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 #include <linux/security.h>
20
21 #include <linux/ceph/libceph.h>
22
23 #ifdef CONFIG_CEPH_FSCACHE
24 #include <linux/fscache.h>
25 #endif
26
27 /* f_type in struct statfs */
28 #define CEPH_SUPER_MAGIC 0x00c36400
29
30 /* large granularity for statfs utilization stats to facilitate
31 * large volume sizes on 32-bit machines. */
32 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */
33 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
34
35 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */
36 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
37 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
38 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
39 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
40 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
41 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
42 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
43 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */
44 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */
45 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */
46 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */
47
48 #define CEPH_MOUNT_OPT_DEFAULT \
49 (CEPH_MOUNT_OPT_DCACHE | \
50 CEPH_MOUNT_OPT_NOCOPYFROM)
51
52 #define ceph_set_mount_opt(fsc, opt) \
53 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
54 #define ceph_clear_mount_opt(fsc, opt) \
55 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
56 #define ceph_test_mount_opt(fsc, opt) \
57 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
58
59 /* max size of osd read request, limited by libceph */
60 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
61 /* osd has a configurable limitaion of max write size.
62 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
63 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN
64 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
65 #define CEPH_MAX_READDIR_DEFAULT 1024
66 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
67 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
68
69 /*
70 * Delay telling the MDS we no longer want caps, in case we reopen
71 * the file. Delay a minimum amount of time, even if we send a cap
72 * message for some other reason. Otherwise, take the oppotunity to
73 * update the mds to avoid sending another message later.
74 */
75 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
76 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
77
78 struct ceph_mount_options {
79 unsigned int flags;
80
81 unsigned int wsize; /* max write size */
82 unsigned int rsize; /* max read size */
83 unsigned int rasize; /* max readahead */
84 unsigned int congestion_kb; /* max writeback in flight */
85 unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
86 int caps_max;
87 unsigned int max_readdir; /* max readdir result (entries) */
88 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
89
90 /*
91 * everything above this point can be memcmp'd; everything below
92 * is handled in compare_mount_options()
93 */
94
95 char *snapdir_name; /* default ".snap" */
96 char *mds_namespace; /* default NULL */
97 char *server_path; /* default NULL (means "/") */
98 char *fscache_uniq; /* default NULL */
99 };
100
101 struct ceph_fs_client {
102 struct super_block *sb;
103
104 struct list_head metric_wakeup;
105
106 struct ceph_mount_options *mount_options;
107 struct ceph_client *client;
108
109 unsigned long mount_state;
110
111 unsigned long last_auto_reconnect;
112 bool blocklisted;
113
114 bool have_copy_from2;
115
116 u32 filp_gen;
117 loff_t max_file_size;
118
119 struct ceph_mds_client *mdsc;
120
121 atomic_long_t writeback_count;
122
123 struct workqueue_struct *inode_wq;
124 struct workqueue_struct *cap_wq;
125
126 #ifdef CONFIG_DEBUG_FS
127 struct dentry *debugfs_dentry_lru, *debugfs_caps;
128 struct dentry *debugfs_congestion_kb;
129 struct dentry *debugfs_bdi;
130 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
131 struct dentry *debugfs_metric;
132 struct dentry *debugfs_mds_sessions;
133 #endif
134
135 #ifdef CONFIG_CEPH_FSCACHE
136 struct fscache_cookie *fscache;
137 #endif
138 };
139
140
141 /*
142 * File i/o capability. This tracks shared state with the metadata
143 * server that allows us to cache or writeback attributes or to read
144 * and write data. For any given inode, we should have one or more
145 * capabilities, one issued by each metadata server, and our
146 * cumulative access is the OR of all issued capabilities.
147 *
148 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
149 * session capability lists.
150 */
151 struct ceph_cap {
152 struct ceph_inode_info *ci;
153 struct rb_node ci_node; /* per-ci cap tree */
154 struct ceph_mds_session *session;
155 struct list_head session_caps; /* per-session caplist */
156 u64 cap_id; /* unique cap id (mds provided) */
157 union {
158 /* in-use caps */
159 struct {
160 int issued; /* latest, from the mds */
161 int implemented; /* implemented superset of
162 issued (for revocation) */
163 int mds; /* mds index for this cap */
164 int mds_wanted; /* caps wanted from this mds */
165 };
166 /* caps to release */
167 struct {
168 u64 cap_ino;
169 int queue_release;
170 };
171 };
172 u32 seq, issue_seq, mseq;
173 u32 cap_gen; /* active/stale cycle */
174 unsigned long last_used;
175 struct list_head caps_item;
176 };
177
178 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */
179 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */
180 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */
181
182 struct ceph_cap_flush {
183 u64 tid;
184 int caps;
185 bool wake; /* wake up flush waiters when finish ? */
186 bool is_capsnap; /* true means capsnap */
187 struct list_head g_list; // global
188 struct list_head i_list; // per inode
189 };
190
191 /*
192 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
193 * we first complete any in-process sync writes and writeback any dirty
194 * data before flushing the snapped state (tracked here) back to the MDS.
195 */
196 struct ceph_cap_snap {
197 refcount_t nref;
198 struct list_head ci_item;
199
200 struct ceph_cap_flush cap_flush;
201
202 u64 follows;
203 int issued, dirty;
204 struct ceph_snap_context *context;
205
206 umode_t mode;
207 kuid_t uid;
208 kgid_t gid;
209
210 struct ceph_buffer *xattr_blob;
211 u64 xattr_version;
212
213 u64 size;
214 u64 change_attr;
215 struct timespec64 mtime, atime, ctime, btime;
216 u64 time_warp_seq;
217 u64 truncate_size;
218 u32 truncate_seq;
219 int writing; /* a sync write is still in progress */
220 int dirty_pages; /* dirty pages awaiting writeback */
221 bool inline_data;
222 bool need_flush;
223 };
224
ceph_put_cap_snap(struct ceph_cap_snap * capsnap)225 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
226 {
227 if (refcount_dec_and_test(&capsnap->nref)) {
228 if (capsnap->xattr_blob)
229 ceph_buffer_put(capsnap->xattr_blob);
230 kfree(capsnap);
231 }
232 }
233
234 /*
235 * The frag tree describes how a directory is fragmented, potentially across
236 * multiple metadata servers. It is also used to indicate points where
237 * metadata authority is delegated, and whether/where metadata is replicated.
238 *
239 * A _leaf_ frag will be present in the i_fragtree IFF there is
240 * delegation info. That is, if mds >= 0 || ndist > 0.
241 */
242 #define CEPH_MAX_DIRFRAG_REP 4
243
244 struct ceph_inode_frag {
245 struct rb_node node;
246
247 /* fragtree state */
248 u32 frag;
249 int split_by; /* i.e. 2^(split_by) children */
250
251 /* delegation and replication info */
252 int mds; /* -1 if same authority as parent */
253 int ndist; /* >0 if replicated */
254 int dist[CEPH_MAX_DIRFRAG_REP];
255 };
256
257 /*
258 * We cache inode xattrs as an encoded blob until they are first used,
259 * at which point we parse them into an rbtree.
260 */
261 struct ceph_inode_xattr {
262 struct rb_node node;
263
264 const char *name;
265 int name_len;
266 const char *val;
267 int val_len;
268 int dirty;
269
270 int should_free_name;
271 int should_free_val;
272 };
273
274 /*
275 * Ceph dentry state
276 */
277 struct ceph_dentry_info {
278 struct dentry *dentry;
279 struct ceph_mds_session *lease_session;
280 struct list_head lease_list;
281 unsigned flags;
282 int lease_shared_gen;
283 u32 lease_gen;
284 u32 lease_seq;
285 unsigned long lease_renew_after, lease_renew_from;
286 unsigned long time;
287 u64 offset;
288 };
289
290 #define CEPH_DENTRY_REFERENCED 1
291 #define CEPH_DENTRY_LEASE_LIST 2
292 #define CEPH_DENTRY_SHRINK_LIST 4
293 #define CEPH_DENTRY_PRIMARY_LINK 8
294
295 struct ceph_inode_xattrs_info {
296 /*
297 * (still encoded) xattr blob. we avoid the overhead of parsing
298 * this until someone actually calls getxattr, etc.
299 *
300 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
301 * NULL means we don't know.
302 */
303 struct ceph_buffer *blob, *prealloc_blob;
304
305 struct rb_root index;
306 bool dirty;
307 int count;
308 int names_size;
309 int vals_size;
310 u64 version, index_version;
311 };
312
313 /*
314 * Ceph inode.
315 */
316 struct ceph_inode_info {
317 struct ceph_vino i_vino; /* ceph ino + snap */
318
319 spinlock_t i_ceph_lock;
320
321 u64 i_version;
322 u64 i_inline_version;
323 u32 i_time_warp_seq;
324
325 unsigned long i_ceph_flags;
326 atomic64_t i_release_count;
327 atomic64_t i_ordered_count;
328 atomic64_t i_complete_seq[2];
329
330 struct ceph_dir_layout i_dir_layout;
331 struct ceph_file_layout i_layout;
332 struct ceph_file_layout i_cached_layout; // for async creates
333 char *i_symlink;
334
335 /* for dirs */
336 struct timespec64 i_rctime;
337 u64 i_rbytes, i_rfiles, i_rsubdirs;
338 u64 i_files, i_subdirs;
339
340 /* quotas */
341 u64 i_max_bytes, i_max_files;
342
343 s32 i_dir_pin;
344
345 struct rb_root i_fragtree;
346 int i_fragtree_nsplits;
347 struct mutex i_fragtree_mutex;
348
349 struct ceph_inode_xattrs_info i_xattrs;
350
351 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
352 * s_mutex. */
353 struct rb_root i_caps; /* cap list */
354 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
355 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
356
357 /*
358 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
359 * is protected by the mdsc->cap_dirty_lock, but each individual item
360 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
361 * requires the mdsc->cap_dirty_lock. List presence for an item can
362 * be tested under the i_ceph_lock. Changing anything requires both.
363 */
364 struct list_head i_dirty_item;
365
366 /*
367 * Link to session's s_cap_flushing list. Protected in a similar
368 * fashion to i_dirty_item, but also by the s_mutex for changes. The
369 * s_cap_flushing list can be walked while holding either the s_mutex
370 * or msdc->cap_dirty_lock. List presence can also be checked while
371 * holding the i_ceph_lock for this inode.
372 */
373 struct list_head i_flushing_item;
374
375 /* we need to track cap writeback on a per-cap-bit basis, to allow
376 * overlapping, pipelined cap flushes to the mds. we can probably
377 * reduce the tid to 8 bits if we're concerned about inode size. */
378 struct ceph_cap_flush *i_prealloc_cap_flush;
379 struct list_head i_cap_flush_list;
380 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
381 unsigned long i_hold_caps_max; /* jiffies */
382 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
383 struct ceph_cap_reservation i_cap_migration_resv;
384 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
385 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
386 dirty|flushing caps */
387 unsigned i_snap_caps; /* cap bits for snapped files */
388
389 unsigned long i_last_rd;
390 unsigned long i_last_wr;
391 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
392
393 struct mutex i_truncate_mutex;
394 u32 i_truncate_seq; /* last truncate to smaller size */
395 u64 i_truncate_size; /* and the size we last truncated down to */
396 int i_truncate_pending; /* still need to call vmtruncate */
397
398 u64 i_max_size; /* max file size authorized by mds */
399 u64 i_reported_size; /* (max_)size reported to or requested of mds */
400 u64 i_wanted_max_size; /* offset we'd like to write too */
401 u64 i_requested_max_size; /* max_size we've requested */
402
403 /* held references to caps */
404 int i_pin_ref;
405 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
406 int i_wrbuffer_ref, i_wrbuffer_ref_head;
407 atomic_t i_filelock_ref;
408 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
409 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
410 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
411
412 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
413 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
414 spinlock_t i_unsafe_lock;
415
416 union {
417 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
418 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
419 };
420 int i_snap_realm_counter; /* snap realm (if caps) */
421 struct list_head i_snap_realm_item;
422 struct list_head i_snap_flush_item;
423 struct timespec64 i_btime;
424 struct timespec64 i_snap_btime;
425
426 struct work_struct i_work;
427 unsigned long i_work_mask;
428
429 #ifdef CONFIG_CEPH_FSCACHE
430 struct fscache_cookie *fscache;
431 u32 i_fscache_gen;
432 #endif
433 struct inode vfs_inode; /* at end */
434 };
435
436 static inline struct ceph_inode_info *
ceph_inode(const struct inode * inode)437 ceph_inode(const struct inode *inode)
438 {
439 return container_of(inode, struct ceph_inode_info, vfs_inode);
440 }
441
442 static inline struct ceph_fs_client *
ceph_inode_to_client(const struct inode * inode)443 ceph_inode_to_client(const struct inode *inode)
444 {
445 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
446 }
447
448 static inline struct ceph_fs_client *
ceph_sb_to_client(const struct super_block * sb)449 ceph_sb_to_client(const struct super_block *sb)
450 {
451 return (struct ceph_fs_client *)sb->s_fs_info;
452 }
453
454 static inline struct ceph_mds_client *
ceph_sb_to_mdsc(const struct super_block * sb)455 ceph_sb_to_mdsc(const struct super_block *sb)
456 {
457 return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
458 }
459
460 static inline struct ceph_vino
ceph_vino(const struct inode * inode)461 ceph_vino(const struct inode *inode)
462 {
463 return ceph_inode(inode)->i_vino;
464 }
465
ceph_ino_to_ino32(u64 vino)466 static inline u32 ceph_ino_to_ino32(u64 vino)
467 {
468 u32 ino = vino & 0xffffffff;
469 ino ^= vino >> 32;
470 if (!ino)
471 ino = 2;
472 return ino;
473 }
474
475 /*
476 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
477 * some arches. We generally do not use this value inside the ceph driver, but
478 * we do want to set it to something, so that generic vfs code has an
479 * appropriate value for tracepoints and the like.
480 */
ceph_vino_to_ino_t(struct ceph_vino vino)481 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
482 {
483 if (sizeof(ino_t) == sizeof(u32))
484 return ceph_ino_to_ino32(vino.ino);
485 return (ino_t)vino.ino;
486 }
487
488 /* for printf-style formatting */
489 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
490
ceph_ino(struct inode * inode)491 static inline u64 ceph_ino(struct inode *inode)
492 {
493 return ceph_inode(inode)->i_vino.ino;
494 }
495
ceph_snap(struct inode * inode)496 static inline u64 ceph_snap(struct inode *inode)
497 {
498 return ceph_inode(inode)->i_vino.snap;
499 }
500
501 /**
502 * ceph_present_ino - format an inode number for presentation to userland
503 * @sb: superblock where the inode lives
504 * @ino: inode number to (possibly) convert
505 *
506 * If the user mounted with the ino32 option, then the 64-bit value needs
507 * to be converted to something that can fit inside 32 bits. Note that
508 * internal kernel code never uses this value, so this is entirely for
509 * userland consumption.
510 */
ceph_present_ino(struct super_block * sb,u64 ino)511 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
512 {
513 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
514 return ceph_ino_to_ino32(ino);
515 return ino;
516 }
517
ceph_present_inode(struct inode * inode)518 static inline u64 ceph_present_inode(struct inode *inode)
519 {
520 return ceph_present_ino(inode->i_sb, ceph_ino(inode));
521 }
522
ceph_ino_compare(struct inode * inode,void * data)523 static inline int ceph_ino_compare(struct inode *inode, void *data)
524 {
525 struct ceph_vino *pvino = (struct ceph_vino *)data;
526 struct ceph_inode_info *ci = ceph_inode(inode);
527 return ci->i_vino.ino == pvino->ino &&
528 ci->i_vino.snap == pvino->snap;
529 }
530
531 /*
532 * The MDS reserves a set of inodes for its own usage. These should never
533 * be accessible by clients, and so the MDS has no reason to ever hand these
534 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
535 *
536 * These come from src/mds/mdstypes.h in the ceph sources.
537 */
538 #define CEPH_MAX_MDS 0x100
539 #define CEPH_NUM_STRAY 10
540 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS)
541 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
542
ceph_vino_is_reserved(const struct ceph_vino vino)543 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
544 {
545 if (vino.ino < CEPH_INO_SYSTEM_BASE &&
546 vino.ino >= CEPH_MDS_INO_MDSDIR_OFFSET) {
547 WARN_RATELIMIT(1, "Attempt to access reserved inode number 0x%llx", vino.ino);
548 return true;
549 }
550 return false;
551 }
552
ceph_find_inode(struct super_block * sb,struct ceph_vino vino)553 static inline struct inode *ceph_find_inode(struct super_block *sb,
554 struct ceph_vino vino)
555 {
556 if (ceph_vino_is_reserved(vino))
557 return NULL;
558
559 /*
560 * NB: The hashval will be run through the fs/inode.c hash function
561 * anyway, so there is no need to squash the inode number down to
562 * 32-bits first. Just use low-order bits on arches with 32-bit long.
563 */
564 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
565 }
566
567
568 /*
569 * Ceph inode.
570 */
571 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
572 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
573 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */
574 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */
575 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */
576 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */
577 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */
578 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */
579 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */
580 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */
581 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */
582 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */
583 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT)
584
585 /*
586 * Masks of ceph inode work.
587 */
588 #define CEPH_I_WORK_WRITEBACK 0 /* writeback */
589 #define CEPH_I_WORK_INVALIDATE_PAGES 1 /* invalidate pages */
590 #define CEPH_I_WORK_VMTRUNCATE 2 /* vmtruncate */
591
592 /*
593 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
594 * and then clear it when they start succeeding. Note that we do a lockless
595 * check first, and only take the lock if it looks like it needs to be changed.
596 * The write submission code just takes this as a hint, so we're not too
597 * worried if a few slip through in either direction.
598 */
ceph_set_error_write(struct ceph_inode_info * ci)599 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
600 {
601 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
602 spin_lock(&ci->i_ceph_lock);
603 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
604 spin_unlock(&ci->i_ceph_lock);
605 }
606 }
607
ceph_clear_error_write(struct ceph_inode_info * ci)608 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
609 {
610 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
611 spin_lock(&ci->i_ceph_lock);
612 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
613 spin_unlock(&ci->i_ceph_lock);
614 }
615 }
616
__ceph_dir_set_complete(struct ceph_inode_info * ci,long long release_count,long long ordered_count)617 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
618 long long release_count,
619 long long ordered_count)
620 {
621 /*
622 * Makes sure operations that setup readdir cache (update page
623 * cache and i_size) are strongly ordered w.r.t. the following
624 * atomic64_set() operations.
625 */
626 smp_mb();
627 atomic64_set(&ci->i_complete_seq[0], release_count);
628 atomic64_set(&ci->i_complete_seq[1], ordered_count);
629 }
630
__ceph_dir_clear_complete(struct ceph_inode_info * ci)631 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
632 {
633 atomic64_inc(&ci->i_release_count);
634 }
635
__ceph_dir_clear_ordered(struct ceph_inode_info * ci)636 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
637 {
638 atomic64_inc(&ci->i_ordered_count);
639 }
640
__ceph_dir_is_complete(struct ceph_inode_info * ci)641 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
642 {
643 return atomic64_read(&ci->i_complete_seq[0]) ==
644 atomic64_read(&ci->i_release_count);
645 }
646
__ceph_dir_is_complete_ordered(struct ceph_inode_info * ci)647 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
648 {
649 return atomic64_read(&ci->i_complete_seq[0]) ==
650 atomic64_read(&ci->i_release_count) &&
651 atomic64_read(&ci->i_complete_seq[1]) ==
652 atomic64_read(&ci->i_ordered_count);
653 }
654
ceph_dir_clear_complete(struct inode * inode)655 static inline void ceph_dir_clear_complete(struct inode *inode)
656 {
657 __ceph_dir_clear_complete(ceph_inode(inode));
658 }
659
ceph_dir_clear_ordered(struct inode * inode)660 static inline void ceph_dir_clear_ordered(struct inode *inode)
661 {
662 __ceph_dir_clear_ordered(ceph_inode(inode));
663 }
664
ceph_dir_is_complete_ordered(struct inode * inode)665 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
666 {
667 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
668 smp_rmb();
669 return ret;
670 }
671
672 /* find a specific frag @f */
673 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
674 u32 f);
675
676 /*
677 * choose fragment for value @v. copy frag content to pfrag, if leaf
678 * exists
679 */
680 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
681 struct ceph_inode_frag *pfrag,
682 int *found);
683
ceph_dentry(const struct dentry * dentry)684 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
685 {
686 return (struct ceph_dentry_info *)dentry->d_fsdata;
687 }
688
689 /*
690 * caps helpers
691 */
__ceph_is_any_real_caps(struct ceph_inode_info * ci)692 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
693 {
694 return !RB_EMPTY_ROOT(&ci->i_caps);
695 }
696
697 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
698 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
699 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
700 int t);
701 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
702 struct ceph_cap *cap);
703
ceph_caps_issued(struct ceph_inode_info * ci)704 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
705 {
706 int issued;
707 spin_lock(&ci->i_ceph_lock);
708 issued = __ceph_caps_issued(ci, NULL);
709 spin_unlock(&ci->i_ceph_lock);
710 return issued;
711 }
712
ceph_caps_issued_mask_metric(struct ceph_inode_info * ci,int mask,int touch)713 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
714 int mask, int touch)
715 {
716 int r;
717 spin_lock(&ci->i_ceph_lock);
718 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
719 spin_unlock(&ci->i_ceph_lock);
720 return r;
721 }
722
__ceph_caps_dirty(struct ceph_inode_info * ci)723 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
724 {
725 return ci->i_dirty_caps | ci->i_flushing_caps;
726 }
727 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
728 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
729 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
730 struct ceph_cap_flush **pcf);
731
732 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
733 struct ceph_cap *ocap, int mask);
734 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
735 extern int __ceph_caps_used(struct ceph_inode_info *ci);
736
__ceph_is_file_opened(struct ceph_inode_info * ci)737 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
738 {
739 return ci->i_nr_by_mode[0];
740 }
741 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
742 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
743
744 /* what the mds thinks we want */
745 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
746
747 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
748 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
749 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
750 struct ceph_mount_options *fsopt);
751 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
752 struct ceph_cap_reservation *ctx, int need);
753 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
754 struct ceph_cap_reservation *ctx);
755 extern void ceph_reservation_status(struct ceph_fs_client *client,
756 int *total, int *avail, int *used,
757 int *reserved, int *min);
758
759
760
761 /*
762 * we keep buffered readdir results attached to file->private_data
763 */
764 #define CEPH_F_SYNC 1
765 #define CEPH_F_ATEND 2
766
767 struct ceph_file_info {
768 short fmode; /* initialized on open */
769 short flags; /* CEPH_F_* */
770
771 spinlock_t rw_contexts_lock;
772 struct list_head rw_contexts;
773
774 u32 filp_gen;
775 atomic_t num_locks;
776 };
777
778 struct ceph_dir_file_info {
779 struct ceph_file_info file_info;
780
781 /* readdir: position within the dir */
782 u32 frag;
783 struct ceph_mds_request *last_readdir;
784
785 /* readdir: position within a frag */
786 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
787 char *last_name; /* last entry in previous chunk */
788 long long dir_release_count;
789 long long dir_ordered_count;
790 int readdir_cache_idx;
791
792 /* used for -o dirstat read() on directory thing */
793 char *dir_info;
794 int dir_info_len;
795 };
796
797 struct ceph_rw_context {
798 struct list_head list;
799 struct task_struct *thread;
800 int caps;
801 };
802
803 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
804 struct ceph_rw_context _name = { \
805 .thread = current, \
806 .caps = _caps, \
807 }
808
ceph_add_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)809 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
810 struct ceph_rw_context *ctx)
811 {
812 spin_lock(&cf->rw_contexts_lock);
813 list_add(&ctx->list, &cf->rw_contexts);
814 spin_unlock(&cf->rw_contexts_lock);
815 }
816
ceph_del_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)817 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
818 struct ceph_rw_context *ctx)
819 {
820 spin_lock(&cf->rw_contexts_lock);
821 list_del(&ctx->list);
822 spin_unlock(&cf->rw_contexts_lock);
823 }
824
825 static inline struct ceph_rw_context*
ceph_find_rw_context(struct ceph_file_info * cf)826 ceph_find_rw_context(struct ceph_file_info *cf)
827 {
828 struct ceph_rw_context *ctx, *found = NULL;
829 spin_lock(&cf->rw_contexts_lock);
830 list_for_each_entry(ctx, &cf->rw_contexts, list) {
831 if (ctx->thread == current) {
832 found = ctx;
833 break;
834 }
835 }
836 spin_unlock(&cf->rw_contexts_lock);
837 return found;
838 }
839
840 struct ceph_readdir_cache_control {
841 struct page *page;
842 struct dentry **dentries;
843 int index;
844 };
845
846 /*
847 * A "snap realm" describes a subset of the file hierarchy sharing
848 * the same set of snapshots that apply to it. The realms themselves
849 * are organized into a hierarchy, such that children inherit (some of)
850 * the snapshots of their parents.
851 *
852 * All inodes within the realm that have capabilities are linked into a
853 * per-realm list.
854 */
855 struct ceph_snap_realm {
856 u64 ino;
857 struct inode *inode;
858 atomic_t nref;
859 struct rb_node node;
860
861 u64 created, seq;
862 u64 parent_ino;
863 u64 parent_since; /* snapid when our current parent became so */
864
865 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
866 u32 num_prior_parent_snaps; /* had prior to parent_since */
867 u64 *snaps; /* snaps specific to this realm */
868 u32 num_snaps;
869
870 struct ceph_snap_realm *parent;
871 struct list_head children; /* list of child realms */
872 struct list_head child_item;
873
874 struct list_head empty_item; /* if i have ref==0 */
875
876 struct list_head dirty_item; /* if realm needs new context */
877
878 /* the current set of snaps for this realm */
879 struct ceph_snap_context *cached_context;
880
881 struct list_head inodes_with_caps;
882 spinlock_t inodes_with_caps_lock;
883 };
884
default_congestion_kb(void)885 static inline int default_congestion_kb(void)
886 {
887 int congestion_kb;
888
889 /*
890 * Copied from NFS
891 *
892 * congestion size, scale with available memory.
893 *
894 * 64MB: 8192k
895 * 128MB: 11585k
896 * 256MB: 16384k
897 * 512MB: 23170k
898 * 1GB: 32768k
899 * 2GB: 46340k
900 * 4GB: 65536k
901 * 8GB: 92681k
902 * 16GB: 131072k
903 *
904 * This allows larger machines to have larger/more transfers.
905 * Limit the default to 256M
906 */
907 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
908 if (congestion_kb > 256*1024)
909 congestion_kb = 256*1024;
910
911 return congestion_kb;
912 }
913
914
915 /* super.c */
916 extern int ceph_force_reconnect(struct super_block *sb);
917 /* snap.c */
918 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
919 u64 ino);
920 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
921 struct ceph_snap_realm *realm);
922 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
923 struct ceph_snap_realm *realm);
924 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
925 void *p, void *e, bool deletion,
926 struct ceph_snap_realm **realm_ret);
927 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
928 struct ceph_mds_session *session,
929 struct ceph_msg *msg);
930 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
931 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
932 struct ceph_cap_snap *capsnap);
933 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
934
935 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
936 u64 snap);
937 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
938 struct ceph_snapid_map *sm);
939 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
940 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
941
942
943 /*
944 * a cap_snap is "pending" if it is still awaiting an in-progress
945 * sync write (that may/may not still update size, mtime, etc.).
946 */
__ceph_have_pending_cap_snap(struct ceph_inode_info * ci)947 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
948 {
949 return !list_empty(&ci->i_cap_snaps) &&
950 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
951 ci_item)->writing;
952 }
953
954 /* inode.c */
955 struct ceph_mds_reply_info_in;
956 struct ceph_mds_reply_dirfrag;
957
958 extern const struct inode_operations ceph_file_iops;
959
960 extern struct inode *ceph_alloc_inode(struct super_block *sb);
961 extern void ceph_evict_inode(struct inode *inode);
962 extern void ceph_free_inode(struct inode *inode);
963
964 extern struct inode *ceph_get_inode(struct super_block *sb,
965 struct ceph_vino vino);
966 extern struct inode *ceph_get_snapdir(struct inode *parent);
967 extern int ceph_fill_file_size(struct inode *inode, int issued,
968 u32 truncate_seq, u64 truncate_size, u64 size);
969 extern void ceph_fill_file_time(struct inode *inode, int issued,
970 u64 time_warp_seq, struct timespec64 *ctime,
971 struct timespec64 *mtime,
972 struct timespec64 *atime);
973 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
974 struct ceph_mds_reply_info_in *iinfo,
975 struct ceph_mds_reply_dirfrag *dirinfo,
976 struct ceph_mds_session *session, int cap_fmode,
977 struct ceph_cap_reservation *caps_reservation);
978 extern int ceph_fill_trace(struct super_block *sb,
979 struct ceph_mds_request *req);
980 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
981 struct ceph_mds_session *session);
982
983 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
984
985 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
986 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
987 extern void ceph_queue_vmtruncate(struct inode *inode);
988 extern void ceph_queue_invalidate(struct inode *inode);
989 extern void ceph_queue_writeback(struct inode *inode);
990 extern void ceph_async_iput(struct inode *inode);
991
992 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
993 int mask, bool force);
ceph_do_getattr(struct inode * inode,int mask,bool force)994 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
995 {
996 return __ceph_do_getattr(inode, NULL, mask, force);
997 }
998 extern int ceph_permission(struct inode *inode, int mask);
999 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1000 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
1001 extern int ceph_getattr(const struct path *path, struct kstat *stat,
1002 u32 request_mask, unsigned int flags);
1003
1004 /* xattr.c */
1005 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1006 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1007 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1008 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1009 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1010 extern const struct xattr_handler *ceph_xattr_handlers[];
1011
1012 struct ceph_acl_sec_ctx {
1013 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1014 void *default_acl;
1015 void *acl;
1016 #endif
1017 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1018 void *sec_ctx;
1019 u32 sec_ctxlen;
1020 #endif
1021 struct ceph_pagelist *pagelist;
1022 };
1023
1024 #ifdef CONFIG_SECURITY
1025 extern bool ceph_security_xattr_deadlock(struct inode *in);
1026 extern bool ceph_security_xattr_wanted(struct inode *in);
1027 #else
ceph_security_xattr_deadlock(struct inode * in)1028 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1029 {
1030 return false;
1031 }
ceph_security_xattr_wanted(struct inode * in)1032 static inline bool ceph_security_xattr_wanted(struct inode *in)
1033 {
1034 return false;
1035 }
1036 #endif
1037
1038 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1039 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1040 struct ceph_acl_sec_ctx *ctx);
ceph_security_invalidate_secctx(struct inode * inode)1041 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1042 {
1043 security_inode_invalidate_secctx(inode);
1044 }
1045 #else
ceph_security_init_secctx(struct dentry * dentry,umode_t mode,struct ceph_acl_sec_ctx * ctx)1046 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1047 struct ceph_acl_sec_ctx *ctx)
1048 {
1049 return 0;
1050 }
ceph_security_invalidate_secctx(struct inode * inode)1051 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1052 {
1053 }
1054 #endif
1055
1056 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1057
1058 /* acl.c */
1059 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1060
1061 struct posix_acl *ceph_get_acl(struct inode *, int);
1062 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type);
1063 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1064 struct ceph_acl_sec_ctx *as_ctx);
1065 void ceph_init_inode_acls(struct inode *inode,
1066 struct ceph_acl_sec_ctx *as_ctx);
1067
ceph_forget_all_cached_acls(struct inode * inode)1068 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1069 {
1070 forget_all_cached_acls(inode);
1071 }
1072
1073 #else
1074
1075 #define ceph_get_acl NULL
1076 #define ceph_set_acl NULL
1077
ceph_pre_init_acls(struct inode * dir,umode_t * mode,struct ceph_acl_sec_ctx * as_ctx)1078 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1079 struct ceph_acl_sec_ctx *as_ctx)
1080 {
1081 return 0;
1082 }
ceph_init_inode_acls(struct inode * inode,struct ceph_acl_sec_ctx * as_ctx)1083 static inline void ceph_init_inode_acls(struct inode *inode,
1084 struct ceph_acl_sec_ctx *as_ctx)
1085 {
1086 }
ceph_acl_chmod(struct dentry * dentry,struct inode * inode)1087 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1088 {
1089 return 0;
1090 }
1091
ceph_forget_all_cached_acls(struct inode * inode)1092 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1093 {
1094 }
1095
1096 #endif
1097
1098 /* caps.c */
1099 extern const char *ceph_cap_string(int c);
1100 extern void ceph_handle_caps(struct ceph_mds_session *session,
1101 struct ceph_msg *msg);
1102 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1103 struct ceph_cap_reservation *ctx);
1104 extern void ceph_add_cap(struct inode *inode,
1105 struct ceph_mds_session *session, u64 cap_id,
1106 unsigned issued, unsigned wanted,
1107 unsigned cap, unsigned seq, u64 realmino, int flags,
1108 struct ceph_cap **new_cap);
1109 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1110 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1111 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1112 struct ceph_cap *cap);
1113 extern int ceph_is_any_caps(struct inode *inode);
1114
1115 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1116 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1117 int datasync);
1118 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1119 struct ceph_mds_session *session);
1120 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1121 struct ceph_mds_session *session);
1122 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1123 struct ceph_inode_info *ci);
1124 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1125 int mds);
1126 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1127 bool snap_rwsem_locked);
1128 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1129 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1130 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1131 int had);
1132 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1133 struct ceph_snap_context *snapc);
1134 extern void __ceph_remove_capsnap(struct inode *inode,
1135 struct ceph_cap_snap *capsnap,
1136 bool *wake_ci, bool *wake_mdsc);
1137 extern void ceph_remove_capsnap(struct inode *inode,
1138 struct ceph_cap_snap *capsnap,
1139 bool *wake_ci, bool *wake_mdsc);
1140 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1141 struct ceph_mds_session **psession);
1142 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1143 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1144 struct ceph_mds_session *session);
1145 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1146 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1147 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1148 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1149 int mds, int drop, int unless, int force);
1150 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1151 struct inode *dir,
1152 int mds, int drop, int unless);
1153
1154 extern int ceph_get_caps(struct file *filp, int need, int want,
1155 loff_t endoff, int *got, struct page **pinned_page);
1156 extern int ceph_try_get_caps(struct inode *inode,
1157 int need, int want, bool nonblock, int *got);
1158
1159 /* for counting open files by mode */
1160 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1161 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1162 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1163 struct ceph_mds_client *mdsc, int fmode);
1164
1165 /* addr.c */
1166 extern const struct address_space_operations ceph_aops;
1167 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1168 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1169 extern int ceph_pool_perm_check(struct inode *inode, int need);
1170 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1171
1172 /* file.c */
1173 extern const struct file_operations ceph_file_fops;
1174
1175 extern int ceph_renew_caps(struct inode *inode, int fmode);
1176 extern int ceph_open(struct inode *inode, struct file *file);
1177 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1178 struct file *file, unsigned flags, umode_t mode);
1179 extern int ceph_release(struct inode *inode, struct file *filp);
1180 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1181 char *data, size_t len);
1182
1183 /* dir.c */
1184 extern const struct file_operations ceph_dir_fops;
1185 extern const struct file_operations ceph_snapdir_fops;
1186 extern const struct inode_operations ceph_dir_iops;
1187 extern const struct inode_operations ceph_snapdir_iops;
1188 extern const struct dentry_operations ceph_dentry_ops;
1189
1190 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1191 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1192 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
1193 struct dentry *dentry, int err);
1194 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1195 struct dentry *dentry, int err);
1196
1197 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1198 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1199 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1200 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1201 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1202 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1203
1204 /* ioctl.c */
1205 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1206
1207 /* export.c */
1208 extern const struct export_operations ceph_export_ops;
1209 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1210
1211 /* locks.c */
1212 extern __init void ceph_flock_init(void);
1213 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1214 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1215 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1216 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1217 struct ceph_filelock *flocks,
1218 int num_fcntl_locks,
1219 int num_flock_locks);
1220 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1221 struct ceph_pagelist *pagelist,
1222 int num_fcntl_locks, int num_flock_locks);
1223
1224 /* debugfs.c */
1225 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1226 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1227
1228 /* quota.c */
__ceph_has_any_quota(struct ceph_inode_info * ci)1229 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1230 {
1231 return ci->i_max_files || ci->i_max_bytes;
1232 }
1233
1234 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1235
__ceph_update_quota(struct ceph_inode_info * ci,u64 max_bytes,u64 max_files)1236 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1237 u64 max_bytes, u64 max_files)
1238 {
1239 bool had_quota, has_quota;
1240 had_quota = __ceph_has_any_quota(ci);
1241 ci->i_max_bytes = max_bytes;
1242 ci->i_max_files = max_files;
1243 has_quota = __ceph_has_any_quota(ci);
1244
1245 if (had_quota != has_quota)
1246 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1247 }
1248
1249 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1250 struct ceph_mds_session *session,
1251 struct ceph_msg *msg);
1252 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1253 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1254 loff_t newlen);
1255 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1256 loff_t newlen);
1257 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1258 struct kstatfs *buf);
1259 extern int ceph_quota_check_rename(struct ceph_mds_client *mdsc,
1260 struct inode *old, struct inode *new);
1261 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1262
1263 #endif /* _FS_CEPH_SUPER_H */
1264