xref: /openbmc/linux/fs/ceph/super.h (revision 34fa67e7)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_SUPER_H
3 #define _FS_CEPH_SUPER_H
4 
5 #include <linux/ceph/ceph_debug.h>
6 
7 #include <asm/unaligned.h>
8 #include <linux/backing-dev.h>
9 #include <linux/completion.h>
10 #include <linux/exportfs.h>
11 #include <linux/fs.h>
12 #include <linux/mempool.h>
13 #include <linux/pagemap.h>
14 #include <linux/wait.h>
15 #include <linux/writeback.h>
16 #include <linux/slab.h>
17 #include <linux/posix_acl.h>
18 #include <linux/refcount.h>
19 #include <linux/security.h>
20 
21 #include <linux/ceph/libceph.h>
22 
23 #ifdef CONFIG_CEPH_FSCACHE
24 #include <linux/fscache.h>
25 #endif
26 
27 /* f_type in struct statfs */
28 #define CEPH_SUPER_MAGIC 0x00c36400
29 
30 /* large granularity for statfs utilization stats to facilitate
31  * large volume sizes on 32-bit machines. */
32 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
33 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
34 
35 #define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
36 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
37 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
38 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
39 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
40 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
41 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
42 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
43 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
44 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
45 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
46 #define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
47 
48 #define CEPH_MOUNT_OPT_DEFAULT			\
49 	(CEPH_MOUNT_OPT_DCACHE |		\
50 	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
51 	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
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_status;
132 	struct dentry *debugfs_mds_sessions;
133 	struct dentry *debugfs_metrics_dir;
134 #endif
135 
136 #ifdef CONFIG_CEPH_FSCACHE
137 	struct fscache_volume *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;
186 	bool wake; /* wake up flush waiters when finish ? */
187 	bool is_capsnap; /* true means capsnap */
188 	struct list_head g_list; // global
189 	struct list_head i_list; // per inode
190 };
191 
192 /*
193  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
194  * we first complete any in-process sync writes and writeback any dirty
195  * data before flushing the snapped state (tracked here) back to the MDS.
196  */
197 struct ceph_cap_snap {
198 	refcount_t nref;
199 	struct list_head ci_item;
200 
201 	struct ceph_cap_flush cap_flush;
202 
203 	u64 follows;
204 	int issued, dirty;
205 	struct ceph_snap_context *context;
206 
207 	umode_t mode;
208 	kuid_t uid;
209 	kgid_t gid;
210 
211 	struct ceph_buffer *xattr_blob;
212 	u64 xattr_version;
213 
214 	u64 size;
215 	u64 change_attr;
216 	struct timespec64 mtime, atime, ctime, btime;
217 	u64 time_warp_seq;
218 	u64 truncate_size;
219 	u32 truncate_seq;
220 	int writing;   /* a sync write is still in progress */
221 	int dirty_pages;     /* dirty pages awaiting writeback */
222 	bool inline_data;
223 	bool need_flush;
224 };
225 
226 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
227 {
228 	if (refcount_dec_and_test(&capsnap->nref)) {
229 		if (capsnap->xattr_blob)
230 			ceph_buffer_put(capsnap->xattr_blob);
231 		kfree(capsnap);
232 	}
233 }
234 
235 /*
236  * The frag tree describes how a directory is fragmented, potentially across
237  * multiple metadata servers.  It is also used to indicate points where
238  * metadata authority is delegated, and whether/where metadata is replicated.
239  *
240  * A _leaf_ frag will be present in the i_fragtree IFF there is
241  * delegation info.  That is, if mds >= 0 || ndist > 0.
242  */
243 #define CEPH_MAX_DIRFRAG_REP 4
244 
245 struct ceph_inode_frag {
246 	struct rb_node node;
247 
248 	/* fragtree state */
249 	u32 frag;
250 	int split_by;         /* i.e. 2^(split_by) children */
251 
252 	/* delegation and replication info */
253 	int mds;              /* -1 if same authority as parent */
254 	int ndist;            /* >0 if replicated */
255 	int dist[CEPH_MAX_DIRFRAG_REP];
256 };
257 
258 /*
259  * We cache inode xattrs as an encoded blob until they are first used,
260  * at which point we parse them into an rbtree.
261  */
262 struct ceph_inode_xattr {
263 	struct rb_node node;
264 
265 	const char *name;
266 	int name_len;
267 	const char *val;
268 	int val_len;
269 	int dirty;
270 
271 	int should_free_name;
272 	int should_free_val;
273 };
274 
275 /*
276  * Ceph dentry state
277  */
278 struct ceph_dentry_info {
279 	struct dentry *dentry;
280 	struct ceph_mds_session *lease_session;
281 	struct list_head lease_list;
282 	unsigned flags;
283 	int lease_shared_gen;
284 	u32 lease_gen;
285 	u32 lease_seq;
286 	unsigned long lease_renew_after, lease_renew_from;
287 	unsigned long time;
288 	u64 offset;
289 };
290 
291 #define CEPH_DENTRY_REFERENCED		1
292 #define CEPH_DENTRY_LEASE_LIST		2
293 #define CEPH_DENTRY_SHRINK_LIST		4
294 #define CEPH_DENTRY_PRIMARY_LINK	8
295 
296 struct ceph_inode_xattrs_info {
297 	/*
298 	 * (still encoded) xattr blob. we avoid the overhead of parsing
299 	 * this until someone actually calls getxattr, etc.
300 	 *
301 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
302 	 * NULL means we don't know.
303 	*/
304 	struct ceph_buffer *blob, *prealloc_blob;
305 
306 	struct rb_root index;
307 	bool dirty;
308 	int count;
309 	int names_size;
310 	int vals_size;
311 	u64 version, index_version;
312 };
313 
314 /*
315  * Ceph inode.
316  */
317 struct ceph_inode_info {
318 	struct ceph_vino i_vino;   /* ceph ino + snap */
319 
320 	spinlock_t i_ceph_lock;
321 
322 	u64 i_version;
323 	u64 i_inline_version;
324 	u32 i_time_warp_seq;
325 
326 	unsigned long i_ceph_flags;
327 	atomic64_t i_release_count;
328 	atomic64_t i_ordered_count;
329 	atomic64_t i_complete_seq[2];
330 
331 	struct ceph_dir_layout i_dir_layout;
332 	struct ceph_file_layout i_layout;
333 	struct ceph_file_layout i_cached_layout;	// for async creates
334 	char *i_symlink;
335 
336 	/* for dirs */
337 	struct timespec64 i_rctime;
338 	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
339 	u64 i_files, i_subdirs;
340 
341 	/* quotas */
342 	u64 i_max_bytes, i_max_files;
343 
344 	s32 i_dir_pin;
345 
346 	struct rb_root i_fragtree;
347 	int i_fragtree_nsplits;
348 	struct mutex i_fragtree_mutex;
349 
350 	struct ceph_inode_xattrs_info i_xattrs;
351 
352 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
353 	 * s_mutex. */
354 	struct rb_root i_caps;           /* cap list */
355 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
356 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
357 
358 	/*
359 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
360 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
361 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
362 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
363 	 * be tested under the i_ceph_lock. Changing anything requires both.
364 	 */
365 	struct list_head i_dirty_item;
366 
367 	/*
368 	 * Link to session's s_cap_flushing list. Protected in a similar
369 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
370 	 * s_cap_flushing list can be walked while holding either the s_mutex
371 	 * or msdc->cap_dirty_lock. List presence can also be checked while
372 	 * holding the i_ceph_lock for this inode.
373 	 */
374 	struct list_head i_flushing_item;
375 
376 	/* we need to track cap writeback on a per-cap-bit basis, to allow
377 	 * overlapping, pipelined cap flushes to the mds.  we can probably
378 	 * reduce the tid to 8 bits if we're concerned about inode size. */
379 	struct ceph_cap_flush *i_prealloc_cap_flush;
380 	struct list_head i_cap_flush_list;
381 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
382 	unsigned long i_hold_caps_max; /* jiffies */
383 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
384 	struct ceph_cap_reservation i_cap_migration_resv;
385 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
386 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
387 						    dirty|flushing caps */
388 	unsigned i_snap_caps;           /* cap bits for snapped files */
389 
390 	unsigned long i_last_rd;
391 	unsigned long i_last_wr;
392 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
393 
394 	struct mutex i_truncate_mutex;
395 	u32 i_truncate_seq;        /* last truncate to smaller size */
396 	u64 i_truncate_size;       /*  and the size we last truncated down to */
397 	int i_truncate_pending;    /*  still need to call vmtruncate */
398 
399 	u64 i_max_size;            /* max file size authorized by mds */
400 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
401 	u64 i_wanted_max_size;     /* offset we'd like to write too */
402 	u64 i_requested_max_size;  /* max_size we've requested */
403 
404 	/* held references to caps */
405 	int i_pin_ref;
406 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
407 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
408 	atomic_t i_filelock_ref;
409 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
410 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
411 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
412 
413 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
414 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
415 	spinlock_t i_unsafe_lock;
416 
417 	union {
418 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
419 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
420 	};
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 	struct inode vfs_inode; /* at end */
433 };
434 
435 static inline struct ceph_inode_info *
436 ceph_inode(const struct inode *inode)
437 {
438 	return container_of(inode, struct ceph_inode_info, vfs_inode);
439 }
440 
441 static inline struct ceph_fs_client *
442 ceph_inode_to_client(const struct inode *inode)
443 {
444 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
445 }
446 
447 static inline struct ceph_fs_client *
448 ceph_sb_to_client(const struct super_block *sb)
449 {
450 	return (struct ceph_fs_client *)sb->s_fs_info;
451 }
452 
453 static inline struct ceph_mds_client *
454 ceph_sb_to_mdsc(const struct super_block *sb)
455 {
456 	return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
457 }
458 
459 static inline struct ceph_vino
460 ceph_vino(const struct inode *inode)
461 {
462 	return ceph_inode(inode)->i_vino;
463 }
464 
465 static inline u32 ceph_ino_to_ino32(u64 vino)
466 {
467 	u32 ino = vino & 0xffffffff;
468 	ino ^= vino >> 32;
469 	if (!ino)
470 		ino = 2;
471 	return ino;
472 }
473 
474 /*
475  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
476  * some arches. We generally do not use this value inside the ceph driver, but
477  * we do want to set it to something, so that generic vfs code has an
478  * appropriate value for tracepoints and the like.
479  */
480 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
481 {
482 	if (sizeof(ino_t) == sizeof(u32))
483 		return ceph_ino_to_ino32(vino.ino);
484 	return (ino_t)vino.ino;
485 }
486 
487 /* for printf-style formatting */
488 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
489 
490 static inline u64 ceph_ino(struct inode *inode)
491 {
492 	return ceph_inode(inode)->i_vino.ino;
493 }
494 
495 static inline u64 ceph_snap(struct inode *inode)
496 {
497 	return ceph_inode(inode)->i_vino.snap;
498 }
499 
500 /**
501  * ceph_present_ino - format an inode number for presentation to userland
502  * @sb: superblock where the inode lives
503  * @ino: inode number to (possibly) convert
504  *
505  * If the user mounted with the ino32 option, then the 64-bit value needs
506  * to be converted to something that can fit inside 32 bits. Note that
507  * internal kernel code never uses this value, so this is entirely for
508  * userland consumption.
509  */
510 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
511 {
512 	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
513 		return ceph_ino_to_ino32(ino);
514 	return ino;
515 }
516 
517 static inline u64 ceph_present_inode(struct inode *inode)
518 {
519 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
520 }
521 
522 static inline int ceph_ino_compare(struct inode *inode, void *data)
523 {
524 	struct ceph_vino *pvino = (struct ceph_vino *)data;
525 	struct ceph_inode_info *ci = ceph_inode(inode);
526 	return ci->i_vino.ino == pvino->ino &&
527 		ci->i_vino.snap == pvino->snap;
528 }
529 
530 /*
531  * The MDS reserves a set of inodes for its own usage. These should never
532  * be accessible by clients, and so the MDS has no reason to ever hand these
533  * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
534  *
535  * These come from src/mds/mdstypes.h in the ceph sources.
536  */
537 #define CEPH_MAX_MDS		0x100
538 #define CEPH_NUM_STRAY		10
539 #define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
540 #define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
541 
542 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
543 {
544 	if (vino.ino < CEPH_INO_SYSTEM_BASE &&
545 	    vino.ino >= CEPH_MDS_INO_MDSDIR_OFFSET) {
546 		WARN_RATELIMIT(1, "Attempt to access reserved inode number 0x%llx", vino.ino);
547 		return true;
548 	}
549 	return false;
550 }
551 
552 static inline struct inode *ceph_find_inode(struct super_block *sb,
553 					    struct ceph_vino vino)
554 {
555 	if (ceph_vino_is_reserved(vino))
556 		return NULL;
557 
558 	/*
559 	 * NB: The hashval will be run through the fs/inode.c hash function
560 	 * anyway, so there is no need to squash the inode number down to
561 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
562 	 */
563 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
564 }
565 
566 
567 /*
568  * Ceph inode.
569  */
570 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
571 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
572 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
573 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
574 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
575 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
576 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
577 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
578 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
579 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
580 #define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
581 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
582 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
583 #define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
584 
585 /*
586  * Masks of ceph inode work.
587  */
588 #define CEPH_I_WORK_WRITEBACK		0
589 #define CEPH_I_WORK_INVALIDATE_PAGES	1
590 #define CEPH_I_WORK_VMTRUNCATE		2
591 #define CEPH_I_WORK_CHECK_CAPS		3
592 #define CEPH_I_WORK_FLUSH_SNAPS		4
593 
594 /*
595  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
596  * and then clear it when they start succeeding. Note that we do a lockless
597  * check first, and only take the lock if it looks like it needs to be changed.
598  * The write submission code just takes this as a hint, so we're not too
599  * worried if a few slip through in either direction.
600  */
601 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
602 {
603 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
604 		spin_lock(&ci->i_ceph_lock);
605 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
606 		spin_unlock(&ci->i_ceph_lock);
607 	}
608 }
609 
610 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
611 {
612 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
613 		spin_lock(&ci->i_ceph_lock);
614 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
615 		spin_unlock(&ci->i_ceph_lock);
616 	}
617 }
618 
619 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
620 					   long long release_count,
621 					   long long ordered_count)
622 {
623 	/*
624 	 * Makes sure operations that setup readdir cache (update page
625 	 * cache and i_size) are strongly ordered w.r.t. the following
626 	 * atomic64_set() operations.
627 	 */
628 	smp_mb();
629 	atomic64_set(&ci->i_complete_seq[0], release_count);
630 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
631 }
632 
633 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
634 {
635 	atomic64_inc(&ci->i_release_count);
636 }
637 
638 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
639 {
640 	atomic64_inc(&ci->i_ordered_count);
641 }
642 
643 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
644 {
645 	return atomic64_read(&ci->i_complete_seq[0]) ==
646 		atomic64_read(&ci->i_release_count);
647 }
648 
649 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
650 {
651 	return  atomic64_read(&ci->i_complete_seq[0]) ==
652 		atomic64_read(&ci->i_release_count) &&
653 		atomic64_read(&ci->i_complete_seq[1]) ==
654 		atomic64_read(&ci->i_ordered_count);
655 }
656 
657 static inline void ceph_dir_clear_complete(struct inode *inode)
658 {
659 	__ceph_dir_clear_complete(ceph_inode(inode));
660 }
661 
662 static inline void ceph_dir_clear_ordered(struct inode *inode)
663 {
664 	__ceph_dir_clear_ordered(ceph_inode(inode));
665 }
666 
667 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
668 {
669 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
670 	smp_rmb();
671 	return ret;
672 }
673 
674 /* find a specific frag @f */
675 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
676 						u32 f);
677 
678 /*
679  * choose fragment for value @v.  copy frag content to pfrag, if leaf
680  * exists
681  */
682 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
683 			    struct ceph_inode_frag *pfrag,
684 			    int *found);
685 
686 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
687 {
688 	return (struct ceph_dentry_info *)dentry->d_fsdata;
689 }
690 
691 /*
692  * caps helpers
693  */
694 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
695 {
696 	return !RB_EMPTY_ROOT(&ci->i_caps);
697 }
698 
699 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
700 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
701 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
702 					  int t);
703 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
704 				    struct ceph_cap *cap);
705 
706 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
707 {
708 	int issued;
709 	spin_lock(&ci->i_ceph_lock);
710 	issued = __ceph_caps_issued(ci, NULL);
711 	spin_unlock(&ci->i_ceph_lock);
712 	return issued;
713 }
714 
715 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
716 					       int mask, int touch)
717 {
718 	int r;
719 	spin_lock(&ci->i_ceph_lock);
720 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
721 	spin_unlock(&ci->i_ceph_lock);
722 	return r;
723 }
724 
725 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
726 {
727 	return ci->i_dirty_caps | ci->i_flushing_caps;
728 }
729 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
730 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
731 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
732 				  struct ceph_cap_flush **pcf);
733 
734 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
735 				      struct ceph_cap *ocap, int mask);
736 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
737 extern int __ceph_caps_used(struct ceph_inode_info *ci);
738 
739 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
740 {
741 	return ci->i_nr_by_mode[0];
742 }
743 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
744 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
745 
746 /* what the mds thinks we want */
747 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
748 
749 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
750 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
751 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
752 				     struct ceph_mount_options *fsopt);
753 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
754 			     struct ceph_cap_reservation *ctx, int need);
755 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
756 			       struct ceph_cap_reservation *ctx);
757 extern void ceph_reservation_status(struct ceph_fs_client *client,
758 				    int *total, int *avail, int *used,
759 				    int *reserved, int *min);
760 
761 
762 
763 /*
764  * we keep buffered readdir results attached to file->private_data
765  */
766 #define CEPH_F_SYNC     1
767 #define CEPH_F_ATEND    2
768 
769 struct ceph_file_info {
770 	short fmode;     /* initialized on open */
771 	short flags;     /* CEPH_F_* */
772 
773 	spinlock_t rw_contexts_lock;
774 	struct list_head rw_contexts;
775 
776 	u32 filp_gen;
777 	atomic_t num_locks;
778 };
779 
780 struct ceph_dir_file_info {
781 	struct ceph_file_info file_info;
782 
783 	/* readdir: position within the dir */
784 	u32 frag;
785 	struct ceph_mds_request *last_readdir;
786 
787 	/* readdir: position within a frag */
788 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
789 	char *last_name;       /* last entry in previous chunk */
790 	long long dir_release_count;
791 	long long dir_ordered_count;
792 	int readdir_cache_idx;
793 
794 	/* used for -o dirstat read() on directory thing */
795 	char *dir_info;
796 	int dir_info_len;
797 };
798 
799 struct ceph_rw_context {
800 	struct list_head list;
801 	struct task_struct *thread;
802 	int caps;
803 };
804 
805 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
806 	struct ceph_rw_context _name = {	\
807 		.thread = current,		\
808 		.caps = _caps,			\
809 	}
810 
811 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
812 				       struct ceph_rw_context *ctx)
813 {
814 	spin_lock(&cf->rw_contexts_lock);
815 	list_add(&ctx->list, &cf->rw_contexts);
816 	spin_unlock(&cf->rw_contexts_lock);
817 }
818 
819 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
820 				       struct ceph_rw_context *ctx)
821 {
822 	spin_lock(&cf->rw_contexts_lock);
823 	list_del(&ctx->list);
824 	spin_unlock(&cf->rw_contexts_lock);
825 }
826 
827 static inline struct ceph_rw_context*
828 ceph_find_rw_context(struct ceph_file_info *cf)
829 {
830 	struct ceph_rw_context *ctx, *found = NULL;
831 	spin_lock(&cf->rw_contexts_lock);
832 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
833 		if (ctx->thread == current) {
834 			found = ctx;
835 			break;
836 		}
837 	}
838 	spin_unlock(&cf->rw_contexts_lock);
839 	return found;
840 }
841 
842 struct ceph_readdir_cache_control {
843 	struct page  *page;
844 	struct dentry **dentries;
845 	int index;
846 };
847 
848 /*
849  * A "snap realm" describes a subset of the file hierarchy sharing
850  * the same set of snapshots that apply to it.  The realms themselves
851  * are organized into a hierarchy, such that children inherit (some of)
852  * the snapshots of their parents.
853  *
854  * All inodes within the realm that have capabilities are linked into a
855  * per-realm list.
856  */
857 struct ceph_snap_realm {
858 	u64 ino;
859 	struct inode *inode;
860 	atomic_t nref;
861 	struct rb_node node;
862 
863 	u64 created, seq;
864 	u64 parent_ino;
865 	u64 parent_since;   /* snapid when our current parent became so */
866 
867 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
868 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
869 	u64 *snaps;                   /* snaps specific to this realm */
870 	u32 num_snaps;
871 
872 	struct ceph_snap_realm *parent;
873 	struct list_head children;       /* list of child realms */
874 	struct list_head child_item;
875 
876 	struct list_head empty_item;     /* if i have ref==0 */
877 
878 	struct list_head dirty_item;     /* if realm needs new context */
879 
880 	/* the current set of snaps for this realm */
881 	struct ceph_snap_context *cached_context;
882 
883 	struct list_head inodes_with_caps;
884 	spinlock_t inodes_with_caps_lock;
885 };
886 
887 static inline int default_congestion_kb(void)
888 {
889 	int congestion_kb;
890 
891 	/*
892 	 * Copied from NFS
893 	 *
894 	 * congestion size, scale with available memory.
895 	 *
896 	 *  64MB:    8192k
897 	 * 128MB:   11585k
898 	 * 256MB:   16384k
899 	 * 512MB:   23170k
900 	 *   1GB:   32768k
901 	 *   2GB:   46340k
902 	 *   4GB:   65536k
903 	 *   8GB:   92681k
904 	 *  16GB:  131072k
905 	 *
906 	 * This allows larger machines to have larger/more transfers.
907 	 * Limit the default to 256M
908 	 */
909 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
910 	if (congestion_kb > 256*1024)
911 		congestion_kb = 256*1024;
912 
913 	return congestion_kb;
914 }
915 
916 
917 /* super.c */
918 extern int ceph_force_reconnect(struct super_block *sb);
919 /* snap.c */
920 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
921 					       u64 ino);
922 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
923 				struct ceph_snap_realm *realm);
924 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
925 				struct ceph_snap_realm *realm);
926 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
927 				  void *p, void *e, bool deletion,
928 				  struct ceph_snap_realm **realm_ret);
929 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
930 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
931 			     struct ceph_mds_session *session,
932 			     struct ceph_msg *msg);
933 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
934 				  struct ceph_cap_snap *capsnap);
935 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
936 
937 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
938 						   u64 snap);
939 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
940 				struct ceph_snapid_map *sm);
941 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
942 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
943 void ceph_umount_begin(struct super_block *sb);
944 
945 
946 /*
947  * a cap_snap is "pending" if it is still awaiting an in-progress
948  * sync write (that may/may not still update size, mtime, etc.).
949  */
950 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
951 {
952 	return !list_empty(&ci->i_cap_snaps) &&
953 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
954 			       ci_item)->writing;
955 }
956 
957 /* inode.c */
958 struct ceph_mds_reply_info_in;
959 struct ceph_mds_reply_dirfrag;
960 
961 extern const struct inode_operations ceph_file_iops;
962 
963 extern struct inode *ceph_alloc_inode(struct super_block *sb);
964 extern void ceph_evict_inode(struct inode *inode);
965 extern void ceph_free_inode(struct inode *inode);
966 
967 extern struct inode *ceph_get_inode(struct super_block *sb,
968 				    struct ceph_vino vino);
969 extern struct inode *ceph_get_snapdir(struct inode *parent);
970 extern int ceph_fill_file_size(struct inode *inode, int issued,
971 			       u32 truncate_seq, u64 truncate_size, u64 size);
972 extern void ceph_fill_file_time(struct inode *inode, int issued,
973 				u64 time_warp_seq, struct timespec64 *ctime,
974 				struct timespec64 *mtime,
975 				struct timespec64 *atime);
976 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
977 		    struct ceph_mds_reply_info_in *iinfo,
978 		    struct ceph_mds_reply_dirfrag *dirinfo,
979 		    struct ceph_mds_session *session, int cap_fmode,
980 		    struct ceph_cap_reservation *caps_reservation);
981 extern int ceph_fill_trace(struct super_block *sb,
982 			   struct ceph_mds_request *req);
983 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
984 				    struct ceph_mds_session *session);
985 
986 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
987 
988 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
989 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
990 
991 void ceph_queue_inode_work(struct inode *inode, int work_bit);
992 
993 static inline void ceph_queue_vmtruncate(struct inode *inode)
994 {
995 	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
996 }
997 
998 static inline void ceph_queue_invalidate(struct inode *inode)
999 {
1000 	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1001 }
1002 
1003 static inline void ceph_queue_writeback(struct inode *inode)
1004 {
1005 	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1006 }
1007 
1008 static inline void ceph_queue_check_caps(struct inode *inode)
1009 {
1010 	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1011 }
1012 
1013 static inline void ceph_queue_flush_snaps(struct inode *inode)
1014 {
1015 	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1016 }
1017 
1018 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1019 			     int mask, bool force);
1020 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1021 {
1022 	return __ceph_do_getattr(inode, NULL, mask, force);
1023 }
1024 extern int ceph_permission(struct user_namespace *mnt_userns,
1025 			   struct inode *inode, int mask);
1026 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1027 extern int ceph_setattr(struct user_namespace *mnt_userns,
1028 			struct dentry *dentry, struct iattr *attr);
1029 extern int ceph_getattr(struct user_namespace *mnt_userns,
1030 			const struct path *path, struct kstat *stat,
1031 			u32 request_mask, unsigned int flags);
1032 void ceph_inode_shutdown(struct inode *inode);
1033 
1034 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1035 {
1036 	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1037 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1038 	int state = READ_ONCE(fsc->mount_state);
1039 
1040 	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1041 }
1042 
1043 /* xattr.c */
1044 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1045 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1046 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1047 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1048 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1049 extern const struct xattr_handler *ceph_xattr_handlers[];
1050 
1051 struct ceph_acl_sec_ctx {
1052 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1053 	void *default_acl;
1054 	void *acl;
1055 #endif
1056 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1057 	void *sec_ctx;
1058 	u32 sec_ctxlen;
1059 #endif
1060 	struct ceph_pagelist *pagelist;
1061 };
1062 
1063 #ifdef CONFIG_SECURITY
1064 extern bool ceph_security_xattr_deadlock(struct inode *in);
1065 extern bool ceph_security_xattr_wanted(struct inode *in);
1066 #else
1067 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1068 {
1069 	return false;
1070 }
1071 static inline bool ceph_security_xattr_wanted(struct inode *in)
1072 {
1073 	return false;
1074 }
1075 #endif
1076 
1077 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1078 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1079 				     struct ceph_acl_sec_ctx *ctx);
1080 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1081 {
1082 	security_inode_invalidate_secctx(inode);
1083 }
1084 #else
1085 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1086 					    struct ceph_acl_sec_ctx *ctx)
1087 {
1088 	return 0;
1089 }
1090 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1091 {
1092 }
1093 #endif
1094 
1095 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1096 
1097 /* acl.c */
1098 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1099 
1100 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1101 int ceph_set_acl(struct user_namespace *mnt_userns,
1102 		 struct inode *inode, struct posix_acl *acl, int type);
1103 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1104 		       struct ceph_acl_sec_ctx *as_ctx);
1105 void ceph_init_inode_acls(struct inode *inode,
1106 			  struct ceph_acl_sec_ctx *as_ctx);
1107 
1108 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1109 {
1110        forget_all_cached_acls(inode);
1111 }
1112 
1113 #else
1114 
1115 #define ceph_get_acl NULL
1116 #define ceph_set_acl NULL
1117 
1118 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1119 				     struct ceph_acl_sec_ctx *as_ctx)
1120 {
1121 	return 0;
1122 }
1123 static inline void ceph_init_inode_acls(struct inode *inode,
1124 					struct ceph_acl_sec_ctx *as_ctx)
1125 {
1126 }
1127 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1128 {
1129 	return 0;
1130 }
1131 
1132 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1133 {
1134 }
1135 
1136 #endif
1137 
1138 /* caps.c */
1139 extern const char *ceph_cap_string(int c);
1140 extern void ceph_handle_caps(struct ceph_mds_session *session,
1141 			     struct ceph_msg *msg);
1142 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1143 				     struct ceph_cap_reservation *ctx);
1144 extern void ceph_add_cap(struct inode *inode,
1145 			 struct ceph_mds_session *session, u64 cap_id,
1146 			 unsigned issued, unsigned wanted,
1147 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1148 			 struct ceph_cap **new_cap);
1149 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1150 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1151 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1152 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1153 			 struct ceph_cap *cap);
1154 extern int ceph_is_any_caps(struct inode *inode);
1155 
1156 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1157 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1158 		      int datasync);
1159 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1160 					  struct ceph_mds_session *session);
1161 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1162 				    struct ceph_mds_session *session);
1163 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1164 				   struct ceph_inode_info *ci);
1165 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1166 					     int mds);
1167 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1168 				bool snap_rwsem_locked);
1169 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1170 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1171 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1172 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1173 					    int had);
1174 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1175 				       struct ceph_snap_context *snapc);
1176 extern void __ceph_remove_capsnap(struct inode *inode,
1177 				  struct ceph_cap_snap *capsnap,
1178 				  bool *wake_ci, bool *wake_mdsc);
1179 extern void ceph_remove_capsnap(struct inode *inode,
1180 				struct ceph_cap_snap *capsnap,
1181 				bool *wake_ci, bool *wake_mdsc);
1182 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1183 			     struct ceph_mds_session **psession);
1184 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1185 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1186 			    struct ceph_mds_session *session);
1187 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1188 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1189 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1190 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1191 				     int mds, int drop, int unless, int force);
1192 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1193 				      struct inode *dir,
1194 				      int mds, int drop, int unless);
1195 
1196 extern int ceph_get_caps(struct file *filp, int need, int want,
1197 			 loff_t endoff, int *got);
1198 extern int ceph_try_get_caps(struct inode *inode,
1199 			     int need, int want, bool nonblock, int *got);
1200 
1201 /* for counting open files by mode */
1202 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1203 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1204 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1205 			       struct ceph_mds_client *mdsc, int fmode);
1206 
1207 /* addr.c */
1208 extern const struct address_space_operations ceph_aops;
1209 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1210 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1211 extern int ceph_pool_perm_check(struct inode *inode, int need);
1212 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1213 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1214 
1215 /* file.c */
1216 extern const struct file_operations ceph_file_fops;
1217 
1218 extern int ceph_renew_caps(struct inode *inode, int fmode);
1219 extern int ceph_open(struct inode *inode, struct file *file);
1220 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1221 			    struct file *file, unsigned flags, umode_t mode);
1222 extern int ceph_release(struct inode *inode, struct file *filp);
1223 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1224 				  char *data, size_t len);
1225 
1226 /* dir.c */
1227 extern const struct file_operations ceph_dir_fops;
1228 extern const struct file_operations ceph_snapdir_fops;
1229 extern const struct inode_operations ceph_dir_iops;
1230 extern const struct inode_operations ceph_snapdir_iops;
1231 extern const struct dentry_operations ceph_dentry_ops;
1232 
1233 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1234 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1235 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1236 			       struct dentry *dentry);
1237 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1238 					 struct dentry *dentry, int err);
1239 
1240 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1241 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1242 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1243 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1244 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1245 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1246 
1247 /* ioctl.c */
1248 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1249 
1250 /* export.c */
1251 extern const struct export_operations ceph_export_ops;
1252 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1253 
1254 /* locks.c */
1255 extern __init void ceph_flock_init(void);
1256 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1257 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1258 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1259 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1260 				       struct ceph_filelock *flocks,
1261 				       int num_fcntl_locks,
1262 				       int num_flock_locks);
1263 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1264 				  struct ceph_pagelist *pagelist,
1265 				  int num_fcntl_locks, int num_flock_locks);
1266 
1267 /* debugfs.c */
1268 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1269 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1270 
1271 /* quota.c */
1272 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1273 {
1274 	return ci->i_max_files || ci->i_max_bytes;
1275 }
1276 
1277 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1278 
1279 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1280 				       u64 max_bytes, u64 max_files)
1281 {
1282 	bool had_quota, has_quota;
1283 	had_quota = __ceph_has_any_quota(ci);
1284 	ci->i_max_bytes = max_bytes;
1285 	ci->i_max_files = max_files;
1286 	has_quota = __ceph_has_any_quota(ci);
1287 
1288 	if (had_quota != has_quota)
1289 		ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1290 }
1291 
1292 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1293 			      struct ceph_mds_session *session,
1294 			      struct ceph_msg *msg);
1295 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1296 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1297 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1298 					     loff_t newlen);
1299 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1300 						loff_t newlen);
1301 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1302 				     struct kstatfs *buf);
1303 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1304 
1305 #endif /* _FS_CEPH_SUPER_H */
1306