_migration_resv;
388 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
389 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
390 						    dirty|flushing caps */
391 	unsigned i_snap_caps;           /* cap bits for snapped files */
392 
393 	unsigned long i_last_rd;
394 	unsigned long i_last_wr;
395 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
396 
397 	struct mutex i_truncate_mutex;
398 	u32 i_truncate_seq;        /* last truncate to smaller size */
399 	u64 i_truncate_size;       /*  and the size we last truncated down to */
400 	int i_truncate_pending;    /*  still need to call vmtruncate */
401 
402 	u64 i_max_size;            /* max file size authorized by mds */
403 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
404 	u64 i_wanted_max_size;     /* offset we'd like to write too */
405 	u64 i_requested_max_size;  /* max_size we've requested */
406 
407 	/* held references to caps */
408 	int i_pin_ref;
409 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
410 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
411 	atomic_t i_filelock_ref;
412 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
413 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
414 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
415 
416 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
417 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
418 	spinlock_t i_unsafe_lock;
419 
420 	union {
421 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
422 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
423 	};
424 	struct list_head i_snap_realm_item;
425 	struct list_head i_snap_flush_item;
426 	struct timespec64 i_btime;
427 	struct timespec64 i_snap_btime;
428 
429 	struct work_struct i_work;
430 	unsigned long  i_work_mask;
431 
432 #ifdef CONFIG_CEPH_FSCACHE
433 	struct fscache_cookie *fscache;
434 #endif
435 	struct inode vfs_inode; /* at end */
436 };
437 
438 static inline struct ceph_inode_info *
439 ceph_inode(const struct inode *inode)
440 {
441 	return container_of(inode, struct ceph_inode_info, vfs_inode);
442 }
443 
444 static inline struct ceph_fs_client *
445 ceph_inode_to_client(const struct inode *inode)
446 {
447 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
448 }
449 
450 static inline struct ceph_fs_client *
451 ceph_sb_to_client(const struct super_block *sb)
452 {
453 	return (struct ceph_fs_client *)sb->s_fs_info;
454 }
455 
456 static inline struct ceph_mds_client *
457 ceph_sb_to_mdsc(const struct super_block *sb)
458 {
459 	return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
460 }
461 
462 static inline struct ceph_vino
463 ceph_vino(const struct inode *inode)
464 {
465 	return ceph_inode(inode)->i_vino;
466 }
467 
468 static inline u32 ceph_ino_to_ino32(u64 vino)
469 {
470 	u32 ino = vino & 0xffffffff;
471 	ino ^= vino >> 32;
472 	if (!ino)
473 		ino = 2;
474 	return ino;
475 }
476 
477 /*
478  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
479  * some arches. We generally do not use this value inside the ceph driver, but
480  * we do want to set it to something, so that generic vfs code has an
481  * appropriate value for tracepoints and the like.
482  */
483 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
484 {
485 	if (sizeof(ino_t) == sizeof(u32))
486 		return ceph_ino_to_ino32(vino.ino);
487 	return (ino_t)vino.ino;
488 }
489 
490 /* for printf-style formatting */
491 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
492 
493 static inline u64 ceph_ino(struct inode *inode)
494 {
495 	return ceph_inode(inode)->i_vino.ino;
496 }
497 
498 static inline u64 ceph_snap(struct inode *inode)
499 {
500 	return ceph_inode(inode)->i_vino.snap;
501 }
502 
503 /**
504  * ceph_present_ino - format an inode number for presentation to userland
505  * @sb: superblock where the inode lives
506  * @ino: inode number to (possibly) convert
507  *
508  * If the user mounted with the ino32 option, then the 64-bit value needs
509  * to be converted to something that can fit inside 32 bits. Note that
510  * internal kernel code never uses this value, so this is entirely for
511  * userland consumption.
512  */
513 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
514 {
515 	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
516 		return ceph_ino_to_ino32(ino);
517 	return ino;
518 }
519 
520 static inline u64 ceph_present_inode(struct inode *inode)
521 {
522 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
523 }
524 
525 static inline int ceph_ino_compare(struct inode *inode, void *data)
526 {
527 	struct ceph_vino *pvino = (struct ceph_vino *)data;
528 	struct ceph_inode_info *ci = ceph_inode(inode);
529 	return ci->i_vino.ino == pvino->ino &&
530 		ci->i_vino.snap == pvino->snap;
531 }
532 
533 /*
534  * The MDS reserves a set of inodes for its own usage. These should never
535  * be accessible by clients, and so the MDS has no reason to ever hand these
536  * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
537  *
538  * These come from src/mds/mdstypes.h in the ceph sources.
539  */
540 #define CEPH_MAX_MDS			0x100
541 #define CEPH_NUM_STRAY			10
542 #define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
543 #define CEPH_MDS_INO_LOG_OFFSET		(2 * CEPH_MAX_MDS)
544 #define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
545 
546 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
547 {
548 	if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
549 	    vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
550 		return false;
551 
552 	/* Don't warn on mdsdirs */
553 	WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
554 			"Attempt to access reserved inode number 0x%llx",
555 			vino.ino);
556 	return true;
557 }
558 
559 static inline struct inode *ceph_find_inode(struct super_block *sb,
560 					    struct ceph_vino vino)
561 {
562 	if (ceph_vino_is_reserved(vino))
563 		return NULL;
564 
565 	/*
566 	 * NB: The hashval will be run through the fs/inode.c hash function
567 	 * anyway, so there is no need to squash the inode number down to
568 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
569 	 */
570 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
571 }
572 
573 
574 /*
575  * Ceph inode.
576  */
577 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
578 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
579 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
580 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
581 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
582 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
583 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
584 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
585 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
586 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
587 #define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
588 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
589 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
590 #define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
591 
592 /*
593  * Masks of ceph inode work.
594  */
595 #define CEPH_I_WORK_WRITEBACK		0
596 #define CEPH_I_WORK_INVALIDATE_PAGES	1
597 #define CEPH_I_WORK_VMTRUNCATE		2
598 #define CEPH_I_WORK_CHECK_CAPS		3
599 #define CEPH_I_WORK_FLUSH_SNAPS		4
600 
601 /*
602  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
603  * and then clear it when they start succeeding. Note that we do a lockless
604  * check first, and only take the lock if it looks like it needs to be changed.
605  * The write submission code just takes this as a hint, so we're not too
606  * worried if a few slip through in either direction.
607  */
608 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
609 {
610 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
611 		spin_lock(&ci->i_ceph_lock);
612 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
613 		spin_unlock(&ci->i_ceph_lock);
614 	}
615 }
616 
617 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
618 {
619 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
620 		spin_lock(&ci->i_ceph_lock);
621 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
622 		spin_unlock(&ci->i_ceph_lock);
623 	}
624 }
625 
626 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
627 					   long long release_count,
628 					   long long ordered_count)
629 {
630 	/*
631 	 * Makes sure operations that setup readdir cache (update page
632 	 * cache and i_size) are strongly ordered w.r.t. the following
633 	 * atomic64_set() operations.
634 	 */
635 	smp_mb();
636 	atomic64_set(&ci->i_complete_seq[0], release_count);
637 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
638 }
639 
640 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
641 {
642 	atomic64_inc(&ci->i_release_count);
643 }
644 
645 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
646 {
647 	atomic64_inc(&ci->i_ordered_count);
648 }
649 
650 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
651 {
652 	return atomic64_read(&ci->i_complete_seq[0]) ==
653 		atomic64_read(&ci->i_release_count);
654 }
655 
656 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
657 {
658 	return  atomic64_read(&ci->i_complete_seq[0]) ==
659 		atomic64_read(&ci->i_release_count) &&
660 		atomic64_read(&ci->i_complete_seq[1]) ==
661 		atomic64_read(&ci->i_ordered_count);
662 }
663 
664 static inline void ceph_dir_clear_complete(struct inode *inode)
665 {
666 	__ceph_dir_clear_complete(ceph_inode(inode));
667 }
668 
669 static inline void ceph_dir_clear_ordered(struct inode *inode)
670 {
671 	__ceph_dir_clear_ordered(ceph_inode(inode));
672 }
673 
674 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
675 {
676 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
677 	smp_rmb();
678 	return ret;
679 }
680 
681 /* find a specific frag @f */
682 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
683 						u32 f);
684 
685 /*
686  * choose fragment for value @v.  copy frag content to pfrag, if leaf
687  * exists
688  */
689 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
690 			    struct ceph_inode_frag *pfrag,
691 			    int *found);
692 
693 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
694 {
695 	return (struct ceph_dentry_info *)dentry->d_fsdata;
696 }
697 
698 /*
699  * caps helpers
700  */
701 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
702 {
703 	return !RB_EMPTY_ROOT(&ci->i_caps);
704 }
705 
706 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
707 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
708 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
709 					  int t);
710 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
711 				    struct ceph_cap *cap);
712 
713 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
714 {
715 	int issued;
716 	spin_lock(&ci->i_ceph_lock);
717 	issued = __ceph_caps_issued(ci, NULL);
718 	spin_unlock(&ci->i_ceph_lock);
719 	return issued;
720 }
721 
722 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
723 					       int mask, int touch)
724 {
725 	int r;
726 	spin_lock(&ci->i_ceph_lock);
727 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
728 	spin_unlock(&ci->i_ceph_lock);
729 	return r;
730 }
731 
732 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
733 {
734 	return ci->i_dirty_caps | ci->i_flushing_caps;
735 }
736 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
737 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
738 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
739 				  struct ceph_cap_flush **pcf);
740 
741 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
742 				      struct ceph_cap *ocap, int mask);
743 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
744 extern int __ceph_caps_used(struct ceph_inode_info *ci);
745 
746 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
747 {
748 	return ci->i_nr_by_mode[0];
749 }
750 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
751 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
752 
753 /* what the mds thinks we want */
754 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
755 
756 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
757 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
758 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
759 				     struct ceph_mount_options *fsopt);
760 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
761 			     struct ceph_cap_reservation *ctx, int need);
762 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
763 			       struct ceph_cap_reservation *ctx);
764 extern void ceph_reservation_status(struct ceph_fs_client *client,
765 				    int *total, int *avail, int *used,
766 				    int *reserved, int *min);
767 
768 
769 
770 /*
771  * we keep buffered readdir results attached to file->private_data
772  */
773 #define CEPH_F_SYNC     1
774 #define CEPH_F_ATEND    2
775 
776 struct ceph_file_info {
777 	short fmode;     /* initialized on open */
778 	short flags;     /* CEPH_F_* */
779 
780 	spinlock_t rw_contexts_lock;
781 	struct list_head rw_contexts;
782 
783 	u32 filp_gen;
784 	atomic_t num_locks;
785 };
786 
787 struct ceph_dir_file_info {
788 	struct ceph_file_info file_info;
789 
790 	/* readdir: position within the dir */
791 	u32 frag;
792 	struct ceph_mds_request *last_readdir;
793 
794 	/* readdir: position within a frag */
795 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
796 	char *last_name;       /* last entry in previous chunk */
797 	long long dir_release_count;
798 	long long dir_ordered_count;
799 	int readdir_cache_idx;
800 
801 	/* used for -o dirstat read() on directory thing */
802 	char *dir_info;
803 	int dir_info_len;
804 };
805 
806 struct ceph_rw_context {
807 	struct list_head list;
808 	struct task_struct *thread;
809 	int caps;
810 };
811 
812 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
813 	struct ceph_rw_context _name = {	\
814 		.thread = current,		\
815 		.caps = _caps,			\
816 	}
817 
818 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
819 				       struct ceph_rw_context *ctx)
820 {
821 	spin_lock(&cf->rw_contexts_lock);
822 	list_add(&ctx->list, &cf->rw_contexts);
823 	spin_unlock(&cf->rw_contexts_lock);
824 }
825 
826 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
827 				       struct ceph_rw_context *ctx)
828 {
829 	spin_lock(&cf->rw_contexts_lock);
830 	list_del(&ctx->list);
831 	spin_unlock(&cf->rw_contexts_lock);
832 }
833 
834 static inline struct ceph_rw_context*
835 ceph_find_rw_context(struct ceph_file_info *cf)
836 {
837 	struct ceph_rw_context *ctx, *found = NULL;
838 	spin_lock(&cf->rw_contexts_lock);
839 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
840 		if (ctx->thread == current) {
841 			found = ctx;
842 			break;
843 		}
844 	}
845 	spin_unlock(&cf->rw_contexts_lock);
846 	return found;
847 }
848 
849 struct ceph_readdir_cache_control {
850 	struct page  *page;
851 	struct dentry **dentries;
852 	int index;
853 };
854 
855 /*
856  * A "snap realm" describes a subset of the file hierarchy sharing
857  * the same set of snapshots that apply to it.  The realms themselves
858  * are organized into a hierarchy, such that children inherit (some of)
859  * the snapshots of their parents.
860  *
861  * All inodes within the realm that have capabilities are linked into a
862  * per-realm list.
863  */
864 struct ceph_snap_realm {
865 	u64 ino;
866 	struct inode *inode;
867 	atomic_t nref;
868 	struct rb_node node;
869 
870 	u64 created, seq;
871 	u64 parent_ino;
872 	u64 parent_since;   /* snapid when our current parent became so */
873 
874 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
875 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
876 	u64 *snaps;                   /* snaps specific to this realm */
877 	u32 num_snaps;
878 
879 	struct ceph_snap_realm *parent;
880 	struct list_head children;       /* list of child realms */
881 	struct list_head child_item;
882 
883 	struct list_head empty_item;     /* if i have ref==0 */
884 
885 	struct list_head dirty_item;     /* if realm needs new context */
886 
887 	struct list_head rebuild_item;   /* rebuild snap realms _downward_ in hierarchy */
888 
889 	/* the current set of snaps for this realm */
890 	struct ceph_snap_context *cached_context;
891 
892 	struct list_head inodes_with_caps;
893 	spinlock_t inodes_with_caps_lock;
894 };
895 
896 static inline int default_congestion_kb(void)
897 {
898 	int congestion_kb;
899 
900 	/*
901 	 * Copied from NFS
902 	 *
903 	 * congestion size, scale with available memory.
904 	 *
905 	 *  64MB:    8192k
906 	 * 128MB:   11585k
907 	 * 256MB:   16384k
908 	 * 512MB:   23170k
909 	 *   1GB:   32768k
910 	 *   2GB:   46340k
911 	 *   4GB:   65536k
912 	 *   8GB:   92681k
913 	 *  16GB:  131072k
914 	 *
915 	 * This allows larger machines to have larger/more transfers.
916 	 * Limit the default to 256M
917 	 */
918 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
919 	if (congestion_kb > 256*1024)
920 		congestion_kb = 256*1024;
921 
922 	return congestion_kb;
923 }
924 
925 
926 /* super.c */
927 extern int ceph_force_reconnect(struct super_block *sb);
928 /* snap.c */
929 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
930 					       u64 ino);
931 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
932 				struct ceph_snap_realm *realm);
933 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
934 				struct ceph_snap_realm *realm);
935 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
936 				  void *p, void *e, bool deletion,
937 				  struct ceph_snap_realm **realm_ret);
938 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
939 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
940 			     struct ceph_mds_session *session,
941 			     struct ceph_msg *msg);
942 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
943 				  struct ceph_cap_snap *capsnap);
944 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
945 
946 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
947 						   u64 snap);
948 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
949 				struct ceph_snapid_map *sm);
950 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
951 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
952 void ceph_umount_begin(struct super_block *sb);
953 
954 
955 /*
956  * a cap_snap is "pending" if it is still awaiting an in-progress
957  * sync write (that may/may not still update size, mtime, etc.).
958  */
959 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
960 {
961 	return !list_empty(&ci->i_cap_snaps) &&
962 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
963 			       ci_item)->writing;
964 }
965 
966 /* inode.c */
967 struct ceph_mds_reply_info_in;
968 struct ceph_mds_reply_dirfrag;
969 
970 extern const struct inode_operations ceph_file_iops;
971 
972 extern struct inode *ceph_alloc_inode(struct super_block *sb);
973 extern void ceph_evict_inode(struct inode *inode);
974 extern void ceph_free_inode(struct inode *inode);
975 
976 extern struct inode *ceph_get_inode(struct super_block *sb,
977 				    struct ceph_vino vino);
978 extern struct inode *ceph_get_snapdir(struct inode *parent);
979 extern int ceph_fill_file_size(struct inode *inode, int issued,
980 			       u32 truncate_seq, u64 truncate_size, u64 size);
981 extern void ceph_fill_file_time(struct inode *inode, int issued,
982 				u64 time_warp_seq, struct timespec64 *ctime,
983 				struct timespec64 *mtime,
984 				struct timespec64 *atime);
985 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
986 		    struct ceph_mds_reply_info_in *iinfo,
987 		    struct ceph_mds_reply_dirfrag *dirinfo,
988 		    struct ceph_mds_session *session, int cap_fmode,
989 		    struct ceph_cap_reservation *caps_reservation);
990 extern int ceph_fill_trace(struct super_block *sb,
991 			   struct ceph_mds_request *req);
992 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
993 				    struct ceph_mds_session *session);
994 
995 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
996 
997 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
998 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
999 
1000 void ceph_queue_inode_work(struct inode *inode, int work_bit);
1001 
1002 static inline void ceph_queue_vmtruncate(struct inode *inode)
1003 {
1004 	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1005 }
1006 
1007 static inline void ceph_queue_invalidate(struct inode *inode)
1008 {
1009 	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1010 }
1011 
1012 static inline void ceph_queue_writeback(struct inode *inode)
1013 {
1014 	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1015 }
1016 
1017 static inline void ceph_queue_check_caps(struct inode *inode)
1018 {
1019 	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1020 }
1021 
1022 static inline void ceph_queue_flush_snaps(struct inode *inode)
1023 {
1024 	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1025 }
1026 
1027 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1028 			     int mask, bool force);
1029 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1030 {
1031 	return __ceph_do_getattr(inode, NULL, mask, force);
1032 }
1033 extern int ceph_permission(struct user_namespace *mnt_userns,
1034 			   struct inode *inode, int mask);
1035 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1036 extern int ceph_setattr(struct user_namespace *mnt_userns,
1037 			struct dentry *dentry, struct iattr *attr);
1038 extern int ceph_getattr(struct user_namespace *mnt_userns,
1039 			const struct path *path, struct kstat *stat,
1040 			u32 request_mask, unsigned int flags);
1041 void ceph_inode_shutdown(struct inode *inode);
1042 
1043 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1044 {
1045 	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1046 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1047 	int state = READ_ONCE(fsc->mount_state);
1048 
1049 	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1050 }
1051 
1052 /* xattr.c */
1053 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1054 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1055 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1056 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1057 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1058 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1059 extern const struct xattr_handler *ceph_xattr_handlers[];
1060 
1061 struct ceph_acl_sec_ctx {
1062 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1063 	void *default_acl;
1064 	void *acl;
1065 #endif
1066 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1067 	void *sec_ctx;
1068 	u32 sec_ctxlen;
1069 #endif
1070 	struct ceph_pagelist *pagelist;
1071 };
1072 
1073 #ifdef CONFIG_SECURITY
1074 extern bool ceph_security_xattr_deadlock(struct inode *in);
1075 extern bool ceph_security_xattr_wanted(struct inode *in);
1076 #else
1077 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1078 {
1079 	return false;
1080 }
1081 static inline bool ceph_security_xattr_wanted(struct inode *in)
1082 {
1083 	return false;
1084 }
1085 #endif
1086 
1087 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1088 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1089 				     struct ceph_acl_sec_ctx *ctx);
1090 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1091 {
1092 	security_inode_invalidate_secctx(inode);
1093 }
1094 #else
1095 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1096 					    struct ceph_acl_sec_ctx *ctx)
1097 {
1098 	return 0;
1099 }
1100 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1101 {
1102 }
1103 #endif
1104 
1105 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1106 
1107 /* acl.c */
1108 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1109 
1110 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1111 int ceph_set_acl(struct user_namespace *mnt_userns,
1112 		 struct inode *inode, struct posix_acl *acl, int type);
1113 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1114 		       struct ceph_acl_sec_ctx *as_ctx);
1115 void ceph_init_inode_acls(struct inode *inode,
1116 			  struct ceph_acl_sec_ctx *as_ctx);
1117 
1118 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1119 {
1120        forget_all_cached_acls(inode);
1121 }
1122 
1123 #else
1124 
1125 #define ceph_get_acl NULL
1126 #define ceph_set_acl NULL
1127 
1128 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1129 				     struct ceph_acl_sec_ctx *as_ctx)
1130 {
1131 	return 0;
1132 }
1133 static inline void ceph_init_inode_acls(struct inode *inode,
1134 					struct ceph_acl_sec_ctx *as_ctx)
1135 {
1136 }
1137 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1138 {
1139 	return 0;
1140 }
1141 
1142 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1143 {
1144 }
1145 
1146 #endif
1147 
1148 /* caps.c */
1149 extern const char *ceph_cap_string(int c);
1150 extern void ceph_handle_caps(struct ceph_mds_session *session,
1151 			     struct ceph_msg *msg);
1152 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1153 				     struct ceph_cap_reservation *ctx);
1154 extern void ceph_add_cap(struct inode *inode,
1155 			 struct ceph_mds_session *session, u64 cap_id,
1156 			 unsigned issued, unsigned wanted,
1157 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1158 			 struct ceph_cap **new_cap);
1159 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1160 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1161 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1162 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1163 			 struct ceph_cap *cap);
1164 extern int ceph_is_any_caps(struct inode *inode);
1165 
1166 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1167 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1168 		      int datasync);
1169 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1170 					  struct ceph_mds_session *session);
1171 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1172 				    struct ceph_mds_session *session);
1173 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1174 				   struct ceph_inode_info *ci);
1175 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1176 					     int mds);
1177 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1178 				bool snap_rwsem_locked);
1179 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1180 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1181 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1182 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1183 					    int had);
1184 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1185 				       struct ceph_snap_context *snapc);
1186 extern void __ceph_remove_capsnap(struct inode *inode,
1187 				  struct ceph_cap_snap *capsnap,
1188 				  bool *wake_ci, bool *wake_mdsc);
1189 extern void ceph_remove_capsnap(struct inode *inode,
1190 				struct ceph_cap_snap *capsnap,
1191 				bool *wake_ci, bool *wake_mdsc);
1192 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1193 			     struct ceph_mds_session **psession);
1194 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1195 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1196 			    struct ceph_mds_session *session);
1197 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1198 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1199 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1200 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1201 				     int mds, int drop, int unless, int force);
1202 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1203 				      struct inode *dir,
1204 				      int mds, int drop, int unless);
1205 
1206 extern int ceph_get_caps(struct file *filp, int need, int want,
1207 			 loff_t endoff, int *got);
1208 extern int ceph_try_get_caps(struct inode *inode,
1209 			     int need, int want, bool nonblock, int *got);
1210 
1211 /* for counting open files by mode */
1212 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1213 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1214 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1215 			       struct ceph_mds_client *mdsc, int fmode);
1216 
1217 /* addr.c */
1218 extern const struct address_space_operations ceph_aops;
1219 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1220 extern int ceph_uninline_data(struct file *file);
1221 extern int ceph_pool_perm_check(struct inode *inode, int need);
1222 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1223 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1224 
1225 /* file.c */
1226 extern const struct file_operations ceph_file_fops;
1227 
1228 extern int ceph_renew_caps(struct inode *inode, int fmode);
1229 extern int ceph_open(struct inode *inode, struct file *file);
1230 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1231 			    struct file *file, unsigned flags, umode_t mode);
1232 extern int ceph_release(struct inode *inode, struct file *filp);
1233 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1234 				  char *data, size_t len);
1235 
1236 /* dir.c */
1237 extern const struct file_operations ceph_dir_fops;
1238 extern const struct file_operations ceph_snapdir_fops;
1239 extern const struct inode_operations ceph_dir_iops;
1240 extern const struct inode_operations ceph_snapdir_iops;
1241 extern const struct dentry_operations ceph_dentry_ops;
1242 
1243 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1244 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1245 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1246 			       struct dentry *dentry);
1247 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1248 					 struct dentry *dentry, int err);
1249 
1250 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1251 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1252 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1253 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1254 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1255 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1256 
1257 /* ioctl.c */
1258 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1259 
1260 /* export.c */
1261 extern const struct export_operations ceph_export_ops;
1262 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1263 
1264 /* locks.c */
1265 extern __init void ceph_flock_init(void);
1266 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1267 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1268 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1269 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1270 				       struct ceph_filelock *flocks,
1271 				       int num_fcntl_locks,
1272 				       int num_flock_locks);
1273 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1274 				  struct ceph_pagelist *pagelist,
1275 				  int num_fcntl_locks, int num_flock_locks);
1276 
1277 /* debugfs.c */
1278 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1279 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1280 
1281 /* quota.c */
1282 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1283 {
1284 	return ci->i_max_files || ci->i_max_bytes;
1285 }
1286 
1287 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1288 
1289 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1290 				       u64 max_bytes, u64 max_files)
1291 {
1292 	bool had_quota, has_quota;
1293 	had_quota = __ceph_has_any_quota(ci);
1294 	ci->i_max_bytes = max_bytes;
1295 	ci->i_max_files = max_files;
1296 	has_quota = __ceph_has_any_quota(ci);
1297 
1298 	if (had_quota != has_quota)
1299 		ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1300 }
1301 
1302 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1303 			      struct ceph_mds_session *session,
1304 			      struct ceph_msg *msg);
1305 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1306 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1307 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1308 					     loff_t newlen);
1309 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1310 						loff_t newlen);
1311 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1312 				     struct kstatfs *buf);
1313 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1314 
1315 #endif /* _FS_CEPH_SUPER_H */
1316