xref: /openbmc/linux/fs/ceph/super.h (revision b78412b8)
1 #ifndef _FS_CEPH_SUPER_H
2 #define _FS_CEPH_SUPER_H
3 
4 #include <linux/ceph/ceph_debug.h>
5 
6 #include <asm/unaligned.h>
7 #include <linux/backing-dev.h>
8 #include <linux/completion.h>
9 #include <linux/exportfs.h>
10 #include <linux/fs.h>
11 #include <linux/mempool.h>
12 #include <linux/pagemap.h>
13 #include <linux/wait.h>
14 #include <linux/writeback.h>
15 #include <linux/slab.h>
16 #include <linux/posix_acl.h>
17 #include <linux/refcount.h>
18 
19 #include <linux/ceph/libceph.h>
20 
21 #ifdef CONFIG_CEPH_FSCACHE
22 #include <linux/fscache.h>
23 #endif
24 
25 /* f_type in struct statfs */
26 #define CEPH_SUPER_MAGIC 0x00c36400
27 
28 /* large granularity for statfs utilization stats to facilitate
29  * large volume sizes on 32-bit machines. */
30 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
31 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
32 
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 
42 #define CEPH_MOUNT_OPT_DEFAULT    CEPH_MOUNT_OPT_DCACHE
43 
44 #define ceph_set_mount_opt(fsc, opt) \
45 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
46 #define ceph_test_mount_opt(fsc, opt) \
47 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
48 
49 /* max size of osd read request, limited by libceph */
50 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
51 /* osd has a configurable limitaion of max write size.
52  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
53 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
54 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
55 #define CEPH_MAX_READDIR_DEFAULT        1024
56 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
57 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
58 
59 /*
60  * Delay telling the MDS we no longer want caps, in case we reopen
61  * the file.  Delay a minimum amount of time, even if we send a cap
62  * message for some other reason.  Otherwise, take the oppotunity to
63  * update the mds to avoid sending another message later.
64  */
65 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
66 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
67 
68 struct ceph_mount_options {
69 	int flags;
70 	int sb_flags;
71 
72 	int wsize;            /* max write size */
73 	int rsize;            /* max read size */
74 	int rasize;           /* max readahead */
75 	int congestion_kb;    /* max writeback in flight */
76 	int caps_wanted_delay_min, caps_wanted_delay_max;
77 	int max_readdir;       /* max readdir result (entires) */
78 	int max_readdir_bytes; /* max readdir result (bytes) */
79 
80 	/*
81 	 * everything above this point can be memcmp'd; everything below
82 	 * is handled in compare_mount_options()
83 	 */
84 
85 	char *snapdir_name;   /* default ".snap" */
86 	char *mds_namespace;  /* default NULL */
87 	char *server_path;    /* default  "/" */
88 	char *fscache_uniq;   /* default NULL */
89 };
90 
91 struct ceph_fs_client {
92 	struct super_block *sb;
93 
94 	struct ceph_mount_options *mount_options;
95 	struct ceph_client *client;
96 
97 	unsigned long mount_state;
98 	int min_caps;                  /* min caps i added */
99 
100 	struct ceph_mds_client *mdsc;
101 
102 	/* writeback */
103 	mempool_t *wb_pagevec_pool;
104 	struct workqueue_struct *wb_wq;
105 	struct workqueue_struct *pg_inv_wq;
106 	struct workqueue_struct *trunc_wq;
107 	atomic_long_t writeback_count;
108 
109 #ifdef CONFIG_DEBUG_FS
110 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
111 	struct dentry *debugfs_congestion_kb;
112 	struct dentry *debugfs_bdi;
113 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
114 	struct dentry *debugfs_mds_sessions;
115 #endif
116 
117 #ifdef CONFIG_CEPH_FSCACHE
118 	struct fscache_cookie *fscache;
119 #endif
120 };
121 
122 
123 /*
124  * File i/o capability.  This tracks shared state with the metadata
125  * server that allows us to cache or writeback attributes or to read
126  * and write data.  For any given inode, we should have one or more
127  * capabilities, one issued by each metadata server, and our
128  * cumulative access is the OR of all issued capabilities.
129  *
130  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
131  * session capability lists.
132  */
133 struct ceph_cap {
134 	struct ceph_inode_info *ci;
135 	struct rb_node ci_node;          /* per-ci cap tree */
136 	struct ceph_mds_session *session;
137 	struct list_head session_caps;   /* per-session caplist */
138 	u64 cap_id;       /* unique cap id (mds provided) */
139 	union {
140 		/* in-use caps */
141 		struct {
142 			int issued;       /* latest, from the mds */
143 			int implemented;  /* implemented superset of
144 					     issued (for revocation) */
145 			int mds, mds_wanted;
146 		};
147 		/* caps to release */
148 		struct {
149 			u64 cap_ino;
150 			int queue_release;
151 		};
152 	};
153 	u32 seq, issue_seq, mseq;
154 	u32 cap_gen;      /* active/stale cycle */
155 	unsigned long last_used;
156 	struct list_head caps_item;
157 };
158 
159 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
160 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
161 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
162 
163 struct ceph_cap_flush {
164 	u64 tid;
165 	int caps; /* 0 means capsnap */
166 	bool wake; /* wake up flush waiters when finish ? */
167 	struct list_head g_list; // global
168 	struct list_head i_list; // per inode
169 };
170 
171 /*
172  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
173  * we first complete any in-process sync writes and writeback any dirty
174  * data before flushing the snapped state (tracked here) back to the MDS.
175  */
176 struct ceph_cap_snap {
177 	refcount_t nref;
178 	struct list_head ci_item;
179 
180 	struct ceph_cap_flush cap_flush;
181 
182 	u64 follows;
183 	int issued, dirty;
184 	struct ceph_snap_context *context;
185 
186 	umode_t mode;
187 	kuid_t uid;
188 	kgid_t gid;
189 
190 	struct ceph_buffer *xattr_blob;
191 	u64 xattr_version;
192 
193 	u64 size;
194 	struct timespec mtime, atime, ctime;
195 	u64 time_warp_seq;
196 	u64 truncate_size;
197 	u32 truncate_seq;
198 	int writing;   /* a sync write is still in progress */
199 	int dirty_pages;     /* dirty pages awaiting writeback */
200 	bool inline_data;
201 	bool need_flush;
202 };
203 
204 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
205 {
206 	if (refcount_dec_and_test(&capsnap->nref)) {
207 		if (capsnap->xattr_blob)
208 			ceph_buffer_put(capsnap->xattr_blob);
209 		kfree(capsnap);
210 	}
211 }
212 
213 /*
214  * The frag tree describes how a directory is fragmented, potentially across
215  * multiple metadata servers.  It is also used to indicate points where
216  * metadata authority is delegated, and whether/where metadata is replicated.
217  *
218  * A _leaf_ frag will be present in the i_fragtree IFF there is
219  * delegation info.  That is, if mds >= 0 || ndist > 0.
220  */
221 #define CEPH_MAX_DIRFRAG_REP 4
222 
223 struct ceph_inode_frag {
224 	struct rb_node node;
225 
226 	/* fragtree state */
227 	u32 frag;
228 	int split_by;         /* i.e. 2^(split_by) children */
229 
230 	/* delegation and replication info */
231 	int mds;              /* -1 if same authority as parent */
232 	int ndist;            /* >0 if replicated */
233 	int dist[CEPH_MAX_DIRFRAG_REP];
234 };
235 
236 /*
237  * We cache inode xattrs as an encoded blob until they are first used,
238  * at which point we parse them into an rbtree.
239  */
240 struct ceph_inode_xattr {
241 	struct rb_node node;
242 
243 	const char *name;
244 	int name_len;
245 	const char *val;
246 	int val_len;
247 	int dirty;
248 
249 	int should_free_name;
250 	int should_free_val;
251 };
252 
253 /*
254  * Ceph dentry state
255  */
256 struct ceph_dentry_info {
257 	struct ceph_mds_session *lease_session;
258 	u32 lease_gen, lease_shared_gen;
259 	u32 lease_seq;
260 	unsigned long lease_renew_after, lease_renew_from;
261 	struct list_head lru;
262 	struct dentry *dentry;
263 	unsigned long time;
264 	u64 offset;
265 };
266 
267 struct ceph_inode_xattrs_info {
268 	/*
269 	 * (still encoded) xattr blob. we avoid the overhead of parsing
270 	 * this until someone actually calls getxattr, etc.
271 	 *
272 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
273 	 * NULL means we don't know.
274 	*/
275 	struct ceph_buffer *blob, *prealloc_blob;
276 
277 	struct rb_root index;
278 	bool dirty;
279 	int count;
280 	int names_size;
281 	int vals_size;
282 	u64 version, index_version;
283 };
284 
285 /*
286  * Ceph inode.
287  */
288 struct ceph_inode_info {
289 	struct ceph_vino i_vino;   /* ceph ino + snap */
290 
291 	spinlock_t i_ceph_lock;
292 
293 	u64 i_version;
294 	u64 i_inline_version;
295 	u32 i_time_warp_seq;
296 
297 	unsigned i_ceph_flags;
298 	atomic64_t i_release_count;
299 	atomic64_t i_ordered_count;
300 	atomic64_t i_complete_seq[2];
301 
302 	struct ceph_dir_layout i_dir_layout;
303 	struct ceph_file_layout i_layout;
304 	char *i_symlink;
305 
306 	/* for dirs */
307 	struct timespec i_rctime;
308 	u64 i_rbytes, i_rfiles, i_rsubdirs;
309 	u64 i_files, i_subdirs;
310 
311 	struct rb_root i_fragtree;
312 	int i_fragtree_nsplits;
313 	struct mutex i_fragtree_mutex;
314 
315 	struct ceph_inode_xattrs_info i_xattrs;
316 
317 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
318 	 * s_mutex. */
319 	struct rb_root i_caps;           /* cap list */
320 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
321 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
322 	struct list_head i_dirty_item, i_flushing_item;
323 	/* we need to track cap writeback on a per-cap-bit basis, to allow
324 	 * overlapping, pipelined cap flushes to the mds.  we can probably
325 	 * reduce the tid to 8 bits if we're concerned about inode size. */
326 	struct ceph_cap_flush *i_prealloc_cap_flush;
327 	struct list_head i_cap_flush_list;
328 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
329 	unsigned long i_hold_caps_min; /* jiffies */
330 	unsigned long i_hold_caps_max; /* jiffies */
331 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
332 	struct ceph_cap_reservation i_cap_migration_resv;
333 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
334 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
335 						    dirty|flushing caps */
336 	unsigned i_snap_caps;           /* cap bits for snapped files */
337 
338 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
339 
340 	struct mutex i_truncate_mutex;
341 	u32 i_truncate_seq;        /* last truncate to smaller size */
342 	u64 i_truncate_size;       /*  and the size we last truncated down to */
343 	int i_truncate_pending;    /*  still need to call vmtruncate */
344 
345 	u64 i_max_size;            /* max file size authorized by mds */
346 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
347 	u64 i_wanted_max_size;     /* offset we'd like to write too */
348 	u64 i_requested_max_size;  /* max_size we've requested */
349 
350 	/* held references to caps */
351 	int i_pin_ref;
352 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
353 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
354 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
355 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
356 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
357 
358 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
359 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
360 	spinlock_t i_unsafe_lock;
361 
362 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
363 	int i_snap_realm_counter; /* snap realm (if caps) */
364 	struct list_head i_snap_realm_item;
365 	struct list_head i_snap_flush_item;
366 
367 	struct work_struct i_wb_work;  /* writeback work */
368 	struct work_struct i_pg_inv_work;  /* page invalidation work */
369 
370 	struct work_struct i_vmtruncate_work;
371 
372 #ifdef CONFIG_CEPH_FSCACHE
373 	struct fscache_cookie *fscache;
374 	u32 i_fscache_gen;
375 #endif
376 	struct inode vfs_inode; /* at end */
377 };
378 
379 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
380 {
381 	return container_of(inode, struct ceph_inode_info, vfs_inode);
382 }
383 
384 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
385 {
386 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
387 }
388 
389 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
390 {
391 	return (struct ceph_fs_client *)sb->s_fs_info;
392 }
393 
394 static inline struct ceph_vino ceph_vino(struct inode *inode)
395 {
396 	return ceph_inode(inode)->i_vino;
397 }
398 
399 /*
400  * ino_t is <64 bits on many architectures, blech.
401  *
402  *               i_ino (kernel inode)   st_ino (userspace)
403  * i386          32                     32
404  * x86_64+ino32  64                     32
405  * x86_64        64                     64
406  */
407 static inline u32 ceph_ino_to_ino32(__u64 vino)
408 {
409 	u32 ino = vino & 0xffffffff;
410 	ino ^= vino >> 32;
411 	if (!ino)
412 		ino = 2;
413 	return ino;
414 }
415 
416 /*
417  * kernel i_ino value
418  */
419 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
420 {
421 #if BITS_PER_LONG == 32
422 	return ceph_ino_to_ino32(vino.ino);
423 #else
424 	return (ino_t)vino.ino;
425 #endif
426 }
427 
428 /*
429  * user-visible ino (stat, filldir)
430  */
431 #if BITS_PER_LONG == 32
432 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
433 {
434 	return ino;
435 }
436 #else
437 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
438 {
439 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
440 		ino = ceph_ino_to_ino32(ino);
441 	return ino;
442 }
443 #endif
444 
445 
446 /* for printf-style formatting */
447 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
448 
449 static inline u64 ceph_ino(struct inode *inode)
450 {
451 	return ceph_inode(inode)->i_vino.ino;
452 }
453 static inline u64 ceph_snap(struct inode *inode)
454 {
455 	return ceph_inode(inode)->i_vino.snap;
456 }
457 
458 static inline int ceph_ino_compare(struct inode *inode, void *data)
459 {
460 	struct ceph_vino *pvino = (struct ceph_vino *)data;
461 	struct ceph_inode_info *ci = ceph_inode(inode);
462 	return ci->i_vino.ino == pvino->ino &&
463 		ci->i_vino.snap == pvino->snap;
464 }
465 
466 static inline struct inode *ceph_find_inode(struct super_block *sb,
467 					    struct ceph_vino vino)
468 {
469 	ino_t t = ceph_vino_to_ino(vino);
470 	return ilookup5(sb, t, ceph_ino_compare, &vino);
471 }
472 
473 
474 /*
475  * Ceph inode.
476  */
477 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
478 #define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
479 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
480 #define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
481 #define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
482 #define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
483 #define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
484 #define CEPH_I_SEC_INITED	(1 << 7)  /* security initialized */
485 #define CEPH_I_CAP_DROPPED	(1 << 8)  /* caps were forcibly dropped */
486 #define CEPH_I_KICK_FLUSH	(1 << 9)  /* kick flushing caps */
487 #define CEPH_I_FLUSH_SNAPS	(1 << 10) /* need flush snapss */
488 #define CEPH_I_ERROR_WRITE	(1 << 11) /* have seen write errors */
489 
490 /*
491  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
492  * and then clear it when they start succeeding. Note that we do a lockless
493  * check first, and only take the lock if it looks like it needs to be changed.
494  * The write submission code just takes this as a hint, so we're not too
495  * worried if a few slip through in either direction.
496  */
497 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
498 {
499 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
500 		spin_lock(&ci->i_ceph_lock);
501 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
502 		spin_unlock(&ci->i_ceph_lock);
503 	}
504 }
505 
506 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
507 {
508 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
509 		spin_lock(&ci->i_ceph_lock);
510 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
511 		spin_unlock(&ci->i_ceph_lock);
512 	}
513 }
514 
515 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
516 					   long long release_count,
517 					   long long ordered_count)
518 {
519 	smp_mb__before_atomic();
520 	atomic64_set(&ci->i_complete_seq[0], release_count);
521 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
522 }
523 
524 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
525 {
526 	atomic64_inc(&ci->i_release_count);
527 }
528 
529 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
530 {
531 	atomic64_inc(&ci->i_ordered_count);
532 }
533 
534 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
535 {
536 	return atomic64_read(&ci->i_complete_seq[0]) ==
537 		atomic64_read(&ci->i_release_count);
538 }
539 
540 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
541 {
542 	return  atomic64_read(&ci->i_complete_seq[0]) ==
543 		atomic64_read(&ci->i_release_count) &&
544 		atomic64_read(&ci->i_complete_seq[1]) ==
545 		atomic64_read(&ci->i_ordered_count);
546 }
547 
548 static inline void ceph_dir_clear_complete(struct inode *inode)
549 {
550 	__ceph_dir_clear_complete(ceph_inode(inode));
551 }
552 
553 static inline void ceph_dir_clear_ordered(struct inode *inode)
554 {
555 	__ceph_dir_clear_ordered(ceph_inode(inode));
556 }
557 
558 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
559 {
560 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
561 	smp_rmb();
562 	return ret;
563 }
564 
565 /* find a specific frag @f */
566 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
567 						u32 f);
568 
569 /*
570  * choose fragment for value @v.  copy frag content to pfrag, if leaf
571  * exists
572  */
573 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
574 			    struct ceph_inode_frag *pfrag,
575 			    int *found);
576 
577 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
578 {
579 	return (struct ceph_dentry_info *)dentry->d_fsdata;
580 }
581 
582 /*
583  * caps helpers
584  */
585 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
586 {
587 	return !RB_EMPTY_ROOT(&ci->i_caps);
588 }
589 
590 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
591 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
592 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
593 				    struct ceph_cap *cap);
594 
595 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
596 {
597 	int issued;
598 	spin_lock(&ci->i_ceph_lock);
599 	issued = __ceph_caps_issued(ci, NULL);
600 	spin_unlock(&ci->i_ceph_lock);
601 	return issued;
602 }
603 
604 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
605 					int touch)
606 {
607 	int r;
608 	spin_lock(&ci->i_ceph_lock);
609 	r = __ceph_caps_issued_mask(ci, mask, touch);
610 	spin_unlock(&ci->i_ceph_lock);
611 	return r;
612 }
613 
614 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
615 {
616 	return ci->i_dirty_caps | ci->i_flushing_caps;
617 }
618 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
619 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
620 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
621 				  struct ceph_cap_flush **pcf);
622 
623 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
624 				      struct ceph_cap *ocap, int mask);
625 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
626 extern int __ceph_caps_used(struct ceph_inode_info *ci);
627 
628 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
629 
630 /*
631  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
632  */
633 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
634 {
635 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
636 	if (w & CEPH_CAP_FILE_BUFFER)
637 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
638 	return w;
639 }
640 
641 /* what the mds thinks we want */
642 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
643 
644 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
645 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
646 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
647 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
648 			     struct ceph_cap_reservation *ctx, int need);
649 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
650 			       struct ceph_cap_reservation *ctx);
651 extern void ceph_reservation_status(struct ceph_fs_client *client,
652 				    int *total, int *avail, int *used,
653 				    int *reserved, int *min);
654 
655 
656 
657 /*
658  * we keep buffered readdir results attached to file->private_data
659  */
660 #define CEPH_F_SYNC     1
661 #define CEPH_F_ATEND    2
662 
663 struct ceph_file_info {
664 	short fmode;     /* initialized on open */
665 	short flags;     /* CEPH_F_* */
666 
667 	/* readdir: position within the dir */
668 	u32 frag;
669 	struct ceph_mds_request *last_readdir;
670 
671 	/* readdir: position within a frag */
672 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
673 	char *last_name;       /* last entry in previous chunk */
674 	long long dir_release_count;
675 	long long dir_ordered_count;
676 	int readdir_cache_idx;
677 
678 	/* used for -o dirstat read() on directory thing */
679 	char *dir_info;
680 	int dir_info_len;
681 };
682 
683 struct ceph_readdir_cache_control {
684 	struct page  *page;
685 	struct dentry **dentries;
686 	int index;
687 };
688 
689 /*
690  * A "snap realm" describes a subset of the file hierarchy sharing
691  * the same set of snapshots that apply to it.  The realms themselves
692  * are organized into a hierarchy, such that children inherit (some of)
693  * the snapshots of their parents.
694  *
695  * All inodes within the realm that have capabilities are linked into a
696  * per-realm list.
697  */
698 struct ceph_snap_realm {
699 	u64 ino;
700 	atomic_t nref;
701 	struct rb_node node;
702 
703 	u64 created, seq;
704 	u64 parent_ino;
705 	u64 parent_since;   /* snapid when our current parent became so */
706 
707 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
708 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
709 	u64 *snaps;                   /* snaps specific to this realm */
710 	u32 num_snaps;
711 
712 	struct ceph_snap_realm *parent;
713 	struct list_head children;       /* list of child realms */
714 	struct list_head child_item;
715 
716 	struct list_head empty_item;     /* if i have ref==0 */
717 
718 	struct list_head dirty_item;     /* if realm needs new context */
719 
720 	/* the current set of snaps for this realm */
721 	struct ceph_snap_context *cached_context;
722 
723 	struct list_head inodes_with_caps;
724 	spinlock_t inodes_with_caps_lock;
725 };
726 
727 static inline int default_congestion_kb(void)
728 {
729 	int congestion_kb;
730 
731 	/*
732 	 * Copied from NFS
733 	 *
734 	 * congestion size, scale with available memory.
735 	 *
736 	 *  64MB:    8192k
737 	 * 128MB:   11585k
738 	 * 256MB:   16384k
739 	 * 512MB:   23170k
740 	 *   1GB:   32768k
741 	 *   2GB:   46340k
742 	 *   4GB:   65536k
743 	 *   8GB:   92681k
744 	 *  16GB:  131072k
745 	 *
746 	 * This allows larger machines to have larger/more transfers.
747 	 * Limit the default to 256M
748 	 */
749 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
750 	if (congestion_kb > 256*1024)
751 		congestion_kb = 256*1024;
752 
753 	return congestion_kb;
754 }
755 
756 
757 
758 /* snap.c */
759 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
760 					       u64 ino);
761 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
762 				struct ceph_snap_realm *realm);
763 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
764 				struct ceph_snap_realm *realm);
765 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
766 				  void *p, void *e, bool deletion,
767 				  struct ceph_snap_realm **realm_ret);
768 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
769 			     struct ceph_mds_session *session,
770 			     struct ceph_msg *msg);
771 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
772 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
773 				  struct ceph_cap_snap *capsnap);
774 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
775 
776 /*
777  * a cap_snap is "pending" if it is still awaiting an in-progress
778  * sync write (that may/may not still update size, mtime, etc.).
779  */
780 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
781 {
782 	return !list_empty(&ci->i_cap_snaps) &&
783 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
784 			       ci_item)->writing;
785 }
786 
787 /* inode.c */
788 extern const struct inode_operations ceph_file_iops;
789 
790 extern struct inode *ceph_alloc_inode(struct super_block *sb);
791 extern void ceph_destroy_inode(struct inode *inode);
792 extern int ceph_drop_inode(struct inode *inode);
793 
794 extern struct inode *ceph_get_inode(struct super_block *sb,
795 				    struct ceph_vino vino);
796 extern struct inode *ceph_get_snapdir(struct inode *parent);
797 extern int ceph_fill_file_size(struct inode *inode, int issued,
798 			       u32 truncate_seq, u64 truncate_size, u64 size);
799 extern void ceph_fill_file_time(struct inode *inode, int issued,
800 				u64 time_warp_seq, struct timespec *ctime,
801 				struct timespec *mtime, struct timespec *atime);
802 extern int ceph_fill_trace(struct super_block *sb,
803 			   struct ceph_mds_request *req);
804 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
805 				    struct ceph_mds_session *session);
806 
807 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
808 
809 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
810 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
811 extern void ceph_queue_vmtruncate(struct inode *inode);
812 
813 extern void ceph_queue_invalidate(struct inode *inode);
814 extern void ceph_queue_writeback(struct inode *inode);
815 
816 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
817 			     int mask, bool force);
818 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
819 {
820 	return __ceph_do_getattr(inode, NULL, mask, force);
821 }
822 extern int ceph_permission(struct inode *inode, int mask);
823 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
824 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
825 extern int ceph_getattr(const struct path *path, struct kstat *stat,
826 			u32 request_mask, unsigned int flags);
827 
828 /* xattr.c */
829 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
830 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
831 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
832 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
833 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
834 extern void __init ceph_xattr_init(void);
835 extern void ceph_xattr_exit(void);
836 extern const struct xattr_handler *ceph_xattr_handlers[];
837 
838 #ifdef CONFIG_SECURITY
839 extern bool ceph_security_xattr_deadlock(struct inode *in);
840 extern bool ceph_security_xattr_wanted(struct inode *in);
841 #else
842 static inline bool ceph_security_xattr_deadlock(struct inode *in)
843 {
844 	return false;
845 }
846 static inline bool ceph_security_xattr_wanted(struct inode *in)
847 {
848 	return false;
849 }
850 #endif
851 
852 /* acl.c */
853 struct ceph_acls_info {
854 	void *default_acl;
855 	void *acl;
856 	struct ceph_pagelist *pagelist;
857 };
858 
859 #ifdef CONFIG_CEPH_FS_POSIX_ACL
860 
861 struct posix_acl *ceph_get_acl(struct inode *, int);
862 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
863 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
864 		       struct ceph_acls_info *info);
865 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
866 void ceph_release_acls_info(struct ceph_acls_info *info);
867 
868 static inline void ceph_forget_all_cached_acls(struct inode *inode)
869 {
870        forget_all_cached_acls(inode);
871 }
872 
873 #else
874 
875 #define ceph_get_acl NULL
876 #define ceph_set_acl NULL
877 
878 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
879 				     struct ceph_acls_info *info)
880 {
881 	return 0;
882 }
883 static inline void ceph_init_inode_acls(struct inode *inode,
884 					struct ceph_acls_info *info)
885 {
886 }
887 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
888 {
889 }
890 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
891 {
892 	return 0;
893 }
894 
895 static inline void ceph_forget_all_cached_acls(struct inode *inode)
896 {
897 }
898 
899 #endif
900 
901 /* caps.c */
902 extern const char *ceph_cap_string(int c);
903 extern void ceph_handle_caps(struct ceph_mds_session *session,
904 			     struct ceph_msg *msg);
905 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
906 				     struct ceph_cap_reservation *ctx);
907 extern void ceph_add_cap(struct inode *inode,
908 			 struct ceph_mds_session *session, u64 cap_id,
909 			 int fmode, unsigned issued, unsigned wanted,
910 			 unsigned cap, unsigned seq, u64 realmino, int flags,
911 			 struct ceph_cap **new_cap);
912 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
913 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
914 			 struct ceph_cap *cap);
915 extern int ceph_is_any_caps(struct inode *inode);
916 
917 extern void ceph_queue_caps_release(struct inode *inode);
918 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
919 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
920 		      int datasync);
921 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
922 					  struct ceph_mds_session *session);
923 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
924 				    struct ceph_mds_session *session);
925 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
926 					     int mds);
927 extern int ceph_get_cap_mds(struct inode *inode);
928 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
929 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
930 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
931 				       struct ceph_snap_context *snapc);
932 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
933 			     struct ceph_mds_session **psession);
934 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
935 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
936 			    struct ceph_mds_session *session);
937 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
938 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
939 
940 extern int ceph_encode_inode_release(void **p, struct inode *inode,
941 				     int mds, int drop, int unless, int force);
942 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
943 				      struct inode *dir,
944 				      int mds, int drop, int unless);
945 
946 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
947 			 loff_t endoff, int *got, struct page **pinned_page);
948 extern int ceph_try_get_caps(struct ceph_inode_info *ci,
949 			     int need, int want, int *got);
950 
951 /* for counting open files by mode */
952 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode);
953 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
954 
955 /* addr.c */
956 extern const struct address_space_operations ceph_aops;
957 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
958 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
959 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
960 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
961 
962 /* file.c */
963 extern const struct file_operations ceph_file_fops;
964 
965 extern int ceph_renew_caps(struct inode *inode);
966 extern int ceph_open(struct inode *inode, struct file *file);
967 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
968 			    struct file *file, unsigned flags, umode_t mode,
969 			    int *opened);
970 extern int ceph_release(struct inode *inode, struct file *filp);
971 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
972 				  char *data, size_t len);
973 
974 /* dir.c */
975 extern const struct file_operations ceph_dir_fops;
976 extern const struct file_operations ceph_snapdir_fops;
977 extern const struct inode_operations ceph_dir_iops;
978 extern const struct inode_operations ceph_snapdir_iops;
979 extern const struct dentry_operations ceph_dentry_ops;
980 
981 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
982 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
983 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
984 			       struct dentry *dentry, int err);
985 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
986 					 struct dentry *dentry, int err);
987 
988 extern void ceph_dentry_lru_add(struct dentry *dn);
989 extern void ceph_dentry_lru_touch(struct dentry *dn);
990 extern void ceph_dentry_lru_del(struct dentry *dn);
991 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
992 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
993 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
994 
995 /* ioctl.c */
996 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
997 
998 /* export.c */
999 extern const struct export_operations ceph_export_ops;
1000 
1001 /* locks.c */
1002 extern __init void ceph_flock_init(void);
1003 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1004 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1005 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1006 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1007 				       struct ceph_filelock *flocks,
1008 				       int num_fcntl_locks,
1009 				       int num_flock_locks);
1010 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1011 				  struct ceph_pagelist *pagelist,
1012 				  int num_fcntl_locks, int num_flock_locks);
1013 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
1014 
1015 /* debugfs.c */
1016 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
1017 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1018 
1019 #endif /* _FS_CEPH_SUPER_H */
1020