xref: /openbmc/linux/fs/ceph/super.h (revision 4a3fad70)
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 
20 #include <linux/ceph/libceph.h>
21 
22 #ifdef CONFIG_CEPH_FSCACHE
23 #include <linux/fscache.h>
24 #endif
25 
26 /* f_type in struct statfs */
27 #define CEPH_SUPER_MAGIC 0x00c36400
28 
29 /* large granularity for statfs utilization stats to facilitate
30  * large volume sizes on 32-bit machines. */
31 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
32 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
33 
34 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
35 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
36 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
37 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
38 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
39 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
40 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
41 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
42 
43 #define CEPH_MOUNT_OPT_DEFAULT    CEPH_MOUNT_OPT_DCACHE
44 
45 #define ceph_set_mount_opt(fsc, opt) \
46 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
47 #define ceph_test_mount_opt(fsc, opt) \
48 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
49 
50 /* max size of osd read request, limited by libceph */
51 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
52 /* osd has a configurable limitaion of max write size.
53  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
54 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
55 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
56 #define CEPH_MAX_READDIR_DEFAULT        1024
57 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
58 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
59 
60 /*
61  * Delay telling the MDS we no longer want caps, in case we reopen
62  * the file.  Delay a minimum amount of time, even if we send a cap
63  * message for some other reason.  Otherwise, take the oppotunity to
64  * update the mds to avoid sending another message later.
65  */
66 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
67 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
68 
69 struct ceph_mount_options {
70 	int flags;
71 	int sb_flags;
72 
73 	int wsize;            /* max write size */
74 	int rsize;            /* max read size */
75 	int rasize;           /* max readahead */
76 	int congestion_kb;    /* max writeback in flight */
77 	int caps_wanted_delay_min, caps_wanted_delay_max;
78 	int max_readdir;       /* max readdir result (entires) */
79 	int max_readdir_bytes; /* max readdir result (bytes) */
80 
81 	/*
82 	 * everything above this point can be memcmp'd; everything below
83 	 * is handled in compare_mount_options()
84 	 */
85 
86 	char *snapdir_name;   /* default ".snap" */
87 	char *mds_namespace;  /* default NULL */
88 	char *server_path;    /* default  "/" */
89 	char *fscache_uniq;   /* default NULL */
90 };
91 
92 struct ceph_fs_client {
93 	struct super_block *sb;
94 
95 	struct ceph_mount_options *mount_options;
96 	struct ceph_client *client;
97 
98 	unsigned long mount_state;
99 	int min_caps;                  /* min caps i added */
100 
101 	struct ceph_mds_client *mdsc;
102 
103 	/* writeback */
104 	mempool_t *wb_pagevec_pool;
105 	struct workqueue_struct *wb_wq;
106 	struct workqueue_struct *pg_inv_wq;
107 	struct workqueue_struct *trunc_wq;
108 	atomic_long_t writeback_count;
109 
110 #ifdef CONFIG_DEBUG_FS
111 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
112 	struct dentry *debugfs_congestion_kb;
113 	struct dentry *debugfs_bdi;
114 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
115 	struct dentry *debugfs_mds_sessions;
116 #endif
117 
118 #ifdef CONFIG_CEPH_FSCACHE
119 	struct fscache_cookie *fscache;
120 #endif
121 };
122 
123 
124 /*
125  * File i/o capability.  This tracks shared state with the metadata
126  * server that allows us to cache or writeback attributes or to read
127  * and write data.  For any given inode, we should have one or more
128  * capabilities, one issued by each metadata server, and our
129  * cumulative access is the OR of all issued capabilities.
130  *
131  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
132  * session capability lists.
133  */
134 struct ceph_cap {
135 	struct ceph_inode_info *ci;
136 	struct rb_node ci_node;          /* per-ci cap tree */
137 	struct ceph_mds_session *session;
138 	struct list_head session_caps;   /* per-session caplist */
139 	u64 cap_id;       /* unique cap id (mds provided) */
140 	union {
141 		/* in-use caps */
142 		struct {
143 			int issued;       /* latest, from the mds */
144 			int implemented;  /* implemented superset of
145 					     issued (for revocation) */
146 			int mds, mds_wanted;
147 		};
148 		/* caps to release */
149 		struct {
150 			u64 cap_ino;
151 			int queue_release;
152 		};
153 	};
154 	u32 seq, issue_seq, mseq;
155 	u32 cap_gen;      /* active/stale cycle */
156 	unsigned long last_used;
157 	struct list_head caps_item;
158 };
159 
160 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
161 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
162 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
163 
164 struct ceph_cap_flush {
165 	u64 tid;
166 	int caps; /* 0 means capsnap */
167 	bool wake; /* wake up flush waiters when finish ? */
168 	struct list_head g_list; // global
169 	struct list_head i_list; // per inode
170 };
171 
172 /*
173  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
174  * we first complete any in-process sync writes and writeback any dirty
175  * data before flushing the snapped state (tracked here) back to the MDS.
176  */
177 struct ceph_cap_snap {
178 	refcount_t nref;
179 	struct list_head ci_item;
180 
181 	struct ceph_cap_flush cap_flush;
182 
183 	u64 follows;
184 	int issued, dirty;
185 	struct ceph_snap_context *context;
186 
187 	umode_t mode;
188 	kuid_t uid;
189 	kgid_t gid;
190 
191 	struct ceph_buffer *xattr_blob;
192 	u64 xattr_version;
193 
194 	u64 size;
195 	struct timespec mtime, atime, ctime;
196 	u64 time_warp_seq;
197 	u64 truncate_size;
198 	u32 truncate_seq;
199 	int writing;   /* a sync write is still in progress */
200 	int dirty_pages;     /* dirty pages awaiting writeback */
201 	bool inline_data;
202 	bool need_flush;
203 };
204 
205 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
206 {
207 	if (refcount_dec_and_test(&capsnap->nref)) {
208 		if (capsnap->xattr_blob)
209 			ceph_buffer_put(capsnap->xattr_blob);
210 		kfree(capsnap);
211 	}
212 }
213 
214 /*
215  * The frag tree describes how a directory is fragmented, potentially across
216  * multiple metadata servers.  It is also used to indicate points where
217  * metadata authority is delegated, and whether/where metadata is replicated.
218  *
219  * A _leaf_ frag will be present in the i_fragtree IFF there is
220  * delegation info.  That is, if mds >= 0 || ndist > 0.
221  */
222 #define CEPH_MAX_DIRFRAG_REP 4
223 
224 struct ceph_inode_frag {
225 	struct rb_node node;
226 
227 	/* fragtree state */
228 	u32 frag;
229 	int split_by;         /* i.e. 2^(split_by) children */
230 
231 	/* delegation and replication info */
232 	int mds;              /* -1 if same authority as parent */
233 	int ndist;            /* >0 if replicated */
234 	int dist[CEPH_MAX_DIRFRAG_REP];
235 };
236 
237 /*
238  * We cache inode xattrs as an encoded blob until they are first used,
239  * at which point we parse them into an rbtree.
240  */
241 struct ceph_inode_xattr {
242 	struct rb_node node;
243 
244 	const char *name;
245 	int name_len;
246 	const char *val;
247 	int val_len;
248 	int dirty;
249 
250 	int should_free_name;
251 	int should_free_val;
252 };
253 
254 /*
255  * Ceph dentry state
256  */
257 struct ceph_dentry_info {
258 	struct ceph_mds_session *lease_session;
259 	u32 lease_gen, lease_shared_gen;
260 	u32 lease_seq;
261 	unsigned long lease_renew_after, lease_renew_from;
262 	struct list_head lru;
263 	struct dentry *dentry;
264 	unsigned long time;
265 	u64 offset;
266 };
267 
268 struct ceph_inode_xattrs_info {
269 	/*
270 	 * (still encoded) xattr blob. we avoid the overhead of parsing
271 	 * this until someone actually calls getxattr, etc.
272 	 *
273 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
274 	 * NULL means we don't know.
275 	*/
276 	struct ceph_buffer *blob, *prealloc_blob;
277 
278 	struct rb_root index;
279 	bool dirty;
280 	int count;
281 	int names_size;
282 	int vals_size;
283 	u64 version, index_version;
284 };
285 
286 /*
287  * Ceph inode.
288  */
289 struct ceph_inode_info {
290 	struct ceph_vino i_vino;   /* ceph ino + snap */
291 
292 	spinlock_t i_ceph_lock;
293 
294 	u64 i_version;
295 	u64 i_inline_version;
296 	u32 i_time_warp_seq;
297 
298 	unsigned i_ceph_flags;
299 	atomic64_t i_release_count;
300 	atomic64_t i_ordered_count;
301 	atomic64_t i_complete_seq[2];
302 
303 	struct ceph_dir_layout i_dir_layout;
304 	struct ceph_file_layout i_layout;
305 	char *i_symlink;
306 
307 	/* for dirs */
308 	struct timespec i_rctime;
309 	u64 i_rbytes, i_rfiles, i_rsubdirs;
310 	u64 i_files, i_subdirs;
311 
312 	struct rb_root i_fragtree;
313 	int i_fragtree_nsplits;
314 	struct mutex i_fragtree_mutex;
315 
316 	struct ceph_inode_xattrs_info i_xattrs;
317 
318 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
319 	 * s_mutex. */
320 	struct rb_root i_caps;           /* cap list */
321 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
322 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
323 	struct list_head i_dirty_item, i_flushing_item;
324 	/* we need to track cap writeback on a per-cap-bit basis, to allow
325 	 * overlapping, pipelined cap flushes to the mds.  we can probably
326 	 * reduce the tid to 8 bits if we're concerned about inode size. */
327 	struct ceph_cap_flush *i_prealloc_cap_flush;
328 	struct list_head i_cap_flush_list;
329 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
330 	unsigned long i_hold_caps_min; /* jiffies */
331 	unsigned long i_hold_caps_max; /* jiffies */
332 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
333 	struct ceph_cap_reservation i_cap_migration_resv;
334 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
335 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
336 						    dirty|flushing caps */
337 	unsigned i_snap_caps;           /* cap bits for snapped files */
338 
339 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
340 
341 	struct mutex i_truncate_mutex;
342 	u32 i_truncate_seq;        /* last truncate to smaller size */
343 	u64 i_truncate_size;       /*  and the size we last truncated down to */
344 	int i_truncate_pending;    /*  still need to call vmtruncate */
345 
346 	u64 i_max_size;            /* max file size authorized by mds */
347 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
348 	u64 i_wanted_max_size;     /* offset we'd like to write too */
349 	u64 i_requested_max_size;  /* max_size we've requested */
350 
351 	/* held references to caps */
352 	int i_pin_ref;
353 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
354 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
355 	atomic_t i_filelock_ref;
356 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
357 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
358 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
359 
360 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
361 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
362 	spinlock_t i_unsafe_lock;
363 
364 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
365 	int i_snap_realm_counter; /* snap realm (if caps) */
366 	struct list_head i_snap_realm_item;
367 	struct list_head i_snap_flush_item;
368 
369 	struct work_struct i_wb_work;  /* writeback work */
370 	struct work_struct i_pg_inv_work;  /* page invalidation work */
371 
372 	struct work_struct i_vmtruncate_work;
373 
374 #ifdef CONFIG_CEPH_FSCACHE
375 	struct fscache_cookie *fscache;
376 	u32 i_fscache_gen;
377 #endif
378 	struct inode vfs_inode; /* at end */
379 };
380 
381 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
382 {
383 	return container_of(inode, struct ceph_inode_info, vfs_inode);
384 }
385 
386 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
387 {
388 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
389 }
390 
391 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
392 {
393 	return (struct ceph_fs_client *)sb->s_fs_info;
394 }
395 
396 static inline struct ceph_vino ceph_vino(struct inode *inode)
397 {
398 	return ceph_inode(inode)->i_vino;
399 }
400 
401 /*
402  * ino_t is <64 bits on many architectures, blech.
403  *
404  *               i_ino (kernel inode)   st_ino (userspace)
405  * i386          32                     32
406  * x86_64+ino32  64                     32
407  * x86_64        64                     64
408  */
409 static inline u32 ceph_ino_to_ino32(__u64 vino)
410 {
411 	u32 ino = vino & 0xffffffff;
412 	ino ^= vino >> 32;
413 	if (!ino)
414 		ino = 2;
415 	return ino;
416 }
417 
418 /*
419  * kernel i_ino value
420  */
421 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
422 {
423 #if BITS_PER_LONG == 32
424 	return ceph_ino_to_ino32(vino.ino);
425 #else
426 	return (ino_t)vino.ino;
427 #endif
428 }
429 
430 /*
431  * user-visible ino (stat, filldir)
432  */
433 #if BITS_PER_LONG == 32
434 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
435 {
436 	return ino;
437 }
438 #else
439 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
440 {
441 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
442 		ino = ceph_ino_to_ino32(ino);
443 	return ino;
444 }
445 #endif
446 
447 
448 /* for printf-style formatting */
449 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
450 
451 static inline u64 ceph_ino(struct inode *inode)
452 {
453 	return ceph_inode(inode)->i_vino.ino;
454 }
455 static inline u64 ceph_snap(struct inode *inode)
456 {
457 	return ceph_inode(inode)->i_vino.snap;
458 }
459 
460 static inline int ceph_ino_compare(struct inode *inode, void *data)
461 {
462 	struct ceph_vino *pvino = (struct ceph_vino *)data;
463 	struct ceph_inode_info *ci = ceph_inode(inode);
464 	return ci->i_vino.ino == pvino->ino &&
465 		ci->i_vino.snap == pvino->snap;
466 }
467 
468 static inline struct inode *ceph_find_inode(struct super_block *sb,
469 					    struct ceph_vino vino)
470 {
471 	ino_t t = ceph_vino_to_ino(vino);
472 	return ilookup5(sb, t, ceph_ino_compare, &vino);
473 }
474 
475 
476 /*
477  * Ceph inode.
478  */
479 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
480 #define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
481 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
482 #define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
483 #define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
484 #define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
485 #define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
486 #define CEPH_I_SEC_INITED	(1 << 7)  /* security initialized */
487 #define CEPH_I_CAP_DROPPED	(1 << 8)  /* caps were forcibly dropped */
488 #define CEPH_I_KICK_FLUSH	(1 << 9)  /* kick flushing caps */
489 #define CEPH_I_FLUSH_SNAPS	(1 << 10) /* need flush snapss */
490 #define CEPH_I_ERROR_WRITE	(1 << 11) /* have seen write errors */
491 #define CEPH_I_ERROR_FILELOCK	(1 << 12) /* have seen file lock errors */
492 
493 
494 /*
495  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
496  * and then clear it when they start succeeding. Note that we do a lockless
497  * check first, and only take the lock if it looks like it needs to be changed.
498  * The write submission code just takes this as a hint, so we're not too
499  * worried if a few slip through in either direction.
500  */
501 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
502 {
503 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
504 		spin_lock(&ci->i_ceph_lock);
505 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
506 		spin_unlock(&ci->i_ceph_lock);
507 	}
508 }
509 
510 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
511 {
512 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
513 		spin_lock(&ci->i_ceph_lock);
514 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
515 		spin_unlock(&ci->i_ceph_lock);
516 	}
517 }
518 
519 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
520 					   long long release_count,
521 					   long long ordered_count)
522 {
523 	smp_mb__before_atomic();
524 	atomic64_set(&ci->i_complete_seq[0], release_count);
525 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
526 }
527 
528 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
529 {
530 	atomic64_inc(&ci->i_release_count);
531 }
532 
533 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
534 {
535 	atomic64_inc(&ci->i_ordered_count);
536 }
537 
538 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
539 {
540 	return atomic64_read(&ci->i_complete_seq[0]) ==
541 		atomic64_read(&ci->i_release_count);
542 }
543 
544 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
545 {
546 	return  atomic64_read(&ci->i_complete_seq[0]) ==
547 		atomic64_read(&ci->i_release_count) &&
548 		atomic64_read(&ci->i_complete_seq[1]) ==
549 		atomic64_read(&ci->i_ordered_count);
550 }
551 
552 static inline void ceph_dir_clear_complete(struct inode *inode)
553 {
554 	__ceph_dir_clear_complete(ceph_inode(inode));
555 }
556 
557 static inline void ceph_dir_clear_ordered(struct inode *inode)
558 {
559 	__ceph_dir_clear_ordered(ceph_inode(inode));
560 }
561 
562 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
563 {
564 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
565 	smp_rmb();
566 	return ret;
567 }
568 
569 /* find a specific frag @f */
570 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
571 						u32 f);
572 
573 /*
574  * choose fragment for value @v.  copy frag content to pfrag, if leaf
575  * exists
576  */
577 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
578 			    struct ceph_inode_frag *pfrag,
579 			    int *found);
580 
581 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
582 {
583 	return (struct ceph_dentry_info *)dentry->d_fsdata;
584 }
585 
586 /*
587  * caps helpers
588  */
589 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
590 {
591 	return !RB_EMPTY_ROOT(&ci->i_caps);
592 }
593 
594 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
595 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
596 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
597 				    struct ceph_cap *cap);
598 
599 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
600 {
601 	int issued;
602 	spin_lock(&ci->i_ceph_lock);
603 	issued = __ceph_caps_issued(ci, NULL);
604 	spin_unlock(&ci->i_ceph_lock);
605 	return issued;
606 }
607 
608 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
609 					int touch)
610 {
611 	int r;
612 	spin_lock(&ci->i_ceph_lock);
613 	r = __ceph_caps_issued_mask(ci, mask, touch);
614 	spin_unlock(&ci->i_ceph_lock);
615 	return r;
616 }
617 
618 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
619 {
620 	return ci->i_dirty_caps | ci->i_flushing_caps;
621 }
622 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
623 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
624 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
625 				  struct ceph_cap_flush **pcf);
626 
627 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
628 				      struct ceph_cap *ocap, int mask);
629 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
630 extern int __ceph_caps_used(struct ceph_inode_info *ci);
631 
632 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
633 
634 /*
635  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
636  */
637 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
638 {
639 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
640 	if (w & CEPH_CAP_FILE_BUFFER)
641 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
642 	return w;
643 }
644 
645 /* what the mds thinks we want */
646 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
647 
648 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
649 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
650 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
651 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
652 			     struct ceph_cap_reservation *ctx, int need);
653 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
654 			       struct ceph_cap_reservation *ctx);
655 extern void ceph_reservation_status(struct ceph_fs_client *client,
656 				    int *total, int *avail, int *used,
657 				    int *reserved, int *min);
658 
659 
660 
661 /*
662  * we keep buffered readdir results attached to file->private_data
663  */
664 #define CEPH_F_SYNC     1
665 #define CEPH_F_ATEND    2
666 
667 struct ceph_file_info {
668 	short fmode;     /* initialized on open */
669 	short flags;     /* CEPH_F_* */
670 
671 	/* readdir: position within the dir */
672 	u32 frag;
673 	struct ceph_mds_request *last_readdir;
674 
675 	/* readdir: position within a frag */
676 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
677 	char *last_name;       /* last entry in previous chunk */
678 	long long dir_release_count;
679 	long long dir_ordered_count;
680 	int readdir_cache_idx;
681 
682 	/* used for -o dirstat read() on directory thing */
683 	char *dir_info;
684 	int dir_info_len;
685 };
686 
687 struct ceph_readdir_cache_control {
688 	struct page  *page;
689 	struct dentry **dentries;
690 	int index;
691 };
692 
693 /*
694  * A "snap realm" describes a subset of the file hierarchy sharing
695  * the same set of snapshots that apply to it.  The realms themselves
696  * are organized into a hierarchy, such that children inherit (some of)
697  * the snapshots of their parents.
698  *
699  * All inodes within the realm that have capabilities are linked into a
700  * per-realm list.
701  */
702 struct ceph_snap_realm {
703 	u64 ino;
704 	atomic_t nref;
705 	struct rb_node node;
706 
707 	u64 created, seq;
708 	u64 parent_ino;
709 	u64 parent_since;   /* snapid when our current parent became so */
710 
711 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
712 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
713 	u64 *snaps;                   /* snaps specific to this realm */
714 	u32 num_snaps;
715 
716 	struct ceph_snap_realm *parent;
717 	struct list_head children;       /* list of child realms */
718 	struct list_head child_item;
719 
720 	struct list_head empty_item;     /* if i have ref==0 */
721 
722 	struct list_head dirty_item;     /* if realm needs new context */
723 
724 	/* the current set of snaps for this realm */
725 	struct ceph_snap_context *cached_context;
726 
727 	struct list_head inodes_with_caps;
728 	spinlock_t inodes_with_caps_lock;
729 };
730 
731 static inline int default_congestion_kb(void)
732 {
733 	int congestion_kb;
734 
735 	/*
736 	 * Copied from NFS
737 	 *
738 	 * congestion size, scale with available memory.
739 	 *
740 	 *  64MB:    8192k
741 	 * 128MB:   11585k
742 	 * 256MB:   16384k
743 	 * 512MB:   23170k
744 	 *   1GB:   32768k
745 	 *   2GB:   46340k
746 	 *   4GB:   65536k
747 	 *   8GB:   92681k
748 	 *  16GB:  131072k
749 	 *
750 	 * This allows larger machines to have larger/more transfers.
751 	 * Limit the default to 256M
752 	 */
753 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
754 	if (congestion_kb > 256*1024)
755 		congestion_kb = 256*1024;
756 
757 	return congestion_kb;
758 }
759 
760 
761 
762 /* snap.c */
763 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
764 					       u64 ino);
765 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
766 				struct ceph_snap_realm *realm);
767 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
768 				struct ceph_snap_realm *realm);
769 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
770 				  void *p, void *e, bool deletion,
771 				  struct ceph_snap_realm **realm_ret);
772 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
773 			     struct ceph_mds_session *session,
774 			     struct ceph_msg *msg);
775 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
776 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
777 				  struct ceph_cap_snap *capsnap);
778 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
779 
780 /*
781  * a cap_snap is "pending" if it is still awaiting an in-progress
782  * sync write (that may/may not still update size, mtime, etc.).
783  */
784 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
785 {
786 	return !list_empty(&ci->i_cap_snaps) &&
787 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
788 			       ci_item)->writing;
789 }
790 
791 /* inode.c */
792 extern const struct inode_operations ceph_file_iops;
793 
794 extern struct inode *ceph_alloc_inode(struct super_block *sb);
795 extern void ceph_destroy_inode(struct inode *inode);
796 extern int ceph_drop_inode(struct inode *inode);
797 
798 extern struct inode *ceph_get_inode(struct super_block *sb,
799 				    struct ceph_vino vino);
800 extern struct inode *ceph_get_snapdir(struct inode *parent);
801 extern int ceph_fill_file_size(struct inode *inode, int issued,
802 			       u32 truncate_seq, u64 truncate_size, u64 size);
803 extern void ceph_fill_file_time(struct inode *inode, int issued,
804 				u64 time_warp_seq, struct timespec *ctime,
805 				struct timespec *mtime, struct timespec *atime);
806 extern int ceph_fill_trace(struct super_block *sb,
807 			   struct ceph_mds_request *req);
808 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
809 				    struct ceph_mds_session *session);
810 
811 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
812 
813 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
814 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
815 extern void ceph_queue_vmtruncate(struct inode *inode);
816 
817 extern void ceph_queue_invalidate(struct inode *inode);
818 extern void ceph_queue_writeback(struct inode *inode);
819 
820 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
821 			     int mask, bool force);
822 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
823 {
824 	return __ceph_do_getattr(inode, NULL, mask, force);
825 }
826 extern int ceph_permission(struct inode *inode, int mask);
827 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
828 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
829 extern int ceph_getattr(const struct path *path, struct kstat *stat,
830 			u32 request_mask, unsigned int flags);
831 
832 /* xattr.c */
833 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
834 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
835 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
836 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
837 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
838 extern void __init ceph_xattr_init(void);
839 extern void ceph_xattr_exit(void);
840 extern const struct xattr_handler *ceph_xattr_handlers[];
841 
842 #ifdef CONFIG_SECURITY
843 extern bool ceph_security_xattr_deadlock(struct inode *in);
844 extern bool ceph_security_xattr_wanted(struct inode *in);
845 #else
846 static inline bool ceph_security_xattr_deadlock(struct inode *in)
847 {
848 	return false;
849 }
850 static inline bool ceph_security_xattr_wanted(struct inode *in)
851 {
852 	return false;
853 }
854 #endif
855 
856 /* acl.c */
857 struct ceph_acls_info {
858 	void *default_acl;
859 	void *acl;
860 	struct ceph_pagelist *pagelist;
861 };
862 
863 #ifdef CONFIG_CEPH_FS_POSIX_ACL
864 
865 struct posix_acl *ceph_get_acl(struct inode *, int);
866 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
867 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
868 		       struct ceph_acls_info *info);
869 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
870 void ceph_release_acls_info(struct ceph_acls_info *info);
871 
872 static inline void ceph_forget_all_cached_acls(struct inode *inode)
873 {
874        forget_all_cached_acls(inode);
875 }
876 
877 #else
878 
879 #define ceph_get_acl NULL
880 #define ceph_set_acl NULL
881 
882 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
883 				     struct ceph_acls_info *info)
884 {
885 	return 0;
886 }
887 static inline void ceph_init_inode_acls(struct inode *inode,
888 					struct ceph_acls_info *info)
889 {
890 }
891 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
892 {
893 }
894 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
895 {
896 	return 0;
897 }
898 
899 static inline void ceph_forget_all_cached_acls(struct inode *inode)
900 {
901 }
902 
903 #endif
904 
905 /* caps.c */
906 extern const char *ceph_cap_string(int c);
907 extern void ceph_handle_caps(struct ceph_mds_session *session,
908 			     struct ceph_msg *msg);
909 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
910 				     struct ceph_cap_reservation *ctx);
911 extern void ceph_add_cap(struct inode *inode,
912 			 struct ceph_mds_session *session, u64 cap_id,
913 			 int fmode, unsigned issued, unsigned wanted,
914 			 unsigned cap, unsigned seq, u64 realmino, int flags,
915 			 struct ceph_cap **new_cap);
916 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
917 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
918 			 struct ceph_cap *cap);
919 extern int ceph_is_any_caps(struct inode *inode);
920 
921 extern void ceph_queue_caps_release(struct inode *inode);
922 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
923 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
924 		      int datasync);
925 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
926 					  struct ceph_mds_session *session);
927 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
928 				    struct ceph_mds_session *session);
929 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
930 					     int mds);
931 extern int ceph_get_cap_mds(struct inode *inode);
932 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
933 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
934 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
935 				       struct ceph_snap_context *snapc);
936 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
937 			     struct ceph_mds_session **psession);
938 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
939 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
940 			    struct ceph_mds_session *session);
941 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
942 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
943 
944 extern int ceph_encode_inode_release(void **p, struct inode *inode,
945 				     int mds, int drop, int unless, int force);
946 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
947 				      struct inode *dir,
948 				      int mds, int drop, int unless);
949 
950 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
951 			 loff_t endoff, int *got, struct page **pinned_page);
952 extern int ceph_try_get_caps(struct ceph_inode_info *ci,
953 			     int need, int want, int *got);
954 
955 /* for counting open files by mode */
956 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode);
957 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
958 
959 /* addr.c */
960 extern const struct address_space_operations ceph_aops;
961 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
962 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
963 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
964 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
965 
966 /* file.c */
967 extern const struct file_operations ceph_file_fops;
968 
969 extern int ceph_renew_caps(struct inode *inode);
970 extern int ceph_open(struct inode *inode, struct file *file);
971 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
972 			    struct file *file, unsigned flags, umode_t mode,
973 			    int *opened);
974 extern int ceph_release(struct inode *inode, struct file *filp);
975 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
976 				  char *data, size_t len);
977 
978 /* dir.c */
979 extern const struct file_operations ceph_dir_fops;
980 extern const struct file_operations ceph_snapdir_fops;
981 extern const struct inode_operations ceph_dir_iops;
982 extern const struct inode_operations ceph_snapdir_iops;
983 extern const struct dentry_operations ceph_dentry_ops;
984 
985 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
986 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
987 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
988 			       struct dentry *dentry, int err);
989 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
990 					 struct dentry *dentry, int err);
991 
992 extern void ceph_dentry_lru_add(struct dentry *dn);
993 extern void ceph_dentry_lru_touch(struct dentry *dn);
994 extern void ceph_dentry_lru_del(struct dentry *dn);
995 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
996 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
997 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
998 
999 /* ioctl.c */
1000 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1001 
1002 /* export.c */
1003 extern const struct export_operations ceph_export_ops;
1004 
1005 /* locks.c */
1006 extern __init void ceph_flock_init(void);
1007 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1008 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1009 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1010 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1011 				       struct ceph_filelock *flocks,
1012 				       int num_fcntl_locks,
1013 				       int num_flock_locks);
1014 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1015 				  struct ceph_pagelist *pagelist,
1016 				  int num_fcntl_locks, int num_flock_locks);
1017 
1018 /* debugfs.c */
1019 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
1020 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1021 
1022 #endif /* _FS_CEPH_SUPER_H */
1023