xref: /openbmc/linux/fs/ceph/super.h (revision 0edbfea5)
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_DCACHE
41 
42 #define ceph_set_mount_opt(fsc, opt) \
43 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
44 #define ceph_test_mount_opt(fsc, opt) \
45 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
46 
47 #define CEPH_RSIZE_DEFAULT             0           /* max read size */
48 #define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
49 #define CEPH_MAX_READDIR_DEFAULT        1024
50 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
51 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
52 
53 struct ceph_mount_options {
54 	int flags;
55 	int sb_flags;
56 
57 	int wsize;            /* max write size */
58 	int rsize;            /* max read size */
59 	int rasize;           /* max readahead */
60 	int congestion_kb;    /* max writeback in flight */
61 	int caps_wanted_delay_min, caps_wanted_delay_max;
62 	int cap_release_safety;
63 	int max_readdir;       /* max readdir result (entires) */
64 	int max_readdir_bytes; /* max readdir result (bytes) */
65 	int mds_namespace;
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 	char *server_path;    /* default  "/" */
74 };
75 
76 struct ceph_fs_client {
77 	struct super_block *sb;
78 
79 	struct ceph_mount_options *mount_options;
80 	struct ceph_client *client;
81 
82 	unsigned long mount_state;
83 	int min_caps;                  /* min caps i added */
84 
85 	struct ceph_mds_client *mdsc;
86 
87 	/* writeback */
88 	mempool_t *wb_pagevec_pool;
89 	struct workqueue_struct *wb_wq;
90 	struct workqueue_struct *pg_inv_wq;
91 	struct workqueue_struct *trunc_wq;
92 	atomic_long_t writeback_count;
93 
94 	struct backing_dev_info backing_dev_info;
95 
96 #ifdef CONFIG_DEBUG_FS
97 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
98 	struct dentry *debugfs_congestion_kb;
99 	struct dentry *debugfs_bdi;
100 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
101 	struct dentry *debugfs_mds_sessions;
102 #endif
103 
104 #ifdef CONFIG_CEPH_FSCACHE
105 	struct fscache_cookie *fscache;
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 	size_t i_pool_ns_len;
291 	char *i_symlink;
292 
293 	/* for dirs */
294 	struct timespec i_rctime;
295 	u64 i_rbytes, i_rfiles, i_rsubdirs;
296 	u64 i_files, i_subdirs;
297 
298 	struct rb_root i_fragtree;
299 	int i_fragtree_nsplits;
300 	struct mutex i_fragtree_mutex;
301 
302 	struct ceph_inode_xattrs_info i_xattrs;
303 
304 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
305 	 * s_mutex. */
306 	struct rb_root i_caps;           /* cap list */
307 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
308 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
309 	struct list_head i_dirty_item, i_flushing_item;
310 	/* we need to track cap writeback on a per-cap-bit basis, to allow
311 	 * overlapping, pipelined cap flushes to the mds.  we can probably
312 	 * reduce the tid to 8 bits if we're concerned about inode size. */
313 	struct ceph_cap_flush *i_prealloc_cap_flush;
314 	struct rb_root i_cap_flush_tree;
315 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
316 	unsigned long i_hold_caps_min; /* jiffies */
317 	unsigned long i_hold_caps_max; /* jiffies */
318 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
319 	struct ceph_cap_reservation i_cap_migration_resv;
320 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
321 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
322 						    dirty|flushing caps */
323 	unsigned i_snap_caps;           /* cap bits for snapped files */
324 
325 	int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
326 
327 	struct mutex i_truncate_mutex;
328 	u32 i_truncate_seq;        /* last truncate to smaller size */
329 	u64 i_truncate_size;       /*  and the size we last truncated down to */
330 	int i_truncate_pending;    /*  still need to call vmtruncate */
331 
332 	u64 i_max_size;            /* max file size authorized by mds */
333 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
334 	u64 i_wanted_max_size;     /* offset we'd like to write too */
335 	u64 i_requested_max_size;  /* max_size we've requested */
336 
337 	/* held references to caps */
338 	int i_pin_ref;
339 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
340 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
341 	u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
342 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
343 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
344 
345 	struct list_head i_unsafe_writes; /* uncommitted sync writes */
346 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
347 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
348 	spinlock_t i_unsafe_lock;
349 
350 	struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
351 	int i_snap_realm_counter; /* snap realm (if caps) */
352 	struct list_head i_snap_realm_item;
353 	struct list_head i_snap_flush_item;
354 
355 	struct work_struct i_wb_work;  /* writeback work */
356 	struct work_struct i_pg_inv_work;  /* page invalidation work */
357 
358 	struct work_struct i_vmtruncate_work;
359 
360 #ifdef CONFIG_CEPH_FSCACHE
361 	struct fscache_cookie *fscache;
362 	u32 i_fscache_gen;
363 #endif
364 	struct inode vfs_inode; /* at end */
365 };
366 
367 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
368 {
369 	return container_of(inode, struct ceph_inode_info, vfs_inode);
370 }
371 
372 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
373 {
374 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
375 }
376 
377 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
378 {
379 	return (struct ceph_fs_client *)sb->s_fs_info;
380 }
381 
382 static inline struct ceph_vino ceph_vino(struct inode *inode)
383 {
384 	return ceph_inode(inode)->i_vino;
385 }
386 
387 /*
388  * ino_t is <64 bits on many architectures, blech.
389  *
390  *               i_ino (kernel inode)   st_ino (userspace)
391  * i386          32                     32
392  * x86_64+ino32  64                     32
393  * x86_64        64                     64
394  */
395 static inline u32 ceph_ino_to_ino32(__u64 vino)
396 {
397 	u32 ino = vino & 0xffffffff;
398 	ino ^= vino >> 32;
399 	if (!ino)
400 		ino = 2;
401 	return ino;
402 }
403 
404 /*
405  * kernel i_ino value
406  */
407 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
408 {
409 #if BITS_PER_LONG == 32
410 	return ceph_ino_to_ino32(vino.ino);
411 #else
412 	return (ino_t)vino.ino;
413 #endif
414 }
415 
416 /*
417  * user-visible ino (stat, filldir)
418  */
419 #if BITS_PER_LONG == 32
420 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
421 {
422 	return ino;
423 }
424 #else
425 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
426 {
427 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
428 		ino = ceph_ino_to_ino32(ino);
429 	return ino;
430 }
431 #endif
432 
433 
434 /* for printf-style formatting */
435 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
436 
437 static inline u64 ceph_ino(struct inode *inode)
438 {
439 	return ceph_inode(inode)->i_vino.ino;
440 }
441 static inline u64 ceph_snap(struct inode *inode)
442 {
443 	return ceph_inode(inode)->i_vino.snap;
444 }
445 
446 static inline int ceph_ino_compare(struct inode *inode, void *data)
447 {
448 	struct ceph_vino *pvino = (struct ceph_vino *)data;
449 	struct ceph_inode_info *ci = ceph_inode(inode);
450 	return ci->i_vino.ino == pvino->ino &&
451 		ci->i_vino.snap == pvino->snap;
452 }
453 
454 static inline struct inode *ceph_find_inode(struct super_block *sb,
455 					    struct ceph_vino vino)
456 {
457 	ino_t t = ceph_vino_to_ino(vino);
458 	return ilookup5(sb, t, ceph_ino_compare, &vino);
459 }
460 
461 
462 /*
463  * Ceph inode.
464  */
465 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
466 #define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
467 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
468 #define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
469 #define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
470 #define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
471 #define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
472 #define CEPH_I_SEC_INITED	(1 << 7)  /* security initialized */
473 #define CEPH_I_CAP_DROPPED	(1 << 8)  /* caps were forcibly dropped */
474 
475 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
476 					   long long release_count,
477 					   long long ordered_count)
478 {
479 	smp_mb__before_atomic();
480 	atomic64_set(&ci->i_complete_seq[0], release_count);
481 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
482 }
483 
484 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
485 {
486 	atomic64_inc(&ci->i_release_count);
487 }
488 
489 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
490 {
491 	atomic64_inc(&ci->i_ordered_count);
492 }
493 
494 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
495 {
496 	return atomic64_read(&ci->i_complete_seq[0]) ==
497 		atomic64_read(&ci->i_release_count);
498 }
499 
500 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
501 {
502 	return  atomic64_read(&ci->i_complete_seq[0]) ==
503 		atomic64_read(&ci->i_release_count) &&
504 		atomic64_read(&ci->i_complete_seq[1]) ==
505 		atomic64_read(&ci->i_ordered_count);
506 }
507 
508 static inline void ceph_dir_clear_complete(struct inode *inode)
509 {
510 	__ceph_dir_clear_complete(ceph_inode(inode));
511 }
512 
513 static inline void ceph_dir_clear_ordered(struct inode *inode)
514 {
515 	__ceph_dir_clear_ordered(ceph_inode(inode));
516 }
517 
518 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
519 {
520 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
521 	smp_rmb();
522 	return ret;
523 }
524 
525 /* find a specific frag @f */
526 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
527 						u32 f);
528 
529 /*
530  * choose fragment for value @v.  copy frag content to pfrag, if leaf
531  * exists
532  */
533 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
534 			    struct ceph_inode_frag *pfrag,
535 			    int *found);
536 
537 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
538 {
539 	return (struct ceph_dentry_info *)dentry->d_fsdata;
540 }
541 
542 /*
543  * caps helpers
544  */
545 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
546 {
547 	return !RB_EMPTY_ROOT(&ci->i_caps);
548 }
549 
550 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
551 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
552 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
553 				    struct ceph_cap *cap);
554 
555 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
556 {
557 	int issued;
558 	spin_lock(&ci->i_ceph_lock);
559 	issued = __ceph_caps_issued(ci, NULL);
560 	spin_unlock(&ci->i_ceph_lock);
561 	return issued;
562 }
563 
564 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
565 					int touch)
566 {
567 	int r;
568 	spin_lock(&ci->i_ceph_lock);
569 	r = __ceph_caps_issued_mask(ci, mask, touch);
570 	spin_unlock(&ci->i_ceph_lock);
571 	return r;
572 }
573 
574 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
575 {
576 	return ci->i_dirty_caps | ci->i_flushing_caps;
577 }
578 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
579 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
580 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
581 				  struct ceph_cap_flush **pcf);
582 
583 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
584 				      struct ceph_cap *ocap, int mask);
585 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
586 extern int __ceph_caps_used(struct ceph_inode_info *ci);
587 
588 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
589 
590 /*
591  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
592  */
593 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
594 {
595 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
596 	if (w & CEPH_CAP_FILE_BUFFER)
597 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
598 	return w;
599 }
600 
601 /* what the mds thinks we want */
602 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
603 
604 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
605 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
606 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
607 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc,
608 			     struct ceph_cap_reservation *ctx, int need);
609 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
610 			       struct ceph_cap_reservation *ctx);
611 extern void ceph_reservation_status(struct ceph_fs_client *client,
612 				    int *total, int *avail, int *used,
613 				    int *reserved, int *min);
614 
615 
616 
617 /*
618  * we keep buffered readdir results attached to file->private_data
619  */
620 #define CEPH_F_SYNC     1
621 #define CEPH_F_ATEND    2
622 
623 struct ceph_file_info {
624 	short fmode;     /* initialized on open */
625 	short flags;     /* CEPH_F_* */
626 
627 	/* readdir: position within the dir */
628 	u32 frag;
629 	struct ceph_mds_request *last_readdir;
630 
631 	/* readdir: position within a frag */
632 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
633 	char *last_name;       /* last entry in previous chunk */
634 	long long dir_release_count;
635 	long long dir_ordered_count;
636 	int readdir_cache_idx;
637 
638 	/* used for -o dirstat read() on directory thing */
639 	char *dir_info;
640 	int dir_info_len;
641 };
642 
643 struct ceph_readdir_cache_control {
644 	struct page  *page;
645 	struct dentry **dentries;
646 	int index;
647 };
648 
649 /*
650  * A "snap realm" describes a subset of the file hierarchy sharing
651  * the same set of snapshots that apply to it.  The realms themselves
652  * are organized into a hierarchy, such that children inherit (some of)
653  * the snapshots of their parents.
654  *
655  * All inodes within the realm that have capabilities are linked into a
656  * per-realm list.
657  */
658 struct ceph_snap_realm {
659 	u64 ino;
660 	atomic_t nref;
661 	struct rb_node node;
662 
663 	u64 created, seq;
664 	u64 parent_ino;
665 	u64 parent_since;   /* snapid when our current parent became so */
666 
667 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
668 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
669 	u64 *snaps;                   /* snaps specific to this realm */
670 	u32 num_snaps;
671 
672 	struct ceph_snap_realm *parent;
673 	struct list_head children;       /* list of child realms */
674 	struct list_head child_item;
675 
676 	struct list_head empty_item;     /* if i have ref==0 */
677 
678 	struct list_head dirty_item;     /* if realm needs new context */
679 
680 	/* the current set of snaps for this realm */
681 	struct ceph_snap_context *cached_context;
682 
683 	struct list_head inodes_with_caps;
684 	spinlock_t inodes_with_caps_lock;
685 };
686 
687 static inline int default_congestion_kb(void)
688 {
689 	int congestion_kb;
690 
691 	/*
692 	 * Copied from NFS
693 	 *
694 	 * congestion size, scale with available memory.
695 	 *
696 	 *  64MB:    8192k
697 	 * 128MB:   11585k
698 	 * 256MB:   16384k
699 	 * 512MB:   23170k
700 	 *   1GB:   32768k
701 	 *   2GB:   46340k
702 	 *   4GB:   65536k
703 	 *   8GB:   92681k
704 	 *  16GB:  131072k
705 	 *
706 	 * This allows larger machines to have larger/more transfers.
707 	 * Limit the default to 256M
708 	 */
709 	congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
710 	if (congestion_kb > 256*1024)
711 		congestion_kb = 256*1024;
712 
713 	return congestion_kb;
714 }
715 
716 
717 
718 /* snap.c */
719 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
720 					       u64 ino);
721 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
722 				struct ceph_snap_realm *realm);
723 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
724 				struct ceph_snap_realm *realm);
725 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
726 				  void *p, void *e, bool deletion,
727 				  struct ceph_snap_realm **realm_ret);
728 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
729 			     struct ceph_mds_session *session,
730 			     struct ceph_msg *msg);
731 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
732 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
733 				  struct ceph_cap_snap *capsnap);
734 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
735 
736 /*
737  * a cap_snap is "pending" if it is still awaiting an in-progress
738  * sync write (that may/may not still update size, mtime, etc.).
739  */
740 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
741 {
742 	return !list_empty(&ci->i_cap_snaps) &&
743 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
744 			       ci_item)->writing;
745 }
746 
747 /* inode.c */
748 extern const struct inode_operations ceph_file_iops;
749 
750 extern struct inode *ceph_alloc_inode(struct super_block *sb);
751 extern void ceph_destroy_inode(struct inode *inode);
752 extern int ceph_drop_inode(struct inode *inode);
753 
754 extern struct inode *ceph_get_inode(struct super_block *sb,
755 				    struct ceph_vino vino);
756 extern struct inode *ceph_get_snapdir(struct inode *parent);
757 extern int ceph_fill_file_size(struct inode *inode, int issued,
758 			       u32 truncate_seq, u64 truncate_size, u64 size);
759 extern void ceph_fill_file_time(struct inode *inode, int issued,
760 				u64 time_warp_seq, struct timespec *ctime,
761 				struct timespec *mtime, struct timespec *atime);
762 extern int ceph_fill_trace(struct super_block *sb,
763 			   struct ceph_mds_request *req,
764 			   struct ceph_mds_session *session);
765 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
766 				    struct ceph_mds_session *session);
767 
768 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
769 
770 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
771 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
772 extern void ceph_queue_vmtruncate(struct inode *inode);
773 
774 extern void ceph_queue_invalidate(struct inode *inode);
775 extern void ceph_queue_writeback(struct inode *inode);
776 
777 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
778 			     int mask, bool force);
779 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
780 {
781 	return __ceph_do_getattr(inode, NULL, mask, force);
782 }
783 extern int ceph_permission(struct inode *inode, int mask);
784 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
785 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
786 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
787 			struct kstat *stat);
788 
789 /* xattr.c */
790 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
791 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
792 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
793 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
794 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
795 extern void __init ceph_xattr_init(void);
796 extern void ceph_xattr_exit(void);
797 extern const struct xattr_handler *ceph_xattr_handlers[];
798 
799 #ifdef CONFIG_SECURITY
800 extern bool ceph_security_xattr_deadlock(struct inode *in);
801 extern bool ceph_security_xattr_wanted(struct inode *in);
802 #else
803 static inline bool ceph_security_xattr_deadlock(struct inode *in)
804 {
805 	return false;
806 }
807 static inline bool ceph_security_xattr_wanted(struct inode *in)
808 {
809 	return false;
810 }
811 #endif
812 
813 /* acl.c */
814 struct ceph_acls_info {
815 	void *default_acl;
816 	void *acl;
817 	struct ceph_pagelist *pagelist;
818 };
819 
820 #ifdef CONFIG_CEPH_FS_POSIX_ACL
821 
822 struct posix_acl *ceph_get_acl(struct inode *, int);
823 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
824 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
825 		       struct ceph_acls_info *info);
826 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
827 void ceph_release_acls_info(struct ceph_acls_info *info);
828 
829 static inline void ceph_forget_all_cached_acls(struct inode *inode)
830 {
831        forget_all_cached_acls(inode);
832 }
833 
834 #else
835 
836 #define ceph_get_acl NULL
837 #define ceph_set_acl NULL
838 
839 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
840 				     struct ceph_acls_info *info)
841 {
842 	return 0;
843 }
844 static inline void ceph_init_inode_acls(struct inode *inode,
845 					struct ceph_acls_info *info)
846 {
847 }
848 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
849 {
850 }
851 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
852 {
853 	return 0;
854 }
855 
856 static inline void ceph_forget_all_cached_acls(struct inode *inode)
857 {
858 }
859 
860 #endif
861 
862 /* caps.c */
863 extern const char *ceph_cap_string(int c);
864 extern void ceph_handle_caps(struct ceph_mds_session *session,
865 			     struct ceph_msg *msg);
866 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
867 				     struct ceph_cap_reservation *ctx);
868 extern void ceph_add_cap(struct inode *inode,
869 			 struct ceph_mds_session *session, u64 cap_id,
870 			 int fmode, unsigned issued, unsigned wanted,
871 			 unsigned cap, unsigned seq, u64 realmino, int flags,
872 			 struct ceph_cap **new_cap);
873 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
874 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
875 			 struct ceph_cap *cap);
876 extern int ceph_is_any_caps(struct inode *inode);
877 
878 extern void ceph_queue_caps_release(struct inode *inode);
879 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
880 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
881 		      int datasync);
882 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
883 					  struct ceph_mds_session *session);
884 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
885 				    struct ceph_mds_session *session);
886 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
887 					     int mds);
888 extern int ceph_get_cap_mds(struct inode *inode);
889 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
890 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
891 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
892 				       struct ceph_snap_context *snapc);
893 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
894 			       struct ceph_mds_session **psession,
895 			       int again);
896 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
897 			    struct ceph_mds_session *session);
898 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
899 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
900 
901 extern int ceph_encode_inode_release(void **p, struct inode *inode,
902 				     int mds, int drop, int unless, int force);
903 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
904 				      int mds, int drop, int unless);
905 
906 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
907 			 loff_t endoff, int *got, struct page **pinned_page);
908 
909 /* for counting open files by mode */
910 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
911 {
912 	ci->i_nr_by_mode[mode]++;
913 }
914 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
915 
916 /* addr.c */
917 extern const struct address_space_operations ceph_aops;
918 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
919 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
920 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
921 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
922 
923 /* file.c */
924 extern const struct file_operations ceph_file_fops;
925 
926 extern int ceph_renew_caps(struct inode *inode);
927 extern int ceph_open(struct inode *inode, struct file *file);
928 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
929 			    struct file *file, unsigned flags, umode_t mode,
930 			    int *opened);
931 extern int ceph_release(struct inode *inode, struct file *filp);
932 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
933 				  char *data, size_t len);
934 /* dir.c */
935 extern const struct file_operations ceph_dir_fops;
936 extern const struct file_operations ceph_snapdir_fops;
937 extern const struct inode_operations ceph_dir_iops;
938 extern const struct inode_operations ceph_snapdir_iops;
939 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
940 	ceph_snapdir_dentry_ops;
941 
942 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
943 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
944 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
945 			       struct dentry *dentry, int err);
946 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
947 					 struct dentry *dentry, int err);
948 
949 extern void ceph_dentry_lru_add(struct dentry *dn);
950 extern void ceph_dentry_lru_touch(struct dentry *dn);
951 extern void ceph_dentry_lru_del(struct dentry *dn);
952 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
953 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
954 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
955 
956 /*
957  * our d_ops vary depending on whether the inode is live,
958  * snapshotted (read-only), or a virtual ".snap" directory.
959  */
960 int ceph_init_dentry(struct dentry *dentry);
961 
962 
963 /* ioctl.c */
964 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
965 
966 /* export.c */
967 extern const struct export_operations ceph_export_ops;
968 
969 /* locks.c */
970 extern __init void ceph_flock_init(void);
971 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
972 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
973 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
974 extern int ceph_encode_locks_to_buffer(struct inode *inode,
975 				       struct ceph_filelock *flocks,
976 				       int num_fcntl_locks,
977 				       int num_flock_locks);
978 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
979 				  struct ceph_pagelist *pagelist,
980 				  int num_fcntl_locks, int num_flock_locks);
981 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
982 
983 /* debugfs.c */
984 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
985 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
986 
987 #endif /* _FS_CEPH_SUPER_H */
988