xref: /linux/fs/ceph/super.h (revision 4e0ae876f77bc01a7e77724dea57b4b82bd53244)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _FS_CEPH_SUPER_H
3 #define _FS_CEPH_SUPER_H
4 
5 #include <linux/ceph/ceph_debug.h>
6 
7 #include <asm/unaligned.h>
8 #include <linux/backing-dev.h>
9 #include <linux/completion.h>
10 #include <linux/exportfs.h>
11 #include <linux/fs.h>
12 #include <linux/mempool.h>
13 #include <linux/pagemap.h>
14 #include <linux/wait.h>
15 #include <linux/writeback.h>
16 #include <linux/slab.h>
17 #include <linux/posix_acl.h>
18 #include <linux/refcount.h>
19 
20 #include <linux/ceph/libceph.h>
21 
22 #ifdef CONFIG_CEPH_FSCACHE
23 #include <linux/fscache.h>
24 #endif
25 
26 /* f_type in struct statfs */
27 #define CEPH_SUPER_MAGIC 0x00c36400
28 
29 /* large granularity for statfs utilization stats to facilitate
30  * large volume sizes on 32-bit machines. */
31 #define CEPH_BLOCK_SHIFT   22  /* 4 MB */
32 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
33 
34 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
35 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
36 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
37 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
38 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
39 #define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
40 #define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
41 #define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
42 #define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
43 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
44 
45 #define CEPH_MOUNT_OPT_DEFAULT			\
46 	(CEPH_MOUNT_OPT_DCACHE |		\
47 	 CEPH_MOUNT_OPT_NOCOPYFROM)
48 
49 #define ceph_set_mount_opt(fsc, opt) \
50 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
51 #define ceph_test_mount_opt(fsc, opt) \
52 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
53 
54 /* max size of osd read request, limited by libceph */
55 #define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
56 /* osd has a configurable limitaion of max write size.
57  * CEPH_MSG_MAX_DATA_LEN should be small enough. */
58 #define CEPH_MAX_WRITE_SIZE		CEPH_MSG_MAX_DATA_LEN
59 #define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
60 #define CEPH_MAX_READDIR_DEFAULT        1024
61 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
62 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
63 
64 /*
65  * Delay telling the MDS we no longer want caps, in case we reopen
66  * the file.  Delay a minimum amount of time, even if we send a cap
67  * message for some other reason.  Otherwise, take the oppotunity to
68  * update the mds to avoid sending another message later.
69  */
70 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
71 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
72 
73 struct ceph_mount_options {
74 	int flags;
75 	int sb_flags;
76 
77 	int wsize;            /* max write size */
78 	int rsize;            /* max read size */
79 	int rasize;           /* max readahead */
80 	int congestion_kb;    /* max writeback in flight */
81 	int caps_wanted_delay_min, caps_wanted_delay_max;
82 	int caps_max;
83 	int max_readdir;       /* max readdir result (entires) */
84 	int max_readdir_bytes; /* max readdir result (bytes) */
85 
86 	/*
87 	 * everything above this point can be memcmp'd; everything below
88 	 * is handled in compare_mount_options()
89 	 */
90 
91 	char *snapdir_name;   /* default ".snap" */
92 	char *mds_namespace;  /* default NULL */
93 	char *server_path;    /* default  "/" */
94 	char *fscache_uniq;   /* default NULL */
95 };
96 
97 struct ceph_fs_client {
98 	struct super_block *sb;
99 
100 	struct ceph_mount_options *mount_options;
101 	struct ceph_client *client;
102 
103 	unsigned long mount_state;
104 	loff_t max_file_size;
105 
106 	struct ceph_mds_client *mdsc;
107 
108 	/* writeback */
109 	mempool_t *wb_pagevec_pool;
110 	atomic_long_t writeback_count;
111 
112 	struct workqueue_struct *wb_wq;
113 	struct workqueue_struct *pg_inv_wq;
114 	struct workqueue_struct *trunc_wq;
115 	struct workqueue_struct *cap_wq;
116 
117 #ifdef CONFIG_DEBUG_FS
118 	struct dentry *debugfs_dentry_lru, *debugfs_caps;
119 	struct dentry *debugfs_congestion_kb;
120 	struct dentry *debugfs_bdi;
121 	struct dentry *debugfs_mdsc, *debugfs_mdsmap;
122 	struct dentry *debugfs_mds_sessions;
123 #endif
124 
125 #ifdef CONFIG_CEPH_FSCACHE
126 	struct fscache_cookie *fscache;
127 #endif
128 };
129 
130 
131 /*
132  * File i/o capability.  This tracks shared state with the metadata
133  * server that allows us to cache or writeback attributes or to read
134  * and write data.  For any given inode, we should have one or more
135  * capabilities, one issued by each metadata server, and our
136  * cumulative access is the OR of all issued capabilities.
137  *
138  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
139  * session capability lists.
140  */
141 struct ceph_cap {
142 	struct ceph_inode_info *ci;
143 	struct rb_node ci_node;          /* per-ci cap tree */
144 	struct ceph_mds_session *session;
145 	struct list_head session_caps;   /* per-session caplist */
146 	u64 cap_id;       /* unique cap id (mds provided) */
147 	union {
148 		/* in-use caps */
149 		struct {
150 			int issued;       /* latest, from the mds */
151 			int implemented;  /* implemented superset of
152 					     issued (for revocation) */
153 			int mds, mds_wanted;
154 		};
155 		/* caps to release */
156 		struct {
157 			u64 cap_ino;
158 			int queue_release;
159 		};
160 	};
161 	u32 seq, issue_seq, mseq;
162 	u32 cap_gen;      /* active/stale cycle */
163 	unsigned long last_used;
164 	struct list_head caps_item;
165 };
166 
167 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
168 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
169 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
170 
171 struct ceph_cap_flush {
172 	u64 tid;
173 	int caps; /* 0 means capsnap */
174 	bool wake; /* wake up flush waiters when finish ? */
175 	struct list_head g_list; // global
176 	struct list_head i_list; // per inode
177 };
178 
179 /*
180  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
181  * we first complete any in-process sync writes and writeback any dirty
182  * data before flushing the snapped state (tracked here) back to the MDS.
183  */
184 struct ceph_cap_snap {
185 	refcount_t nref;
186 	struct list_head ci_item;
187 
188 	struct ceph_cap_flush cap_flush;
189 
190 	u64 follows;
191 	int issued, dirty;
192 	struct ceph_snap_context *context;
193 
194 	umode_t mode;
195 	kuid_t uid;
196 	kgid_t gid;
197 
198 	struct ceph_buffer *xattr_blob;
199 	u64 xattr_version;
200 
201 	u64 size;
202 	struct timespec64 mtime, atime, ctime;
203 	u64 time_warp_seq;
204 	u64 truncate_size;
205 	u32 truncate_seq;
206 	int writing;   /* a sync write is still in progress */
207 	int dirty_pages;     /* dirty pages awaiting writeback */
208 	bool inline_data;
209 	bool need_flush;
210 };
211 
212 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
213 {
214 	if (refcount_dec_and_test(&capsnap->nref)) {
215 		if (capsnap->xattr_blob)
216 			ceph_buffer_put(capsnap->xattr_blob);
217 		kfree(capsnap);
218 	}
219 }
220 
221 /*
222  * The frag tree describes how a directory is fragmented, potentially across
223  * multiple metadata servers.  It is also used to indicate points where
224  * metadata authority is delegated, and whether/where metadata is replicated.
225  *
226  * A _leaf_ frag will be present in the i_fragtree IFF there is
227  * delegation info.  That is, if mds >= 0 || ndist > 0.
228  */
229 #define CEPH_MAX_DIRFRAG_REP 4
230 
231 struct ceph_inode_frag {
232 	struct rb_node node;
233 
234 	/* fragtree state */
235 	u32 frag;
236 	int split_by;         /* i.e. 2^(split_by) children */
237 
238 	/* delegation and replication info */
239 	int mds;              /* -1 if same authority as parent */
240 	int ndist;            /* >0 if replicated */
241 	int dist[CEPH_MAX_DIRFRAG_REP];
242 };
243 
244 /*
245  * We cache inode xattrs as an encoded blob until they are first used,
246  * at which point we parse them into an rbtree.
247  */
248 struct ceph_inode_xattr {
249 	struct rb_node node;
250 
251 	const char *name;
252 	int name_len;
253 	const char *val;
254 	int val_len;
255 	int dirty;
256 
257 	int should_free_name;
258 	int should_free_val;
259 };
260 
261 /*
262  * Ceph dentry state
263  */
264 struct ceph_dentry_info {
265 	struct dentry *dentry;
266 	struct ceph_mds_session *lease_session;
267 	struct list_head lease_list;
268 	unsigned flags;
269 	int lease_shared_gen;
270 	u32 lease_gen;
271 	u32 lease_seq;
272 	unsigned long lease_renew_after, lease_renew_from;
273 	unsigned long time;
274 	u64 offset;
275 };
276 
277 #define CEPH_DENTRY_REFERENCED		1
278 #define CEPH_DENTRY_LEASE_LIST		2
279 #define CEPH_DENTRY_SHRINK_LIST		4
280 
281 struct ceph_inode_xattrs_info {
282 	/*
283 	 * (still encoded) xattr blob. we avoid the overhead of parsing
284 	 * this until someone actually calls getxattr, etc.
285 	 *
286 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
287 	 * NULL means we don't know.
288 	*/
289 	struct ceph_buffer *blob, *prealloc_blob;
290 
291 	struct rb_root index;
292 	bool dirty;
293 	int count;
294 	int names_size;
295 	int vals_size;
296 	u64 version, index_version;
297 };
298 
299 /*
300  * Ceph inode.
301  */
302 struct ceph_inode_info {
303 	struct ceph_vino i_vino;   /* ceph ino + snap */
304 
305 	spinlock_t i_ceph_lock;
306 
307 	u64 i_version;
308 	u64 i_inline_version;
309 	u32 i_time_warp_seq;
310 
311 	unsigned i_ceph_flags;
312 	atomic64_t i_release_count;
313 	atomic64_t i_ordered_count;
314 	atomic64_t i_complete_seq[2];
315 
316 	struct ceph_dir_layout i_dir_layout;
317 	struct ceph_file_layout i_layout;
318 	char *i_symlink;
319 
320 	/* for dirs */
321 	struct timespec64 i_rctime;
322 	u64 i_rbytes, i_rfiles, i_rsubdirs;
323 	u64 i_files, i_subdirs;
324 
325 	/* quotas */
326 	u64 i_max_bytes, i_max_files;
327 
328 	s32 i_dir_pin;
329 
330 	struct rb_root i_fragtree;
331 	int i_fragtree_nsplits;
332 	struct mutex i_fragtree_mutex;
333 
334 	struct ceph_inode_xattrs_info i_xattrs;
335 
336 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
337 	 * s_mutex. */
338 	struct rb_root i_caps;           /* cap list */
339 	struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
340 	unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
341 	struct list_head i_dirty_item, i_flushing_item;
342 	/* we need to track cap writeback on a per-cap-bit basis, to allow
343 	 * overlapping, pipelined cap flushes to the mds.  we can probably
344 	 * reduce the tid to 8 bits if we're concerned about inode size. */
345 	struct ceph_cap_flush *i_prealloc_cap_flush;
346 	struct list_head i_cap_flush_list;
347 	wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
348 	unsigned long i_hold_caps_min; /* jiffies */
349 	unsigned long i_hold_caps_max; /* jiffies */
350 	struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
351 	struct ceph_cap_reservation i_cap_migration_resv;
352 	struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
353 	struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
354 						    dirty|flushing caps */
355 	unsigned i_snap_caps;           /* cap bits for snapped files */
356 
357 	int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
358 
359 	struct mutex i_truncate_mutex;
360 	u32 i_truncate_seq;        /* last truncate to smaller size */
361 	u64 i_truncate_size;       /*  and the size we last truncated down to */
362 	int i_truncate_pending;    /*  still need to call vmtruncate */
363 
364 	u64 i_max_size;            /* max file size authorized by mds */
365 	u64 i_reported_size; /* (max_)size reported to or requested of mds */
366 	u64 i_wanted_max_size;     /* offset we'd like to write too */
367 	u64 i_requested_max_size;  /* max_size we've requested */
368 
369 	/* held references to caps */
370 	int i_pin_ref;
371 	int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
372 	int i_wrbuffer_ref, i_wrbuffer_ref_head;
373 	atomic_t i_filelock_ref;
374 	atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
375 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
376 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
377 
378 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
379 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
380 	spinlock_t i_unsafe_lock;
381 
382 	union {
383 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
384 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
385 	};
386 	int i_snap_realm_counter; /* snap realm (if caps) */
387 	struct list_head i_snap_realm_item;
388 	struct list_head i_snap_flush_item;
389 
390 	struct work_struct i_wb_work;  /* writeback work */
391 	struct work_struct i_pg_inv_work;  /* page invalidation work */
392 
393 	struct work_struct i_vmtruncate_work;
394 
395 #ifdef CONFIG_CEPH_FSCACHE
396 	struct fscache_cookie *fscache;
397 	u32 i_fscache_gen;
398 #endif
399 	struct inode vfs_inode; /* at end */
400 };
401 
402 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
403 {
404 	return container_of(inode, struct ceph_inode_info, vfs_inode);
405 }
406 
407 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
408 {
409 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
410 }
411 
412 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
413 {
414 	return (struct ceph_fs_client *)sb->s_fs_info;
415 }
416 
417 static inline struct ceph_vino ceph_vino(struct inode *inode)
418 {
419 	return ceph_inode(inode)->i_vino;
420 }
421 
422 /*
423  * ino_t is <64 bits on many architectures, blech.
424  *
425  *               i_ino (kernel inode)   st_ino (userspace)
426  * i386          32                     32
427  * x86_64+ino32  64                     32
428  * x86_64        64                     64
429  */
430 static inline u32 ceph_ino_to_ino32(__u64 vino)
431 {
432 	u32 ino = vino & 0xffffffff;
433 	ino ^= vino >> 32;
434 	if (!ino)
435 		ino = 2;
436 	return ino;
437 }
438 
439 /*
440  * kernel i_ino value
441  */
442 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
443 {
444 #if BITS_PER_LONG == 32
445 	return ceph_ino_to_ino32(vino.ino);
446 #else
447 	return (ino_t)vino.ino;
448 #endif
449 }
450 
451 /*
452  * user-visible ino (stat, filldir)
453  */
454 #if BITS_PER_LONG == 32
455 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
456 {
457 	return ino;
458 }
459 #else
460 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
461 {
462 	if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
463 		ino = ceph_ino_to_ino32(ino);
464 	return ino;
465 }
466 #endif
467 
468 
469 /* for printf-style formatting */
470 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
471 
472 static inline u64 ceph_ino(struct inode *inode)
473 {
474 	return ceph_inode(inode)->i_vino.ino;
475 }
476 static inline u64 ceph_snap(struct inode *inode)
477 {
478 	return ceph_inode(inode)->i_vino.snap;
479 }
480 
481 static inline int ceph_ino_compare(struct inode *inode, void *data)
482 {
483 	struct ceph_vino *pvino = (struct ceph_vino *)data;
484 	struct ceph_inode_info *ci = ceph_inode(inode);
485 	return ci->i_vino.ino == pvino->ino &&
486 		ci->i_vino.snap == pvino->snap;
487 }
488 
489 static inline struct inode *ceph_find_inode(struct super_block *sb,
490 					    struct ceph_vino vino)
491 {
492 	ino_t t = ceph_vino_to_ino(vino);
493 	return ilookup5(sb, t, ceph_ino_compare, &vino);
494 }
495 
496 
497 /*
498  * Ceph inode.
499  */
500 #define CEPH_I_DIR_ORDERED	(1 << 0)  /* dentries in dir are ordered */
501 #define CEPH_I_NODELAY		(1 << 1)  /* do not delay cap release */
502 #define CEPH_I_FLUSH		(1 << 2)  /* do not delay flush of dirty metadata */
503 #define CEPH_I_NOFLUSH		(1 << 3)  /* do not flush dirty caps */
504 #define CEPH_I_POOL_PERM	(1 << 4)  /* pool rd/wr bits are valid */
505 #define CEPH_I_POOL_RD		(1 << 5)  /* can read from pool */
506 #define CEPH_I_POOL_WR		(1 << 6)  /* can write to pool */
507 #define CEPH_I_SEC_INITED	(1 << 7)  /* security initialized */
508 #define CEPH_I_CAP_DROPPED	(1 << 8)  /* caps were forcibly dropped */
509 #define CEPH_I_KICK_FLUSH	(1 << 9)  /* kick flushing caps */
510 #define CEPH_I_FLUSH_SNAPS	(1 << 10) /* need flush snapss */
511 #define CEPH_I_ERROR_WRITE	(1 << 11) /* have seen write errors */
512 #define CEPH_I_ERROR_FILELOCK	(1 << 12) /* have seen file lock errors */
513 
514 
515 /*
516  * We set the ERROR_WRITE bit when we start seeing write errors on an inode
517  * and then clear it when they start succeeding. Note that we do a lockless
518  * check first, and only take the lock if it looks like it needs to be changed.
519  * The write submission code just takes this as a hint, so we're not too
520  * worried if a few slip through in either direction.
521  */
522 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
523 {
524 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
525 		spin_lock(&ci->i_ceph_lock);
526 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
527 		spin_unlock(&ci->i_ceph_lock);
528 	}
529 }
530 
531 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
532 {
533 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
534 		spin_lock(&ci->i_ceph_lock);
535 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
536 		spin_unlock(&ci->i_ceph_lock);
537 	}
538 }
539 
540 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
541 					   long long release_count,
542 					   long long ordered_count)
543 {
544 	smp_mb__before_atomic();
545 	atomic64_set(&ci->i_complete_seq[0], release_count);
546 	atomic64_set(&ci->i_complete_seq[1], ordered_count);
547 }
548 
549 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
550 {
551 	atomic64_inc(&ci->i_release_count);
552 }
553 
554 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
555 {
556 	atomic64_inc(&ci->i_ordered_count);
557 }
558 
559 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
560 {
561 	return atomic64_read(&ci->i_complete_seq[0]) ==
562 		atomic64_read(&ci->i_release_count);
563 }
564 
565 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
566 {
567 	return  atomic64_read(&ci->i_complete_seq[0]) ==
568 		atomic64_read(&ci->i_release_count) &&
569 		atomic64_read(&ci->i_complete_seq[1]) ==
570 		atomic64_read(&ci->i_ordered_count);
571 }
572 
573 static inline void ceph_dir_clear_complete(struct inode *inode)
574 {
575 	__ceph_dir_clear_complete(ceph_inode(inode));
576 }
577 
578 static inline void ceph_dir_clear_ordered(struct inode *inode)
579 {
580 	__ceph_dir_clear_ordered(ceph_inode(inode));
581 }
582 
583 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
584 {
585 	bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
586 	smp_rmb();
587 	return ret;
588 }
589 
590 /* find a specific frag @f */
591 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
592 						u32 f);
593 
594 /*
595  * choose fragment for value @v.  copy frag content to pfrag, if leaf
596  * exists
597  */
598 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
599 			    struct ceph_inode_frag *pfrag,
600 			    int *found);
601 
602 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
603 {
604 	return (struct ceph_dentry_info *)dentry->d_fsdata;
605 }
606 
607 /*
608  * caps helpers
609  */
610 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
611 {
612 	return !RB_EMPTY_ROOT(&ci->i_caps);
613 }
614 
615 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
616 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
617 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
618 				    struct ceph_cap *cap);
619 
620 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
621 {
622 	int issued;
623 	spin_lock(&ci->i_ceph_lock);
624 	issued = __ceph_caps_issued(ci, NULL);
625 	spin_unlock(&ci->i_ceph_lock);
626 	return issued;
627 }
628 
629 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
630 					int touch)
631 {
632 	int r;
633 	spin_lock(&ci->i_ceph_lock);
634 	r = __ceph_caps_issued_mask(ci, mask, touch);
635 	spin_unlock(&ci->i_ceph_lock);
636 	return r;
637 }
638 
639 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
640 {
641 	return ci->i_dirty_caps | ci->i_flushing_caps;
642 }
643 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
644 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
645 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
646 				  struct ceph_cap_flush **pcf);
647 
648 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
649 				      struct ceph_cap *ocap, int mask);
650 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
651 extern int __ceph_caps_used(struct ceph_inode_info *ci);
652 
653 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
654 
655 /*
656  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
657  */
658 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
659 {
660 	int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
661 	if (w & CEPH_CAP_FILE_BUFFER)
662 		w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
663 	return w;
664 }
665 
666 /* what the mds thinks we want */
667 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
668 
669 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
670 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
671 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
672 				     struct ceph_mount_options *fsopt);
673 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
674 			     struct ceph_cap_reservation *ctx, int need);
675 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
676 			       struct ceph_cap_reservation *ctx);
677 extern void ceph_reservation_status(struct ceph_fs_client *client,
678 				    int *total, int *avail, int *used,
679 				    int *reserved, int *min);
680 
681 
682 
683 /*
684  * we keep buffered readdir results attached to file->private_data
685  */
686 #define CEPH_F_SYNC     1
687 #define CEPH_F_ATEND    2
688 
689 struct ceph_file_info {
690 	short fmode;     /* initialized on open */
691 	short flags;     /* CEPH_F_* */
692 
693 	spinlock_t rw_contexts_lock;
694 	struct list_head rw_contexts;
695 };
696 
697 struct ceph_dir_file_info {
698 	struct ceph_file_info file_info;
699 
700 	/* readdir: position within the dir */
701 	u32 frag;
702 	struct ceph_mds_request *last_readdir;
703 
704 	/* readdir: position within a frag */
705 	unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
706 	char *last_name;       /* last entry in previous chunk */
707 	long long dir_release_count;
708 	long long dir_ordered_count;
709 	int readdir_cache_idx;
710 
711 	/* used for -o dirstat read() on directory thing */
712 	char *dir_info;
713 	int dir_info_len;
714 };
715 
716 struct ceph_rw_context {
717 	struct list_head list;
718 	struct task_struct *thread;
719 	int caps;
720 };
721 
722 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps)	\
723 	struct ceph_rw_context _name = {	\
724 		.thread = current,		\
725 		.caps = _caps,			\
726 	}
727 
728 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
729 				       struct ceph_rw_context *ctx)
730 {
731 	spin_lock(&cf->rw_contexts_lock);
732 	list_add(&ctx->list, &cf->rw_contexts);
733 	spin_unlock(&cf->rw_contexts_lock);
734 }
735 
736 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
737 				       struct ceph_rw_context *ctx)
738 {
739 	spin_lock(&cf->rw_contexts_lock);
740 	list_del(&ctx->list);
741 	spin_unlock(&cf->rw_contexts_lock);
742 }
743 
744 static inline struct ceph_rw_context*
745 ceph_find_rw_context(struct ceph_file_info *cf)
746 {
747 	struct ceph_rw_context *ctx, *found = NULL;
748 	spin_lock(&cf->rw_contexts_lock);
749 	list_for_each_entry(ctx, &cf->rw_contexts, list) {
750 		if (ctx->thread == current) {
751 			found = ctx;
752 			break;
753 		}
754 	}
755 	spin_unlock(&cf->rw_contexts_lock);
756 	return found;
757 }
758 
759 struct ceph_readdir_cache_control {
760 	struct page  *page;
761 	struct dentry **dentries;
762 	int index;
763 };
764 
765 /*
766  * A "snap realm" describes a subset of the file hierarchy sharing
767  * the same set of snapshots that apply to it.  The realms themselves
768  * are organized into a hierarchy, such that children inherit (some of)
769  * the snapshots of their parents.
770  *
771  * All inodes within the realm that have capabilities are linked into a
772  * per-realm list.
773  */
774 struct ceph_snap_realm {
775 	u64 ino;
776 	struct inode *inode;
777 	atomic_t nref;
778 	struct rb_node node;
779 
780 	u64 created, seq;
781 	u64 parent_ino;
782 	u64 parent_since;   /* snapid when our current parent became so */
783 
784 	u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
785 	u32 num_prior_parent_snaps;   /*  had prior to parent_since */
786 	u64 *snaps;                   /* snaps specific to this realm */
787 	u32 num_snaps;
788 
789 	struct ceph_snap_realm *parent;
790 	struct list_head children;       /* list of child realms */
791 	struct list_head child_item;
792 
793 	struct list_head empty_item;     /* if i have ref==0 */
794 
795 	struct list_head dirty_item;     /* if realm needs new context */
796 
797 	/* the current set of snaps for this realm */
798 	struct ceph_snap_context *cached_context;
799 
800 	struct list_head inodes_with_caps;
801 	spinlock_t inodes_with_caps_lock;
802 };
803 
804 static inline int default_congestion_kb(void)
805 {
806 	int congestion_kb;
807 
808 	/*
809 	 * Copied from NFS
810 	 *
811 	 * congestion size, scale with available memory.
812 	 *
813 	 *  64MB:    8192k
814 	 * 128MB:   11585k
815 	 * 256MB:   16384k
816 	 * 512MB:   23170k
817 	 *   1GB:   32768k
818 	 *   2GB:   46340k
819 	 *   4GB:   65536k
820 	 *   8GB:   92681k
821 	 *  16GB:  131072k
822 	 *
823 	 * This allows larger machines to have larger/more transfers.
824 	 * Limit the default to 256M
825 	 */
826 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
827 	if (congestion_kb > 256*1024)
828 		congestion_kb = 256*1024;
829 
830 	return congestion_kb;
831 }
832 
833 
834 
835 /* snap.c */
836 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
837 					       u64 ino);
838 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
839 				struct ceph_snap_realm *realm);
840 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
841 				struct ceph_snap_realm *realm);
842 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
843 				  void *p, void *e, bool deletion,
844 				  struct ceph_snap_realm **realm_ret);
845 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
846 			     struct ceph_mds_session *session,
847 			     struct ceph_msg *msg);
848 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
849 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
850 				  struct ceph_cap_snap *capsnap);
851 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
852 
853 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
854 						   u64 snap);
855 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
856 				struct ceph_snapid_map *sm);
857 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
858 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
859 
860 
861 /*
862  * a cap_snap is "pending" if it is still awaiting an in-progress
863  * sync write (that may/may not still update size, mtime, etc.).
864  */
865 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
866 {
867 	return !list_empty(&ci->i_cap_snaps) &&
868 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
869 			       ci_item)->writing;
870 }
871 
872 /* inode.c */
873 extern const struct inode_operations ceph_file_iops;
874 
875 extern struct inode *ceph_alloc_inode(struct super_block *sb);
876 extern void ceph_destroy_inode(struct inode *inode);
877 extern int ceph_drop_inode(struct inode *inode);
878 
879 extern struct inode *ceph_get_inode(struct super_block *sb,
880 				    struct ceph_vino vino);
881 extern struct inode *ceph_get_snapdir(struct inode *parent);
882 extern int ceph_fill_file_size(struct inode *inode, int issued,
883 			       u32 truncate_seq, u64 truncate_size, u64 size);
884 extern void ceph_fill_file_time(struct inode *inode, int issued,
885 				u64 time_warp_seq, struct timespec64 *ctime,
886 				struct timespec64 *mtime,
887 				struct timespec64 *atime);
888 extern int ceph_fill_trace(struct super_block *sb,
889 			   struct ceph_mds_request *req);
890 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
891 				    struct ceph_mds_session *session);
892 
893 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
894 
895 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
896 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
897 extern void ceph_queue_vmtruncate(struct inode *inode);
898 
899 extern void ceph_queue_invalidate(struct inode *inode);
900 extern void ceph_queue_writeback(struct inode *inode);
901 
902 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
903 			     int mask, bool force);
904 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
905 {
906 	return __ceph_do_getattr(inode, NULL, mask, force);
907 }
908 extern int ceph_permission(struct inode *inode, int mask);
909 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
910 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
911 extern int ceph_getattr(const struct path *path, struct kstat *stat,
912 			u32 request_mask, unsigned int flags);
913 
914 /* xattr.c */
915 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
916 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
917 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
918 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
919 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
920 extern void __init ceph_xattr_init(void);
921 extern void ceph_xattr_exit(void);
922 extern const struct xattr_handler *ceph_xattr_handlers[];
923 
924 #ifdef CONFIG_SECURITY
925 extern bool ceph_security_xattr_deadlock(struct inode *in);
926 extern bool ceph_security_xattr_wanted(struct inode *in);
927 #else
928 static inline bool ceph_security_xattr_deadlock(struct inode *in)
929 {
930 	return false;
931 }
932 static inline bool ceph_security_xattr_wanted(struct inode *in)
933 {
934 	return false;
935 }
936 #endif
937 
938 /* acl.c */
939 struct ceph_acls_info {
940 	void *default_acl;
941 	void *acl;
942 	struct ceph_pagelist *pagelist;
943 };
944 
945 #ifdef CONFIG_CEPH_FS_POSIX_ACL
946 
947 struct posix_acl *ceph_get_acl(struct inode *, int);
948 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
949 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
950 		       struct ceph_acls_info *info);
951 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info);
952 void ceph_release_acls_info(struct ceph_acls_info *info);
953 
954 static inline void ceph_forget_all_cached_acls(struct inode *inode)
955 {
956        forget_all_cached_acls(inode);
957 }
958 
959 #else
960 
961 #define ceph_get_acl NULL
962 #define ceph_set_acl NULL
963 
964 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
965 				     struct ceph_acls_info *info)
966 {
967 	return 0;
968 }
969 static inline void ceph_init_inode_acls(struct inode *inode,
970 					struct ceph_acls_info *info)
971 {
972 }
973 static inline void ceph_release_acls_info(struct ceph_acls_info *info)
974 {
975 }
976 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
977 {
978 	return 0;
979 }
980 
981 static inline void ceph_forget_all_cached_acls(struct inode *inode)
982 {
983 }
984 
985 #endif
986 
987 /* caps.c */
988 extern const char *ceph_cap_string(int c);
989 extern void ceph_handle_caps(struct ceph_mds_session *session,
990 			     struct ceph_msg *msg);
991 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
992 				     struct ceph_cap_reservation *ctx);
993 extern void ceph_add_cap(struct inode *inode,
994 			 struct ceph_mds_session *session, u64 cap_id,
995 			 int fmode, unsigned issued, unsigned wanted,
996 			 unsigned cap, unsigned seq, u64 realmino, int flags,
997 			 struct ceph_cap **new_cap);
998 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
999 extern void __ceph_remove_caps(struct inode* inode);
1000 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1001 			 struct ceph_cap *cap);
1002 extern int ceph_is_any_caps(struct inode *inode);
1003 
1004 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1005 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1006 		      int datasync);
1007 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1008 					  struct ceph_mds_session *session);
1009 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1010 				    struct ceph_mds_session *session);
1011 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1012 					     int mds);
1013 extern int ceph_get_cap_mds(struct inode *inode);
1014 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1015 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1016 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1017 				       struct ceph_snap_context *snapc);
1018 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1019 			     struct ceph_mds_session **psession);
1020 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1021 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1022 			    struct ceph_mds_session *session);
1023 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1024 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1025 extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1026 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1027 				     int mds, int drop, int unless, int force);
1028 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1029 				      struct inode *dir,
1030 				      int mds, int drop, int unless);
1031 
1032 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
1033 			 loff_t endoff, int *got, struct page **pinned_page);
1034 extern int ceph_try_get_caps(struct ceph_inode_info *ci,
1035 			     int need, int want, bool nonblock, int *got);
1036 
1037 /* for counting open files by mode */
1038 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode);
1039 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
1040 
1041 /* addr.c */
1042 extern const struct address_space_operations ceph_aops;
1043 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1044 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1045 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need);
1046 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1047 
1048 /* file.c */
1049 extern const struct file_operations ceph_file_fops;
1050 
1051 extern int ceph_renew_caps(struct inode *inode);
1052 extern int ceph_open(struct inode *inode, struct file *file);
1053 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1054 			    struct file *file, unsigned flags, umode_t mode);
1055 extern int ceph_release(struct inode *inode, struct file *filp);
1056 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1057 				  char *data, size_t len);
1058 
1059 /* dir.c */
1060 extern const struct file_operations ceph_dir_fops;
1061 extern const struct file_operations ceph_snapdir_fops;
1062 extern const struct inode_operations ceph_dir_iops;
1063 extern const struct inode_operations ceph_snapdir_iops;
1064 extern const struct dentry_operations ceph_dentry_ops;
1065 
1066 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1067 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1068 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1069 			       struct dentry *dentry, int err);
1070 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1071 					 struct dentry *dentry, int err);
1072 
1073 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1074 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1075 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1076 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1077 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1078 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1079 
1080 /* ioctl.c */
1081 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1082 
1083 /* export.c */
1084 extern const struct export_operations ceph_export_ops;
1085 
1086 /* locks.c */
1087 extern __init void ceph_flock_init(void);
1088 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1089 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1090 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1091 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1092 				       struct ceph_filelock *flocks,
1093 				       int num_fcntl_locks,
1094 				       int num_flock_locks);
1095 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1096 				  struct ceph_pagelist *pagelist,
1097 				  int num_fcntl_locks, int num_flock_locks);
1098 
1099 /* debugfs.c */
1100 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
1101 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1102 
1103 /* quota.c */
1104 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1105 {
1106 	return ci->i_max_files || ci->i_max_bytes;
1107 }
1108 
1109 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1110 
1111 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1112 				       u64 max_bytes, u64 max_files)
1113 {
1114 	bool had_quota, has_quota;
1115 	had_quota = __ceph_has_any_quota(ci);
1116 	ci->i_max_bytes = max_bytes;
1117 	ci->i_max_files = max_files;
1118 	has_quota = __ceph_has_any_quota(ci);
1119 
1120 	if (had_quota != has_quota)
1121 		ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1122 }
1123 
1124 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1125 			      struct ceph_mds_session *session,
1126 			      struct ceph_msg *msg);
1127 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1128 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1129 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1130 					     loff_t newlen);
1131 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1132 						loff_t newlen);
1133 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1134 				     struct kstatfs *buf);
1135 
1136 #endif /* _FS_CEPH_SUPER_H */
1137