xref: /linux/fs/ceph/super.h (revision b37042b2bb7cd751f03b73afb90364a418d870f4)
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 #include <linux/netfs.h>
21 #include <linux/fscache.h>
22 #include <linux/hashtable.h>
23 
24 #include <linux/ceph/libceph.h>
25 
26 /* large granularity for statfs utilization stats to facilitate
27  * large volume sizes on 32-bit machines. */
28 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
29 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
30 #define CEPH_4K_BLOCK_SHIFT 12  /* 4 KB */
31 
32 #define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
33 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
34 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
35 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
36 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
37 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
38 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
39 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
40 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
41 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
42 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
43 #define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
44 #define CEPH_MOUNT_OPT_NOPAGECACHE     (1<<16) /* bypass pagecache altogether */
45 
46 #define CEPH_MOUNT_OPT_DEFAULT			\
47 	(CEPH_MOUNT_OPT_DCACHE |		\
48 	 CEPH_MOUNT_OPT_NOCOPYFROM |		\
49 	 CEPH_MOUNT_OPT_ASYNC_DIROPS)
50 
51 #define ceph_set_mount_opt(fsc, opt) \
52 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
53 #define ceph_clear_mount_opt(fsc, opt) \
54 	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
55 #define ceph_test_mount_opt(fsc, opt) \
56 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
57 
58 /* max size of osd read request, limited by libceph */
59 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
60 /* osd has a configurable limitaion of max write size.
61  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
62 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
63 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
64 #define CEPH_MAX_READDIR_DEFAULT        1024
65 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
66 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
67 
68 /*
69  * Delay telling the MDS we no longer want caps, in case we reopen
70  * the file.  Delay a minimum amount of time, even if we send a cap
71  * message for some other reason.  Otherwise, take the oppotunity to
72  * update the mds to avoid sending another message later.
73  */
74 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
75 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
76 
77 struct ceph_mount_options {
78 	unsigned int flags;
79 
80 	unsigned int wsize;            /* max write size */
81 	unsigned int rsize;            /* max read size */
82 	unsigned int rasize;           /* max readahead */
83 	unsigned int congestion_kb;    /* max writeback in flight */
84 	unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
85 	int caps_max;
86 	unsigned int max_readdir;       /* max readdir result (entries) */
87 	unsigned int max_readdir_bytes; /* max readdir result (bytes) */
88 
89 	bool new_dev_syntax;
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 	char *mon_addr;
101 };
102 
103 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
104 
105 struct ceph_fs_client {
106 	struct super_block *sb;
107 
108 	struct list_head metric_wakeup;
109 
110 	struct ceph_mount_options *mount_options;
111 	struct ceph_client *client;
112 
113 	int mount_state;
114 
115 	bool blocklisted;
116 
117 	bool have_copy_from2;
118 
119 	u32 filp_gen;
120 	loff_t max_file_size;
121 
122 	struct ceph_mds_client *mdsc;
123 
124 	atomic_long_t writeback_count;
125 	bool write_congested;
126 
127 	struct workqueue_struct *inode_wq;
128 	struct workqueue_struct *cap_wq;
129 
130 	DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
131 	spinlock_t async_unlink_conflict_lock;
132 
133 #ifdef CONFIG_DEBUG_FS
134 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
135 	struct dentry *debugfs_congestion_kb;
136 	struct dentry *debugfs_bdi;
137 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
138 	struct dentry *debugfs_status;
139 	struct dentry *debugfs_mds_sessions;
140 	struct dentry *debugfs_metrics_dir;
141 #endif
142 
143 #ifdef CONFIG_CEPH_FSCACHE
144 	struct fscache_volume *fscache;
145 #endif
146 };
147 
148 
149 /*
150  * File i/o capability.  This tracks shared state with the metadata
151  * server that allows us to cache or writeback attributes or to read
152  * and write data.  For any given inode, we should have one or more
153  * capabilities, one issued by each metadata server, and our
154  * cumulative access is the OR of all issued capabilities.
155  *
156  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
157  * session capability lists.
158  */
159 struct ceph_cap {
160 	struct ceph_inode_info *ci;
161 	struct rb_node ci_node;          /* per-ci cap tree */
162 	struct ceph_mds_session *session;
163 	struct list_head session_caps;   /* per-session caplist */
164 	u64 cap_id;       /* unique cap id (mds provided) */
165 	union {
166 		/* in-use caps */
167 		struct {
168 			int issued;       /* latest, from the mds */
169 			int implemented;  /* implemented superset of
170 					     issued (for revocation) */
171 			int mds;	  /* mds index for this cap */
172 			int mds_wanted;   /* caps wanted from this mds */
173 		};
174 		/* caps to release */
175 		struct {
176 			u64 cap_ino;
177 			int queue_release;
178 		};
179 	};
180 	u32 seq, issue_seq, mseq;
181 	u32 cap_gen;      /* active/stale cycle */
182 	unsigned long last_used;
183 	struct list_head caps_item;
184 };
185 
186 #define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
187 #define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
188 #define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
189 
190 struct ceph_cap_flush {
191 	u64 tid;
192 	int caps;
193 	bool wake; /* wake up flush waiters when finish ? */
194 	bool is_capsnap; /* true means capsnap */
195 	struct list_head g_list; // global
196 	struct list_head i_list; // per inode
197 };
198 
199 /*
200  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
201  * we first complete any in-process sync writes and writeback any dirty
202  * data before flushing the snapped state (tracked here) back to the MDS.
203  */
204 struct ceph_cap_snap {
205 	refcount_t nref;
206 	struct list_head ci_item;
207 
208 	struct ceph_cap_flush cap_flush;
209 
210 	u64 follows;
211 	int issued, dirty;
212 	struct ceph_snap_context *context;
213 
214 	umode_t mode;
215 	kuid_t uid;
216 	kgid_t gid;
217 
218 	struct ceph_buffer *xattr_blob;
219 	u64 xattr_version;
220 
221 	u64 size;
222 	u64 change_attr;
223 	struct timespec64 mtime, atime, ctime, btime;
224 	u64 time_warp_seq;
225 	u64 truncate_size;
226 	u32 truncate_seq;
227 	int writing;   /* a sync write is still in progress */
228 	int dirty_pages;     /* dirty pages awaiting writeback */
229 	bool inline_data;
230 	bool need_flush;
231 };
232 
233 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
234 {
235 	if (refcount_dec_and_test(&capsnap->nref)) {
236 		if (capsnap->xattr_blob)
237 			ceph_buffer_put(capsnap->xattr_blob);
238 		kmem_cache_free(ceph_cap_snap_cachep, capsnap);
239 	}
240 }
241 
242 /*
243  * The frag tree describes how a directory is fragmented, potentially across
244  * multiple metadata servers.  It is also used to indicate points where
245  * metadata authority is delegated, and whether/where metadata is replicated.
246  *
247  * A _leaf_ frag will be present in the i_fragtree IFF there is
248  * delegation info.  That is, if mds >= 0 || ndist > 0.
249  */
250 #define CEPH_MAX_DIRFRAG_REP 4
251 
252 struct ceph_inode_frag {
253 	struct rb_node node;
254 
255 	/* fragtree state */
256 	u32 frag;
257 	int split_by;         /* i.e. 2^(split_by) children */
258 
259 	/* delegation and replication info */
260 	int mds;              /* -1 if same authority as parent */
261 	int ndist;            /* >0 if replicated */
262 	int dist[CEPH_MAX_DIRFRAG_REP];
263 };
264 
265 /*
266  * We cache inode xattrs as an encoded blob until they are first used,
267  * at which point we parse them into an rbtree.
268  */
269 struct ceph_inode_xattr {
270 	struct rb_node node;
271 
272 	const char *name;
273 	int name_len;
274 	const char *val;
275 	int val_len;
276 	int dirty;
277 
278 	int should_free_name;
279 	int should_free_val;
280 };
281 
282 /*
283  * Ceph dentry state
284  */
285 struct ceph_dentry_info {
286 	struct dentry *dentry;
287 	struct ceph_mds_session *lease_session;
288 	struct list_head lease_list;
289 	struct hlist_node hnode;
290 	unsigned long flags;
291 	int lease_shared_gen;
292 	u32 lease_gen;
293 	u32 lease_seq;
294 	unsigned long lease_renew_after, lease_renew_from;
295 	unsigned long time;
296 	u64 offset;
297 };
298 
299 #define CEPH_DENTRY_REFERENCED		(1 << 0)
300 #define CEPH_DENTRY_LEASE_LIST		(1 << 1)
301 #define CEPH_DENTRY_SHRINK_LIST		(1 << 2)
302 #define CEPH_DENTRY_PRIMARY_LINK	(1 << 3)
303 #define CEPH_DENTRY_ASYNC_UNLINK_BIT	(4)
304 #define CEPH_DENTRY_ASYNC_UNLINK	(1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
305 #define CEPH_DENTRY_ASYNC_CREATE_BIT	(5)
306 #define CEPH_DENTRY_ASYNC_CREATE	(1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
307 
308 struct ceph_inode_xattrs_info {
309 	/*
310 	 * (still encoded) xattr blob. we avoid the overhead of parsing
311 	 * this until someone actually calls getxattr, etc.
312 	 *
313 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
314 	 * NULL means we don't know.
315 	*/
316 	struct ceph_buffer *blob, *prealloc_blob;
317 
318 	struct rb_root index;
319 	bool dirty;
320 	int count;
321 	int names_size;
322 	int vals_size;
323 	u64 version, index_version;
324 };
325 
326 /*
327  * Ceph inode.
328  */
329 struct ceph_inode_info {
330 	struct netfs_inode netfs; /* Netfslib context and vfs inode */
331 	struct ceph_vino i_vino;   /* ceph ino + snap */
332 
333 	spinlock_t i_ceph_lock;
334 
335 	u64 i_version;
336 	u64 i_inline_version;
337 	u32 i_time_warp_seq;
338 
339 	unsigned long i_ceph_flags;
340 	atomic64_t i_release_count;
341 	atomic64_t i_ordered_count;
342 	atomic64_t i_complete_seq[2];
343 
344 	struct ceph_dir_layout i_dir_layout;
345 	struct ceph_file_layout i_layout;
346 	struct ceph_file_layout i_cached_layout;	// for async creates
347 	char *i_symlink;
348 
349 	/* for dirs */
350 	struct timespec64 i_rctime;
351 	u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
352 	u64 i_files, i_subdirs;
353 
354 	/* quotas */
355 	u64 i_max_bytes, i_max_files;
356 
357 	s32 i_dir_pin;
358 
359 	struct rb_root i_fragtree;
360 	int i_fragtree_nsplits;
361 	struct mutex i_fragtree_mutex;
362 
363 	struct ceph_inode_xattrs_info i_xattrs;
364 
365 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
366 	 * s_mutex. */
367 	struct rb_root i_caps;           /* cap list */
368 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
369 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
370 
371 	/*
372 	 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
373 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
374 	 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
375 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
376 	 * be tested under the i_ceph_lock. Changing anything requires both.
377 	 */
378 	struct list_head i_dirty_item;
379 
380 	/*
381 	 * Link to session's s_cap_flushing list. Protected in a similar
382 	 * fashion to i_dirty_item, but also by the s_mutex for changes. The
383 	 * s_cap_flushing list can be walked while holding either the s_mutex
384 	 * or msdc->cap_dirty_lock. List presence can also be checked while
385 	 * holding the i_ceph_lock for this inode.
386 	 */
387 	struct list_head i_flushing_item;
388 
389 	/* we need to track cap writeback on a per-cap-bit basis, to allow
390 	 * overlapping, pipelined cap flushes to the mds.  we can probably
391 	 * reduce the tid to 8 bits if we're concerned about inode size. */
392 	struct ceph_cap_flush *i_prealloc_cap_flush;
393 	struct list_head i_cap_flush_list;
394 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
395 	unsigned long i_hold_caps_max; /* jiffies */
396 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
397 	struct ceph_cap_reservation i_cap_migration_resv;
398 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
399 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
400 						    dirty|flushing caps */
401 	unsigned i_snap_caps;           /* cap bits for snapped files */
402 
403 	unsigned long i_last_rd;
404 	unsigned long i_last_wr;
405 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
406 
407 	struct mutex i_truncate_mutex;
408 	u32 i_truncate_seq;        /* last truncate to smaller size */
409 	u64 i_truncate_size;       /*  and the size we last truncated down to */
410 	int i_truncate_pending;    /*  still need to call vmtruncate */
411 
412 	u64 i_max_size;            /* max file size authorized by mds */
413 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
414 	u64 i_wanted_max_size;     /* offset we'd like to write too */
415 	u64 i_requested_max_size;  /* max_size we've requested */
416 
417 	/* held references to caps */
418 	int i_pin_ref;
419 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
420 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
421 	atomic_t i_filelock_ref;
422 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
423 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
424 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
425 
426 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
427 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
428 	spinlock_t i_unsafe_lock;
429 
430 	union {
431 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
432 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
433 	};
434 	struct list_head i_snap_realm_item;
435 	struct list_head i_snap_flush_item;
436 	struct timespec64 i_btime;
437 	struct timespec64 i_snap_btime;
438 
439 	struct work_struct i_work;
440 	unsigned long  i_work_mask;
441 };
442 
443 static inline struct ceph_inode_info *
444 ceph_inode(const struct inode *inode)
445 {
446 	return container_of(inode, struct ceph_inode_info, netfs.inode);
447 }
448 
449 static inline struct ceph_fs_client *
450 ceph_inode_to_client(const struct inode *inode)
451 {
452 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
453 }
454 
455 static inline struct ceph_fs_client *
456 ceph_sb_to_client(const struct super_block *sb)
457 {
458 	return (struct ceph_fs_client *)sb->s_fs_info;
459 }
460 
461 static inline struct ceph_mds_client *
462 ceph_sb_to_mdsc(const struct super_block *sb)
463 {
464 	return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
465 }
466 
467 static inline struct ceph_vino
468 ceph_vino(const struct inode *inode)
469 {
470 	return ceph_inode(inode)->i_vino;
471 }
472 
473 static inline u32 ceph_ino_to_ino32(u64 vino)
474 {
475 	u32 ino = vino & 0xffffffff;
476 	ino ^= vino >> 32;
477 	if (!ino)
478 		ino = 2;
479 	return ino;
480 }
481 
482 /*
483  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
484  * some arches. We generally do not use this value inside the ceph driver, but
485  * we do want to set it to something, so that generic vfs code has an
486  * appropriate value for tracepoints and the like.
487  */
488 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
489 {
490 	if (sizeof(ino_t) == sizeof(u32))
491 		return ceph_ino_to_ino32(vino.ino);
492 	return (ino_t)vino.ino;
493 }
494 
495 /* for printf-style formatting */
496 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
497 
498 static inline u64 ceph_ino(struct inode *inode)
499 {
500 	return ceph_inode(inode)->i_vino.ino;
501 }
502 
503 static inline u64 ceph_snap(struct inode *inode)
504 {
505 	return ceph_inode(inode)->i_vino.snap;
506 }
507 
508 /**
509  * ceph_present_ino - format an inode number for presentation to userland
510  * @sb: superblock where the inode lives
511  * @ino: inode number to (possibly) convert
512  *
513  * If the user mounted with the ino32 option, then the 64-bit value needs
514  * to be converted to something that can fit inside 32 bits. Note that
515  * internal kernel code never uses this value, so this is entirely for
516  * userland consumption.
517  */
518 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
519 {
520 	if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
521 		return ceph_ino_to_ino32(ino);
522 	return ino;
523 }
524 
525 static inline u64 ceph_present_inode(struct inode *inode)
526 {
527 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));
528 }
529 
530 static inline int ceph_ino_compare(struct inode *inode, void *data)
531 {
532 	struct ceph_vino *pvino = (struct ceph_vino *)data;
533 	struct ceph_inode_info *ci = ceph_inode(inode);
534 	return ci->i_vino.ino == pvino->ino &&
535 		ci->i_vino.snap == pvino->snap;
536 }
537 
538 /*
539  * The MDS reserves a set of inodes for its own usage. These should never
540  * be accessible by clients, and so the MDS has no reason to ever hand these
541  * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
542  *
543  * These come from src/mds/mdstypes.h in the ceph sources.
544  */
545 #define CEPH_MAX_MDS			0x100
546 #define CEPH_NUM_STRAY			10
547 #define CEPH_MDS_INO_MDSDIR_OFFSET	(1 * CEPH_MAX_MDS)
548 #define CEPH_MDS_INO_LOG_OFFSET		(2 * CEPH_MAX_MDS)
549 #define CEPH_INO_SYSTEM_BASE		((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
550 
551 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
552 {
553 	if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
554 	    vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
555 		return false;
556 
557 	/* Don't warn on mdsdirs */
558 	WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
559 			"Attempt to access reserved inode number 0x%llx",
560 			vino.ino);
561 	return true;
562 }
563 
564 static inline struct inode *ceph_find_inode(struct super_block *sb,
565 					    struct ceph_vino vino)
566 {
567 	if (ceph_vino_is_reserved(vino))
568 		return NULL;
569 
570 	/*
571 	 * NB: The hashval will be run through the fs/inode.c hash function
572 	 * anyway, so there is no need to squash the inode number down to
573 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.
574 	 */
575 	return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
576 }
577 
578 
579 /*
580  * Ceph inode.
581  */
582 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
583 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
584 #define CEPH_I_POOL_PERM	(1 << 3)  /* pool rd/wr bits are valid */
585 #define CEPH_I_POOL_RD		(1 << 4)  /* can read from pool */
586 #define CEPH_I_POOL_WR		(1 << 5)  /* can write to pool */
587 #define CEPH_I_SEC_INITED	(1 << 6)  /* security initialized */
588 #define CEPH_I_KICK_FLUSH	(1 << 7)  /* kick flushing caps */
589 #define CEPH_I_FLUSH_SNAPS	(1 << 8)  /* need flush snapss */
590 #define CEPH_I_ERROR_WRITE	(1 << 9) /* have seen write errors */
591 #define CEPH_I_ERROR_FILELOCK	(1 << 10) /* have seen file lock errors */
592 #define CEPH_I_ODIRECT		(1 << 11) /* inode in direct I/O mode */
593 #define CEPH_ASYNC_CREATE_BIT	(12)	  /* async create in flight for this */
594 #define CEPH_I_ASYNC_CREATE	(1 << CEPH_ASYNC_CREATE_BIT)
595 #define CEPH_I_SHUTDOWN		(1 << 13) /* inode is no longer usable */
596 
597 /*
598  * Masks of ceph inode work.
599  */
600 #define CEPH_I_WORK_WRITEBACK		0
601 #define CEPH_I_WORK_INVALIDATE_PAGES	1
602 #define CEPH_I_WORK_VMTRUNCATE		2
603 #define CEPH_I_WORK_CHECK_CAPS		3
604 #define CEPH_I_WORK_FLUSH_SNAPS		4
605 
606 /*
607  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
608  * and then clear it when they start succeeding. Note that we do a lockless
609  * check first, and only take the lock if it looks like it needs to be changed.
610  * The write submission code just takes this as a hint, so we're not too
611  * worried if a few slip through in either direction.
612  */
613 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
614 {
615 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
616 		spin_lock(&ci->i_ceph_lock);
617 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
618 		spin_unlock(&ci->i_ceph_lock);
619 	}
620 }
621 
622 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
623 {
624 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
625 		spin_lock(&ci->i_ceph_lock);
626 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
627 		spin_unlock(&ci->i_ceph_lock);
628 	}
629 }
630 
631 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
632 					   long long release_count,
633 					   long long ordered_count)
634 {
635 	/*
636 	 * Makes sure operations that setup readdir cache (update page
637 	 * cache and i_size) are strongly ordered w.r.t. the following
638 	 * atomic64_set() operations.
639 	 */
640 	smp_mb();
641 	atomic64_set(&ci->i_complete_seq[0], release_count);
642 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
643 }
644 
645 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
646 {
647 	atomic64_inc(&ci->i_release_count);
648 }
649 
650 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
651 {
652 	atomic64_inc(&ci->i_ordered_count);
653 }
654 
655 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
656 {
657 	return atomic64_read(&ci->i_complete_seq[0]) ==
658 		atomic64_read(&ci->i_release_count);
659 }
660 
661 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
662 {
663 	return  atomic64_read(&ci->i_complete_seq[0]) ==
664 		atomic64_read(&ci->i_release_count) &&
665 		atomic64_read(&ci->i_complete_seq[1]) ==
666 		atomic64_read(&ci->i_ordered_count);
667 }
668 
669 static inline void ceph_dir_clear_complete(struct inode *inode)
670 {
671 	__ceph_dir_clear_complete(ceph_inode(inode));
672 }
673 
674 static inline void ceph_dir_clear_ordered(struct inode *inode)
675 {
676 	__ceph_dir_clear_ordered(ceph_inode(inode));
677 }
678 
679 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
680 {
681 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
682 	smp_rmb();
683 	return ret;
684 }
685 
686 /* find a specific frag @f */
687 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
688 						u32 f);
689 
690 /*
691  * choose fragment for value @v.  copy frag content to pfrag, if leaf
692  * exists
693  */
694 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
695 			    struct ceph_inode_frag *pfrag,
696 			    int *found);
697 
698 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
699 {
700 	return (struct ceph_dentry_info *)dentry->d_fsdata;
701 }
702 
703 /*
704  * caps helpers
705  */
706 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
707 {
708 	return !RB_EMPTY_ROOT(&ci->i_caps);
709 }
710 
711 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
712 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
713 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
714 					  int t);
715 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
716 				    struct ceph_cap *cap);
717 
718 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
719 {
720 	int issued;
721 	spin_lock(&ci->i_ceph_lock);
722 	issued = __ceph_caps_issued(ci, NULL);
723 	spin_unlock(&ci->i_ceph_lock);
724 	return issued;
725 }
726 
727 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
728 					       int mask, int touch)
729 {
730 	int r;
731 	spin_lock(&ci->i_ceph_lock);
732 	r = __ceph_caps_issued_mask_metric(ci, mask, touch);
733 	spin_unlock(&ci->i_ceph_lock);
734 	return r;
735 }
736 
737 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
738 {
739 	return ci->i_dirty_caps | ci->i_flushing_caps;
740 }
741 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
742 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
743 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
744 				  struct ceph_cap_flush **pcf);
745 
746 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
747 				      struct ceph_cap *ocap, int mask);
748 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
749 extern int __ceph_caps_used(struct ceph_inode_info *ci);
750 
751 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
752 {
753 	return ci->i_nr_by_mode[0];
754 }
755 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
756 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
757 
758 /* what the mds thinks we want */
759 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
760 
761 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
762 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
763 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
764 				     struct ceph_mount_options *fsopt);
765 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
766 			     struct ceph_cap_reservation *ctx, int need);
767 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
768 			       struct ceph_cap_reservation *ctx);
769 extern void ceph_reservation_status(struct ceph_fs_client *client,
770 				    int *total, int *avail, int *used,
771 				    int *reserved, int *min);
772 extern void change_auth_cap_ses(struct ceph_inode_info *ci,
773 				struct ceph_mds_session *session);
774 
775 
776 
777 /*
778  * we keep buffered readdir results attached to file->private_data
779  */
780 #define CEPH_F_SYNC     1
781 #define CEPH_F_ATEND    2
782 
783 struct ceph_file_info {
784 	short fmode;     /* initialized on open */
785 	short flags;     /* CEPH_F_* */
786 
787 	spinlock_t rw_contexts_lock;
788 	struct list_head rw_contexts;
789 
790 	u32 filp_gen;
791 	atomic_t num_locks;
792 };
793 
794 struct ceph_dir_file_info {
795 	struct ceph_file_info file_info;
796 
797 	/* readdir: position within the dir */
798 	u32 frag;
799 	struct ceph_mds_request *last_readdir;
800 
801 	/* readdir: position within a frag */
802 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
803 	char *last_name;       /* last entry in previous chunk */
804 	long long dir_release_count;
805 	long long dir_ordered_count;
806 	int readdir_cache_idx;
807 
808 	/* used for -o dirstat read() on directory thing */
809 	char *dir_info;
810 	int dir_info_len;
811 };
812 
813 struct ceph_rw_context {
814 	struct list_head list;
815 	struct task_struct *thread;
816 	int caps;
817 };
818 
819 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
820 	struct ceph_rw_context _name = {	\
821 		.thread = current,		\
822 		.caps = _caps,			\
823 	}
824 
825 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
826 				       struct ceph_rw_context *ctx)
827 {
828 	spin_lock(&cf->rw_contexts_lock);
829 	list_add(&ctx->list, &cf->rw_contexts);
830 	spin_unlock(&cf->rw_contexts_lock);
831 }
832 
833 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
834 				       struct ceph_rw_context *ctx)
835 {
836 	spin_lock(&cf->rw_contexts_lock);
837 	list_del(&ctx->list);
838 	spin_unlock(&cf->rw_contexts_lock);
839 }
840 
841 static inline struct ceph_rw_context*
842 ceph_find_rw_context(struct ceph_file_info *cf)
843 {
844 	struct ceph_rw_context *ctx, *found = NULL;
845 	spin_lock(&cf->rw_contexts_lock);
846 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
847 		if (ctx->thread == current) {
848 			found = ctx;
849 			break;
850 		}
851 	}
852 	spin_unlock(&cf->rw_contexts_lock);
853 	return found;
854 }
855 
856 struct ceph_readdir_cache_control {
857 	struct page  *page;
858 	struct dentry **dentries;
859 	int index;
860 };
861 
862 /*
863  * A "snap realm" describes a subset of the file hierarchy sharing
864  * the same set of snapshots that apply to it.  The realms themselves
865  * are organized into a hierarchy, such that children inherit (some of)
866  * the snapshots of their parents.
867  *
868  * All inodes within the realm that have capabilities are linked into a
869  * per-realm list.
870  */
871 struct ceph_snap_realm {
872 	u64 ino;
873 	struct inode *inode;
874 	atomic_t nref;
875 	struct rb_node node;
876 
877 	u64 created, seq;
878 	u64 parent_ino;
879 	u64 parent_since;   /* snapid when our current parent became so */
880 
881 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
882 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
883 	u64 *snaps;                   /* snaps specific to this realm */
884 	u32 num_snaps;
885 
886 	struct ceph_snap_realm *parent;
887 	struct list_head children;       /* list of child realms */
888 	struct list_head child_item;
889 
890 	struct list_head empty_item;     /* if i have ref==0 */
891 
892 	struct list_head dirty_item;     /* if realm needs new context */
893 
894 	struct list_head rebuild_item;   /* rebuild snap realms _downward_ in hierarchy */
895 
896 	/* the current set of snaps for this realm */
897 	struct ceph_snap_context *cached_context;
898 
899 	struct list_head inodes_with_caps;
900 	spinlock_t inodes_with_caps_lock;
901 };
902 
903 static inline int default_congestion_kb(void)
904 {
905 	int congestion_kb;
906 
907 	/*
908 	 * Copied from NFS
909 	 *
910 	 * congestion size, scale with available memory.
911 	 *
912 	 *  64MB:    8192k
913 	 * 128MB:   11585k
914 	 * 256MB:   16384k
915 	 * 512MB:   23170k
916 	 *   1GB:   32768k
917 	 *   2GB:   46340k
918 	 *   4GB:   65536k
919 	 *   8GB:   92681k
920 	 *  16GB:  131072k
921 	 *
922 	 * This allows larger machines to have larger/more transfers.
923 	 * Limit the default to 256M
924 	 */
925 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
926 	if (congestion_kb > 256*1024)
927 		congestion_kb = 256*1024;
928 
929 	return congestion_kb;
930 }
931 
932 
933 /* super.c */
934 extern int ceph_force_reconnect(struct super_block *sb);
935 /* snap.c */
936 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
937 					       u64 ino);
938 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
939 				struct ceph_snap_realm *realm);
940 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
941 				struct ceph_snap_realm *realm);
942 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
943 				  void *p, void *e, bool deletion,
944 				  struct ceph_snap_realm **realm_ret);
945 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
946 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
947 			     struct ceph_mds_session *session,
948 			     struct ceph_msg *msg);
949 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
950 				  struct ceph_cap_snap *capsnap);
951 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
952 
953 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
954 						   u64 snap);
955 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
956 				struct ceph_snapid_map *sm);
957 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
958 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
959 void ceph_umount_begin(struct super_block *sb);
960 
961 
962 /*
963  * a cap_snap is "pending" if it is still awaiting an in-progress
964  * sync write (that may/may not still update size, mtime, etc.).
965  */
966 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
967 {
968 	return !list_empty(&ci->i_cap_snaps) &&
969 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
970 			       ci_item)->writing;
971 }
972 
973 /* inode.c */
974 struct ceph_mds_reply_info_in;
975 struct ceph_mds_reply_dirfrag;
976 
977 extern const struct inode_operations ceph_file_iops;
978 
979 extern struct inode *ceph_alloc_inode(struct super_block *sb);
980 extern void ceph_evict_inode(struct inode *inode);
981 extern void ceph_free_inode(struct inode *inode);
982 
983 extern struct inode *ceph_get_inode(struct super_block *sb,
984 				    struct ceph_vino vino);
985 extern struct inode *ceph_get_snapdir(struct inode *parent);
986 extern int ceph_fill_file_size(struct inode *inode, int issued,
987 			       u32 truncate_seq, u64 truncate_size, u64 size);
988 extern void ceph_fill_file_time(struct inode *inode, int issued,
989 				u64 time_warp_seq, struct timespec64 *ctime,
990 				struct timespec64 *mtime,
991 				struct timespec64 *atime);
992 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
993 		    struct ceph_mds_reply_info_in *iinfo,
994 		    struct ceph_mds_reply_dirfrag *dirinfo,
995 		    struct ceph_mds_session *session, int cap_fmode,
996 		    struct ceph_cap_reservation *caps_reservation);
997 extern int ceph_fill_trace(struct super_block *sb,
998 			   struct ceph_mds_request *req);
999 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1000 				    struct ceph_mds_session *session);
1001 
1002 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
1003 
1004 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
1005 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
1006 
1007 void ceph_queue_inode_work(struct inode *inode, int work_bit);
1008 
1009 static inline void ceph_queue_vmtruncate(struct inode *inode)
1010 {
1011 	ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1012 }
1013 
1014 static inline void ceph_queue_invalidate(struct inode *inode)
1015 {
1016 	ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1017 }
1018 
1019 static inline void ceph_queue_writeback(struct inode *inode)
1020 {
1021 	ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1022 }
1023 
1024 static inline void ceph_queue_check_caps(struct inode *inode)
1025 {
1026 	ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1027 }
1028 
1029 static inline void ceph_queue_flush_snaps(struct inode *inode)
1030 {
1031 	ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1032 }
1033 
1034 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1035 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1036 			     int mask, bool force);
1037 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1038 {
1039 	return __ceph_do_getattr(inode, NULL, mask, force);
1040 }
1041 extern int ceph_permission(struct user_namespace *mnt_userns,
1042 			   struct inode *inode, int mask);
1043 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1044 extern int ceph_setattr(struct user_namespace *mnt_userns,
1045 			struct dentry *dentry, struct iattr *attr);
1046 extern int ceph_getattr(struct user_namespace *mnt_userns,
1047 			const struct path *path, struct kstat *stat,
1048 			u32 request_mask, unsigned int flags);
1049 void ceph_inode_shutdown(struct inode *inode);
1050 
1051 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1052 {
1053 	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1054 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1055 	int state = READ_ONCE(fsc->mount_state);
1056 
1057 	return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1058 }
1059 
1060 /* xattr.c */
1061 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1062 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1063 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1064 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1065 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1066 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1067 extern const struct xattr_handler *ceph_xattr_handlers[];
1068 
1069 struct ceph_acl_sec_ctx {
1070 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1071 	void *default_acl;
1072 	void *acl;
1073 #endif
1074 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1075 	void *sec_ctx;
1076 	u32 sec_ctxlen;
1077 #endif
1078 	struct ceph_pagelist *pagelist;
1079 };
1080 
1081 #ifdef CONFIG_SECURITY
1082 extern bool ceph_security_xattr_deadlock(struct inode *in);
1083 extern bool ceph_security_xattr_wanted(struct inode *in);
1084 #else
1085 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1086 {
1087 	return false;
1088 }
1089 static inline bool ceph_security_xattr_wanted(struct inode *in)
1090 {
1091 	return false;
1092 }
1093 #endif
1094 
1095 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1096 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1097 				     struct ceph_acl_sec_ctx *ctx);
1098 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1099 {
1100 	security_inode_invalidate_secctx(inode);
1101 }
1102 #else
1103 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1104 					    struct ceph_acl_sec_ctx *ctx)
1105 {
1106 	return 0;
1107 }
1108 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1109 {
1110 }
1111 #endif
1112 
1113 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1114 
1115 /* acl.c */
1116 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1117 
1118 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1119 int ceph_set_acl(struct user_namespace *mnt_userns,
1120 		 struct inode *inode, struct posix_acl *acl, int type);
1121 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1122 		       struct ceph_acl_sec_ctx *as_ctx);
1123 void ceph_init_inode_acls(struct inode *inode,
1124 			  struct ceph_acl_sec_ctx *as_ctx);
1125 
1126 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1127 {
1128        forget_all_cached_acls(inode);
1129 }
1130 
1131 #else
1132 
1133 #define ceph_get_acl NULL
1134 #define ceph_set_acl NULL
1135 
1136 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1137 				     struct ceph_acl_sec_ctx *as_ctx)
1138 {
1139 	return 0;
1140 }
1141 static inline void ceph_init_inode_acls(struct inode *inode,
1142 					struct ceph_acl_sec_ctx *as_ctx)
1143 {
1144 }
1145 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1146 {
1147 	return 0;
1148 }
1149 
1150 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1151 {
1152 }
1153 
1154 #endif
1155 
1156 /* caps.c */
1157 extern const char *ceph_cap_string(int c);
1158 extern void ceph_handle_caps(struct ceph_mds_session *session,
1159 			     struct ceph_msg *msg);
1160 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1161 				     struct ceph_cap_reservation *ctx);
1162 extern void ceph_add_cap(struct inode *inode,
1163 			 struct ceph_mds_session *session, u64 cap_id,
1164 			 unsigned issued, unsigned wanted,
1165 			 unsigned cap, unsigned seq, u64 realmino, int flags,
1166 			 struct ceph_cap **new_cap);
1167 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1168 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1169 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1170 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1171 			 struct ceph_cap *cap);
1172 extern int ceph_is_any_caps(struct inode *inode);
1173 
1174 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1175 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1176 		      int datasync);
1177 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1178 					  struct ceph_mds_session *session);
1179 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1180 				    struct ceph_mds_session *session);
1181 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1182 				   struct ceph_inode_info *ci);
1183 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1184 					     int mds);
1185 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1186 				bool snap_rwsem_locked);
1187 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1188 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1189 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1190 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1191 					    int had);
1192 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1193 				       struct ceph_snap_context *snapc);
1194 extern void __ceph_remove_capsnap(struct inode *inode,
1195 				  struct ceph_cap_snap *capsnap,
1196 				  bool *wake_ci, bool *wake_mdsc);
1197 extern void ceph_remove_capsnap(struct inode *inode,
1198 				struct ceph_cap_snap *capsnap,
1199 				bool *wake_ci, bool *wake_mdsc);
1200 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1201 			     struct ceph_mds_session **psession);
1202 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1203 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1204 			    struct ceph_mds_session *session);
1205 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1206 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1207 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1208 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1209 				     int mds, int drop, int unless, int force);
1210 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1211 				      struct inode *dir,
1212 				      int mds, int drop, int unless);
1213 
1214 extern int ceph_get_caps(struct file *filp, int need, int want,
1215 			 loff_t endoff, int *got);
1216 extern int ceph_try_get_caps(struct inode *inode,
1217 			     int need, int want, bool nonblock, int *got);
1218 
1219 /* for counting open files by mode */
1220 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1221 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1222 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1223 			       struct ceph_mds_client *mdsc, int fmode);
1224 
1225 /* addr.c */
1226 extern const struct address_space_operations ceph_aops;
1227 extern const struct netfs_request_ops ceph_netfs_ops;
1228 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1229 extern int ceph_uninline_data(struct file *file);
1230 extern int ceph_pool_perm_check(struct inode *inode, int need);
1231 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1232 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1233 
1234 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1235 {
1236 	if (ci->i_inline_version == CEPH_INLINE_NONE ||
1237 	    ci->i_inline_version == 1) /* initial version, no data */
1238 		return false;
1239 	return true;
1240 }
1241 
1242 /* file.c */
1243 extern const struct file_operations ceph_file_fops;
1244 
1245 extern int ceph_renew_caps(struct inode *inode, int fmode);
1246 extern int ceph_open(struct inode *inode, struct file *file);
1247 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1248 			    struct file *file, unsigned flags, umode_t mode);
1249 extern int ceph_release(struct inode *inode, struct file *filp);
1250 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1251 				  char *data, size_t len);
1252 
1253 /* dir.c */
1254 extern const struct file_operations ceph_dir_fops;
1255 extern const struct file_operations ceph_snapdir_fops;
1256 extern const struct inode_operations ceph_dir_iops;
1257 extern const struct inode_operations ceph_snapdir_iops;
1258 extern const struct dentry_operations ceph_dentry_ops;
1259 
1260 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1261 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1262 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1263 			       struct dentry *dentry);
1264 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1265 					 struct dentry *dentry, int err);
1266 
1267 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1268 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1269 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1270 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1271 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1272 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1273 
1274 /* ioctl.c */
1275 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1276 
1277 /* export.c */
1278 extern const struct export_operations ceph_export_ops;
1279 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1280 
1281 /* locks.c */
1282 extern __init void ceph_flock_init(void);
1283 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1284 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1285 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1286 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1287 				       struct ceph_filelock *flocks,
1288 				       int num_fcntl_locks,
1289 				       int num_flock_locks);
1290 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1291 				  struct ceph_pagelist *pagelist,
1292 				  int num_fcntl_locks, int num_flock_locks);
1293 
1294 /* debugfs.c */
1295 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1296 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1297 
1298 /* quota.c */
1299 
1300 enum quota_get_realm {
1301 	QUOTA_GET_MAX_FILES,
1302 	QUOTA_GET_MAX_BYTES,
1303 	QUOTA_GET_ANY
1304 };
1305 
1306 static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1307 				    enum quota_get_realm which)
1308 {
1309 	bool has_quota = false;
1310 
1311 	switch (which) {
1312 	case QUOTA_GET_MAX_BYTES:
1313 		has_quota = !!ci->i_max_bytes;
1314 		break;
1315 	case QUOTA_GET_MAX_FILES:
1316 		has_quota = !!ci->i_max_files;
1317 		break;
1318 	default:
1319 		has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1320 	}
1321 	return has_quota;
1322 }
1323 
1324 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1325 
1326 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1327 				       u64 max_bytes, u64 max_files)
1328 {
1329 	bool had_quota, has_quota;
1330 	had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1331 	ci->i_max_bytes = max_bytes;
1332 	ci->i_max_files = max_files;
1333 	has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1334 
1335 	if (had_quota != has_quota)
1336 		ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1337 }
1338 
1339 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1340 			      struct ceph_mds_session *session,
1341 			      struct ceph_msg *msg);
1342 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1343 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1344 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1345 					     loff_t newlen);
1346 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1347 						loff_t newlen);
1348 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1349 				     struct kstatfs *buf);
1350 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1351 
1352 #endif /* _FS_CEPH_SUPER_H */
1353