xref: /openbmc/linux/fs/ceph/super.h (revision 9cfc5c90)
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 
18 #include <linux/ceph/libceph.h>
19 
20 #ifdef CONFIG_CEPH_FSCACHE
21 #include <linux/fscache.h>
22 #endif
23 
24 /* f_type in struct statfs */
25 #define CEPH_SUPER_MAGIC 0x00c36400
26 
27 /* large granularity for statfs utilization stats to facilitate
28  * large volume sizes on 32-bit machines. */
29 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
30 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
31 
32 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
33 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
34 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
35 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
36 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
37 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
38 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
39 
40 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES | \
41 				   CEPH_MOUNT_OPT_DCACHE)
42 
43 #define ceph_set_mount_opt(fsc, opt) \
44 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
45 #define ceph_test_mount_opt(fsc, opt) \
46 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
47 
48 #define CEPH_RSIZE_DEFAULT             0           /* max read size */
49 #define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
50 #define CEPH_MAX_READDIR_DEFAULT        1024
51 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
52 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
53 
54 struct ceph_mount_options {
55 	int flags;
56 	int sb_flags;
57 
58 	int wsize;            /* max write size */
59 	int rsize;            /* max read size */
60 	int rasize;           /* max readahead */
61 	int congestion_kb;    /* max writeback in flight */
62 	int caps_wanted_delay_min, caps_wanted_delay_max;
63 	int cap_release_safety;
64 	int max_readdir;       /* max readdir result (entires) */
65 	int max_readdir_bytes; /* max readdir result (bytes) */
66 
67 	/*
68 	 * everything above this point can be memcmp'd; everything below
69 	 * is handled in compare_mount_options()
70 	 */
71 
72 	char *snapdir_name;   /* default ".snap" */
73 };
74 
75 struct ceph_fs_client {
76 	struct super_block *sb;
77 
78 	struct ceph_mount_options *mount_options;
79 	struct ceph_client *client;
80 
81 	unsigned long mount_state;
82 	int min_caps;                  /* min caps i added */
83 
84 	struct ceph_mds_client *mdsc;
85 
86 	/* writeback */
87 	mempool_t *wb_pagevec_pool;
88 	struct workqueue_struct *wb_wq;
89 	struct workqueue_struct *pg_inv_wq;
90 	struct workqueue_struct *trunc_wq;
91 	atomic_long_t writeback_count;
92 
93 	struct backing_dev_info backing_dev_info;
94 
95 #ifdef CONFIG_DEBUG_FS
96 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
97 	struct dentry *debugfs_congestion_kb;
98 	struct dentry *debugfs_bdi;
99 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
100 	struct dentry *debugfs_mds_sessions;
101 #endif
102 
103 #ifdef CONFIG_CEPH_FSCACHE
104 	struct fscache_cookie *fscache;
105 	struct workqueue_struct *revalidate_wq;
106 #endif
107 };
108 
109 
110 /*
111  * File i/o capability.  This tracks shared state with the metadata
112  * server that allows us to cache or writeback attributes or to read
113  * and write data.  For any given inode, we should have one or more
114  * capabilities, one issued by each metadata server, and our
115  * cumulative access is the OR of all issued capabilities.
116  *
117  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
118  * session capability lists.
119  */
120 struct ceph_cap {
121 	struct ceph_inode_info *ci;
122 	struct rb_node ci_node;          /* per-ci cap tree */
123 	struct ceph_mds_session *session;
124 	struct list_head session_caps;   /* per-session caplist */
125 	u64 cap_id;       /* unique cap id (mds provided) */
126 	union {
127 		/* in-use caps */
128 		struct {
129 			int issued;       /* latest, from the mds */
130 			int implemented;  /* implemented superset of
131 					     issued (for revocation) */
132 			int mds, mds_wanted;
133 		};
134 		/* caps to release */
135 		struct {
136 			u64 cap_ino;
137 			int queue_release;
138 		};
139 	};
140 	u32 seq, issue_seq, mseq;
141 	u32 cap_gen;      /* active/stale cycle */
142 	unsigned long last_used;
143 	struct list_head caps_item;
144 };
145 
146 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
147 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
148 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
149 
150 /*
151  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
152  * we first complete any in-process sync writes and writeback any dirty
153  * data before flushing the snapped state (tracked here) back to the MDS.
154  */
155 struct ceph_cap_snap {
156 	atomic_t nref;
157 	struct ceph_inode_info *ci;
158 	struct list_head ci_item, flushing_item;
159 
160 	u64 follows, flush_tid;
161 	int issued, dirty;
162 	struct ceph_snap_context *context;
163 
164 	umode_t mode;
165 	kuid_t uid;
166 	kgid_t gid;
167 
168 	struct ceph_buffer *xattr_blob;
169 	u64 xattr_version;
170 
171 	u64 size;
172 	struct timespec mtime, atime, ctime;
173 	u64 time_warp_seq;
174 	int writing;   /* a sync write is still in progress */
175 	int dirty_pages;     /* dirty pages awaiting writeback */
176 	bool inline_data;
177 	bool need_flush;
178 };
179 
180 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
181 {
182 	if (atomic_dec_and_test(&capsnap->nref)) {
183 		if (capsnap->xattr_blob)
184 			ceph_buffer_put(capsnap->xattr_blob);
185 		kfree(capsnap);
186 	}
187 }
188 
189 struct ceph_cap_flush {
190 	u64 tid;
191 	int caps;
192 	struct rb_node g_node; // global
193 	union {
194 		struct rb_node i_node; // inode
195 		struct list_head list;
196 	};
197 };
198 
199 /*
200  * The frag tree describes how a directory is fragmented, potentially across
201  * multiple metadata servers.  It is also used to indicate points where
202  * metadata authority is delegated, and whether/where metadata is replicated.
203  *
204  * A _leaf_ frag will be present in the i_fragtree IFF there is
205  * delegation info.  That is, if mds >= 0 || ndist > 0.
206  */
207 #define CEPH_MAX_DIRFRAG_REP 4
208 
209 struct ceph_inode_frag {
210 	struct rb_node node;
211 
212 	/* fragtree state */
213 	u32 frag;
214 	int split_by;         /* i.e. 2^(split_by) children */
215 
216 	/* delegation and replication info */
217 	int mds;              /* -1 if same authority as parent */
218 	int ndist;            /* >0 if replicated */
219 	int dist[CEPH_MAX_DIRFRAG_REP];
220 };
221 
222 /*
223  * We cache inode xattrs as an encoded blob until they are first used,
224  * at which point we parse them into an rbtree.
225  */
226 struct ceph_inode_xattr {
227 	struct rb_node node;
228 
229 	const char *name;
230 	int name_len;
231 	const char *val;
232 	int val_len;
233 	int dirty;
234 
235 	int should_free_name;
236 	int should_free_val;
237 };
238 
239 /*
240  * Ceph dentry state
241  */
242 struct ceph_dentry_info {
243 	struct ceph_mds_session *lease_session;
244 	u32 lease_gen, lease_shared_gen;
245 	u32 lease_seq;
246 	unsigned long lease_renew_after, lease_renew_from;
247 	struct list_head lru;
248 	struct dentry *dentry;
249 	u64 time;
250 	u64 offset;
251 };
252 
253 struct ceph_inode_xattrs_info {
254 	/*
255 	 * (still encoded) xattr blob. we avoid the overhead of parsing
256 	 * this until someone actually calls getxattr, etc.
257 	 *
258 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
259 	 * NULL means we don't know.
260 	*/
261 	struct ceph_buffer *blob, *prealloc_blob;
262 
263 	struct rb_root index;
264 	bool dirty;
265 	int count;
266 	int names_size;
267 	int vals_size;
268 	u64 version, index_version;
269 };
270 
271 /*
272  * Ceph inode.
273  */
274 struct ceph_inode_info {
275 	struct ceph_vino i_vino;   /* ceph ino + snap */
276 
277 	spinlock_t i_ceph_lock;
278 
279 	u64 i_version;
280 	u64 i_inline_version;
281 	u32 i_time_warp_seq;
282 
283 	unsigned i_ceph_flags;
284 	atomic64_t i_release_count;
285 	atomic64_t i_ordered_count;
286 	atomic64_t i_complete_seq[2];
287 
288 	struct ceph_dir_layout i_dir_layout;
289 	struct ceph_file_layout i_layout;
290 	char *i_symlink;
291 
292 	/* for dirs */
293 	struct timespec i_rctime;
294 	u64 i_rbytes, i_rfiles, i_rsubdirs;
295 	u64 i_files, i_subdirs;
296 
297 	struct rb_root i_fragtree;
298 	struct mutex i_fragtree_mutex;
299 
300 	struct ceph_inode_xattrs_info i_xattrs;
301 
302 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
303 	 * s_mutex. */
304 	struct rb_root i_caps;           /* cap list */
305 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
306 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
307 	struct list_head i_dirty_item, i_flushing_item;
308 	/* we need to track cap writeback on a per-cap-bit basis, to allow
309 	 * overlapping, pipelined cap flushes to the mds.  we can probably
310 	 * reduce the tid to 8 bits if we're concerned about inode size. */
311 	struct ceph_cap_flush *i_prealloc_cap_flush;
312 	struct rb_root i_cap_flush_tree;
313 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
314 	unsigned long i_hold_caps_min; /* jiffies */
315 	unsigned long i_hold_caps_max; /* jiffies */
316 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
317 	struct ceph_cap_reservation i_cap_migration_resv;
318 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
319 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
320 						    dirty|flushing caps */
321 	unsigned i_snap_caps;           /* cap bits for snapped files */
322 
323 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
324 
325 	struct mutex i_truncate_mutex;
326 	u32 i_truncate_seq;        /* last truncate to smaller size */
327 	u64 i_truncate_size;       /*  and the size we last truncated down to */
328 	int i_truncate_pending;    /*  still need to call vmtruncate */
329 
330 	u64 i_max_size;            /* max file size authorized by mds */
331 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
332 	u64 i_wanted_max_size;     /* offset we'd like to write too */
333 	u64 i_requested_max_size;  /* max_size we've requested */
334 
335 	/* held references to caps */
336 	int i_pin_ref;
337 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
338 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
339 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
340 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
341 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
342 
343 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
344 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
345 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
346 	spinlock_t i_unsafe_lock;
347 
348 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
349 	int i_snap_realm_counter; /* snap realm (if caps) */
350 	struct list_head i_snap_realm_item;
351 	struct list_head i_snap_flush_item;
352 
353 	struct work_struct i_wb_work;  /* writeback work */
354 	struct work_struct i_pg_inv_work;  /* page invalidation work */
355 
356 	struct work_struct i_vmtruncate_work;
357 
358 #ifdef CONFIG_CEPH_FSCACHE
359 	struct fscache_cookie *fscache;
360 	u32 i_fscache_gen; /* sequence, for delayed fscache validate */
361 	struct work_struct i_revalidate_work;
362 #endif
363 	struct inode vfs_inode; /* at end */
364 };
365 
366 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
367 {
368 	return container_of(inode, struct ceph_inode_info, vfs_inode);
369 }
370 
371 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
372 {
373 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
374 }
375 
376 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
377 {
378 	return (struct ceph_fs_client *)sb->s_fs_info;
379 }
380 
381 static inline struct ceph_vino ceph_vino(struct inode *inode)
382 {
383 	return ceph_inode(inode)->i_vino;
384 }
385 
386 /*
387  * ino_t is <64 bits on many architectures, blech.
388  *
389  *               i_ino (kernel inode)   st_ino (userspace)
390  * i386          32                     32
391  * x86_64+ino32  64                     32
392  * x86_64        64                     64
393  */
394 static inline u32 ceph_ino_to_ino32(__u64 vino)
395 {
396 	u32 ino = vino & 0xffffffff;
397 	ino ^= vino >> 32;
398 	if (!ino)
399 		ino = 2;
400 	return ino;
401 }
402 
403 /*
404  * kernel i_ino value
405  */
406 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
407 {
408 #if BITS_PER_LONG == 32
409 	return ceph_ino_to_ino32(vino.ino);
410 #else
411 	return (ino_t)vino.ino;
412 #endif
413 }
414 
415 /*
416  * user-visible ino (stat, filldir)
417  */
418 #if BITS_PER_LONG == 32
419 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
420 {
421 	return ino;
422 }
423 #else
424 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
425 {
426 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
427 		ino = ceph_ino_to_ino32(ino);
428 	return ino;
429 }
430 #endif
431 
432 
433 /* for printf-style formatting */
434 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
435 
436 static inline u64 ceph_ino(struct inode *inode)
437 {
438 	return ceph_inode(inode)->i_vino.ino;
439 }
440 static inline u64 ceph_snap(struct inode *inode)
441 {
442 	return ceph_inode(inode)->i_vino.snap;
443 }
444 
445 static inline int ceph_ino_compare(struct inode *inode, void *data)
446 {
447 	struct ceph_vino *pvino = (struct ceph_vino *)data;
448 	struct ceph_inode_info *ci = ceph_inode(inode);
449 	return ci->i_vino.ino == pvino->ino &&
450 		ci->i_vino.snap == pvino->snap;
451 }
452 
453 static inline struct inode *ceph_find_inode(struct super_block *sb,
454 					    struct ceph_vino vino)
455 {
456 	ino_t t = ceph_vino_to_ino(vino);
457 	return ilookup5(sb, t, ceph_ino_compare, &vino);
458 }
459 
460 
461 /*
462  * Ceph inode.
463  */
464 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
465 #define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
466 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
467 #define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
468 #define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
469 #define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
470 #define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
471 
472 
473 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
474 					   long long release_count,
475 					   long long ordered_count)
476 {
477 	smp_mb__before_atomic();
478 	atomic64_set(&ci->i_complete_seq[0], release_count);
479 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
480 }
481 
482 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
483 {
484 	atomic64_inc(&ci->i_release_count);
485 }
486 
487 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
488 {
489 	atomic64_inc(&ci->i_ordered_count);
490 }
491 
492 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
493 {
494 	return atomic64_read(&ci->i_complete_seq[0]) ==
495 		atomic64_read(&ci->i_release_count);
496 }
497 
498 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
499 {
500 	return  atomic64_read(&ci->i_complete_seq[0]) ==
501 		atomic64_read(&ci->i_release_count) &&
502 		atomic64_read(&ci->i_complete_seq[1]) ==
503 		atomic64_read(&ci->i_ordered_count);
504 }
505 
506 static inline void ceph_dir_clear_complete(struct inode *inode)
507 {
508 	__ceph_dir_clear_complete(ceph_inode(inode));
509 }
510 
511 static inline void ceph_dir_clear_ordered(struct inode *inode)
512 {
513 	__ceph_dir_clear_ordered(ceph_inode(inode));
514 }
515 
516 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
517 {
518 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
519 	smp_rmb();
520 	return ret;
521 }
522 
523 /* find a specific frag @f */
524 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
525 						u32 f);
526 
527 /*
528  * choose fragment for value @v.  copy frag content to pfrag, if leaf
529  * exists
530  */
531 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
532 			    struct ceph_inode_frag *pfrag,
533 			    int *found);
534 
535 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
536 {
537 	return (struct ceph_dentry_info *)dentry->d_fsdata;
538 }
539 
540 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
541 {
542 	return ((loff_t)frag << 32) | (loff_t)off;
543 }
544 
545 /*
546  * caps helpers
547  */
548 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
549 {
550 	return !RB_EMPTY_ROOT(&ci->i_caps);
551 }
552 
553 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
554 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
555 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
556 				    struct ceph_cap *cap);
557 
558 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
559 {
560 	int issued;
561 	spin_lock(&ci->i_ceph_lock);
562 	issued = __ceph_caps_issued(ci, NULL);
563 	spin_unlock(&ci->i_ceph_lock);
564 	return issued;
565 }
566 
567 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
568 					int touch)
569 {
570 	int r;
571 	spin_lock(&ci->i_ceph_lock);
572 	r = __ceph_caps_issued_mask(ci, mask, touch);
573 	spin_unlock(&ci->i_ceph_lock);
574 	return r;
575 }
576 
577 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
578 {
579 	return ci->i_dirty_caps | ci->i_flushing_caps;
580 }
581 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
582 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
583 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
584 				  struct ceph_cap_flush **pcf);
585 
586 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
587 				      struct ceph_cap *ocap, int mask);
588 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
589 extern int __ceph_caps_used(struct ceph_inode_info *ci);
590 
591 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
592 
593 /*
594  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
595  */
596 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
597 {
598 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
599 	if (w & CEPH_CAP_FILE_BUFFER)
600 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
601 	return w;
602 }
603 
604 /* what the mds thinks we want */
605 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
606 
607 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
608 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
609 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
610 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
611 			     struct ceph_cap_reservation *ctx, int need);
612 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
613 			       struct ceph_cap_reservation *ctx);
614 extern void ceph_reservation_status(struct ceph_fs_client *client,
615 				    int *total, int *avail, int *used,
616 				    int *reserved, int *min);
617 
618 
619 
620 /*
621  * we keep buffered readdir results attached to file->private_data
622  */
623 #define CEPH_F_SYNC     1
624 #define CEPH_F_ATEND    2
625 
626 struct ceph_file_info {
627 	short fmode;     /* initialized on open */
628 	short flags;     /* CEPH_F_* */
629 
630 	/* readdir: position within the dir */
631 	u32 frag;
632 	struct ceph_mds_request *last_readdir;
633 
634 	/* readdir: position within a frag */
635 	unsigned offset;       /* offset of last chunk, adjusted for . and .. */
636 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
637 	char *last_name;       /* last entry in previous chunk */
638 	long long dir_release_count;
639 	long long dir_ordered_count;
640 	int readdir_cache_idx;
641 
642 	/* used for -o dirstat read() on directory thing */
643 	char *dir_info;
644 	int dir_info_len;
645 };
646 
647 struct ceph_readdir_cache_control {
648 	struct page  *page;
649 	struct dentry **dentries;
650 	int index;
651 };
652 
653 /*
654  * A "snap realm" describes a subset of the file hierarchy sharing
655  * the same set of snapshots that apply to it.  The realms themselves
656  * are organized into a hierarchy, such that children inherit (some of)
657  * the snapshots of their parents.
658  *
659  * All inodes within the realm that have capabilities are linked into a
660  * per-realm list.
661  */
662 struct ceph_snap_realm {
663 	u64 ino;
664 	atomic_t nref;
665 	struct rb_node node;
666 
667 	u64 created, seq;
668 	u64 parent_ino;
669 	u64 parent_since;   /* snapid when our current parent became so */
670 
671 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
672 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
673 	u64 *snaps;                   /* snaps specific to this realm */
674 	u32 num_snaps;
675 
676 	struct ceph_snap_realm *parent;
677 	struct list_head children;       /* list of child realms */
678 	struct list_head child_item;
679 
680 	struct list_head empty_item;     /* if i have ref==0 */
681 
682 	struct list_head dirty_item;     /* if realm needs new context */
683 
684 	/* the current set of snaps for this realm */
685 	struct ceph_snap_context *cached_context;
686 
687 	struct list_head inodes_with_caps;
688 	spinlock_t inodes_with_caps_lock;
689 };
690 
691 static inline int default_congestion_kb(void)
692 {
693 	int congestion_kb;
694 
695 	/*
696 	 * Copied from NFS
697 	 *
698 	 * congestion size, scale with available memory.
699 	 *
700 	 *  64MB:    8192k
701 	 * 128MB:   11585k
702 	 * 256MB:   16384k
703 	 * 512MB:   23170k
704 	 *   1GB:   32768k
705 	 *   2GB:   46340k
706 	 *   4GB:   65536k
707 	 *   8GB:   92681k
708 	 *  16GB:  131072k
709 	 *
710 	 * This allows larger machines to have larger/more transfers.
711 	 * Limit the default to 256M
712 	 */
713 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
714 	if (congestion_kb > 256*1024)
715 		congestion_kb = 256*1024;
716 
717 	return congestion_kb;
718 }
719 
720 
721 
722 /* snap.c */
723 extern struct ceph_snap_context *ceph_empty_snapc;
724 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
725 					       u64 ino);
726 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
727 				struct ceph_snap_realm *realm);
728 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
729 				struct ceph_snap_realm *realm);
730 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
731 				  void *p, void *e, bool deletion,
732 				  struct ceph_snap_realm **realm_ret);
733 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
734 			     struct ceph_mds_session *session,
735 			     struct ceph_msg *msg);
736 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
737 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
738 				  struct ceph_cap_snap *capsnap);
739 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
740 extern int ceph_snap_init(void);
741 extern void ceph_snap_exit(void);
742 
743 /*
744  * a cap_snap is "pending" if it is still awaiting an in-progress
745  * sync write (that may/may not still update size, mtime, etc.).
746  */
747 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
748 {
749 	return !list_empty(&ci->i_cap_snaps) &&
750 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
751 			       ci_item)->writing;
752 }
753 
754 /* inode.c */
755 extern const struct inode_operations ceph_file_iops;
756 
757 extern struct inode *ceph_alloc_inode(struct super_block *sb);
758 extern void ceph_destroy_inode(struct inode *inode);
759 extern int ceph_drop_inode(struct inode *inode);
760 
761 extern struct inode *ceph_get_inode(struct super_block *sb,
762 				    struct ceph_vino vino);
763 extern struct inode *ceph_get_snapdir(struct inode *parent);
764 extern int ceph_fill_file_size(struct inode *inode, int issued,
765 			       u32 truncate_seq, u64 truncate_size, u64 size);
766 extern void ceph_fill_file_time(struct inode *inode, int issued,
767 				u64 time_warp_seq, struct timespec *ctime,
768 				struct timespec *mtime, struct timespec *atime);
769 extern int ceph_fill_trace(struct super_block *sb,
770 			   struct ceph_mds_request *req,
771 			   struct ceph_mds_session *session);
772 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
773 				    struct ceph_mds_session *session);
774 
775 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
776 
777 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
778 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
779 extern void ceph_queue_vmtruncate(struct inode *inode);
780 
781 extern void ceph_queue_invalidate(struct inode *inode);
782 extern void ceph_queue_writeback(struct inode *inode);
783 
784 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
785 			     int mask, bool force);
786 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
787 {
788 	return __ceph_do_getattr(inode, NULL, mask, force);
789 }
790 extern int ceph_permission(struct inode *inode, int mask);
791 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
792 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
793 			struct kstat *stat);
794 
795 /* xattr.c */
796 extern int ceph_setxattr(struct dentry *, const char *, const void *,
797 			 size_t, int);
798 int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int);
799 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
800 int __ceph_removexattr(struct dentry *, const char *);
801 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
802 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
803 extern int ceph_removexattr(struct dentry *, const char *);
804 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
805 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
806 extern void __init ceph_xattr_init(void);
807 extern void ceph_xattr_exit(void);
808 extern const struct xattr_handler *ceph_xattr_handlers[];
809 
810 /* acl.c */
811 struct ceph_acls_info {
812 	void *default_acl;
813 	void *acl;
814 	struct ceph_pagelist *pagelist;
815 };
816 
817 #ifdef CONFIG_CEPH_FS_POSIX_ACL
818 
819 struct posix_acl *ceph_get_acl(struct inode *, int);
820 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
821 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
822 		       struct ceph_acls_info *info);
823 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
824 void ceph_release_acls_info(struct ceph_acls_info *info);
825 
826 static inline void ceph_forget_all_cached_acls(struct inode *inode)
827 {
828        forget_all_cached_acls(inode);
829 }
830 
831 #else
832 
833 #define ceph_get_acl NULL
834 #define ceph_set_acl NULL
835 
836 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
837 				     struct ceph_acls_info *info)
838 {
839 	return 0;
840 }
841 static inline void ceph_init_inode_acls(struct inode *inode,
842 					struct ceph_acls_info *info)
843 {
844 }
845 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
846 {
847 }
848 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
849 {
850 	return 0;
851 }
852 
853 static inline void ceph_forget_all_cached_acls(struct inode *inode)
854 {
855 }
856 
857 #endif
858 
859 /* caps.c */
860 extern const char *ceph_cap_string(int c);
861 extern void ceph_handle_caps(struct ceph_mds_session *session,
862 			     struct ceph_msg *msg);
863 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
864 				     struct ceph_cap_reservation *ctx);
865 extern void ceph_add_cap(struct inode *inode,
866 			 struct ceph_mds_session *session, u64 cap_id,
867 			 int fmode, unsigned issued, unsigned wanted,
868 			 unsigned cap, unsigned seq, u64 realmino, int flags,
869 			 struct ceph_cap **new_cap);
870 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
871 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
872 			 struct ceph_cap *cap);
873 extern int ceph_is_any_caps(struct inode *inode);
874 
875 extern void ceph_queue_caps_release(struct inode *inode);
876 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
877 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
878 		      int datasync);
879 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
880 					  struct ceph_mds_session *session);
881 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
882 				    struct ceph_mds_session *session);
883 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
884 					     int mds);
885 extern int ceph_get_cap_mds(struct inode *inode);
886 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
887 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
888 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
889 				       struct ceph_snap_context *snapc);
890 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
891 			       struct ceph_mds_session **psession,
892 			       int again);
893 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
894 			    struct ceph_mds_session *session);
895 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
896 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
897 
898 extern int ceph_encode_inode_release(void **p, struct inode *inode,
899 				     int mds, int drop, int unless, int force);
900 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
901 				      int mds, int drop, int unless);
902 
903 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
904 			 loff_t endoff, int *got, struct page **pinned_page);
905 
906 /* for counting open files by mode */
907 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
908 {
909 	ci->i_nr_by_mode[mode]++;
910 }
911 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
912 
913 /* addr.c */
914 extern const struct address_space_operations ceph_aops;
915 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
916 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
917 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
918 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
919 
920 /* file.c */
921 extern const struct file_operations ceph_file_fops;
922 
923 extern int ceph_open(struct inode *inode, struct file *file);
924 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
925 			    struct file *file, unsigned flags, umode_t mode,
926 			    int *opened);
927 extern int ceph_release(struct inode *inode, struct file *filp);
928 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
929 				  char *data, size_t len);
930 /* dir.c */
931 extern const struct file_operations ceph_dir_fops;
932 extern const struct file_operations ceph_snapdir_fops;
933 extern const struct inode_operations ceph_dir_iops;
934 extern const struct inode_operations ceph_snapdir_iops;
935 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
936 	ceph_snapdir_dentry_ops;
937 
938 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
939 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
940 			       struct dentry *dentry, int err);
941 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
942 					 struct dentry *dentry, int err);
943 
944 extern void ceph_dentry_lru_add(struct dentry *dn);
945 extern void ceph_dentry_lru_touch(struct dentry *dn);
946 extern void ceph_dentry_lru_del(struct dentry *dn);
947 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
948 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
949 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
950 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
951 
952 /*
953  * our d_ops vary depending on whether the inode is live,
954  * snapshotted (read-only), or a virtual ".snap" directory.
955  */
956 int ceph_init_dentry(struct dentry *dentry);
957 
958 
959 /* ioctl.c */
960 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
961 
962 /* export.c */
963 extern const struct export_operations ceph_export_ops;
964 
965 /* locks.c */
966 extern __init void ceph_flock_init(void);
967 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
968 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
969 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
970 extern int ceph_encode_locks_to_buffer(struct inode *inode,
971 				       struct ceph_filelock *flocks,
972 				       int num_fcntl_locks,
973 				       int num_flock_locks);
974 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
975 				  struct ceph_pagelist *pagelist,
976 				  int num_fcntl_locks, int num_flock_locks);
977 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
978 
979 /* debugfs.c */
980 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
981 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
982 
983 #endif /* _FS_CEPH_SUPER_H */
984