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 #define FSCACHE_USE_NEW_IO_API
25 #include <linux/fscache.h>
26 #endif
27
28 /* f_type in struct statfs */
29 #define CEPH_SUPER_MAGIC 0x00c36400
30
31 /* large granularity for statfs utilization stats to facilitate
32 * large volume sizes on 32-bit machines. */
33 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */
34 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
35
36 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */
37 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
38 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
39 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
40 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
41 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
42 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
43 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
44 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */
45 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */
46 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */
47 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */
48
49 #define CEPH_MOUNT_OPT_DEFAULT \
50 (CEPH_MOUNT_OPT_DCACHE | \
51 CEPH_MOUNT_OPT_NOCOPYFROM | \
52 CEPH_MOUNT_OPT_ASYNC_DIROPS)
53
54 #define ceph_set_mount_opt(fsc, opt) \
55 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
56 #define ceph_clear_mount_opt(fsc, opt) \
57 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
58 #define ceph_test_mount_opt(fsc, opt) \
59 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
60
61 /* max size of osd read request, limited by libceph */
62 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
63 /* osd has a configurable limitaion of max write size.
64 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
65 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN
66 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
67 #define CEPH_MAX_READDIR_DEFAULT 1024
68 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
69 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
70
71 /*
72 * Delay telling the MDS we no longer want caps, in case we reopen
73 * the file. Delay a minimum amount of time, even if we send a cap
74 * message for some other reason. Otherwise, take the oppotunity to
75 * update the mds to avoid sending another message later.
76 */
77 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
78 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
79
80 struct ceph_mount_options {
81 unsigned int flags;
82
83 unsigned int wsize; /* max write size */
84 unsigned int rsize; /* max read size */
85 unsigned int rasize; /* max readahead */
86 unsigned int congestion_kb; /* max writeback in flight */
87 unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
88 int caps_max;
89 unsigned int max_readdir; /* max readdir result (entries) */
90 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
91
92 /*
93 * everything above this point can be memcmp'd; everything below
94 * is handled in compare_mount_options()
95 */
96
97 char *snapdir_name; /* default ".snap" */
98 char *mds_namespace; /* default NULL */
99 char *server_path; /* default NULL (means "/") */
100 char *fscache_uniq; /* default NULL */
101 };
102
103 struct ceph_fs_client {
104 struct super_block *sb;
105
106 struct list_head metric_wakeup;
107
108 struct ceph_mount_options *mount_options;
109 struct ceph_client *client;
110
111 int mount_state;
112
113 bool blocklisted;
114
115 bool have_copy_from2;
116
117 u32 filp_gen;
118 loff_t max_file_size;
119
120 struct ceph_mds_client *mdsc;
121
122 atomic_long_t writeback_count;
123
124 struct workqueue_struct *inode_wq;
125 struct workqueue_struct *cap_wq;
126
127 #ifdef CONFIG_DEBUG_FS
128 struct dentry *debugfs_dentry_lru, *debugfs_caps;
129 struct dentry *debugfs_congestion_kb;
130 struct dentry *debugfs_bdi;
131 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
132 struct dentry *debugfs_status;
133 struct dentry *debugfs_mds_sessions;
134 struct dentry *debugfs_metrics_dir;
135 #endif
136
137 #ifdef CONFIG_CEPH_FSCACHE
138 struct fscache_cookie *fscache;
139 #endif
140 };
141
142
143 /*
144 * File i/o capability. This tracks shared state with the metadata
145 * server that allows us to cache or writeback attributes or to read
146 * and write data. For any given inode, we should have one or more
147 * capabilities, one issued by each metadata server, and our
148 * cumulative access is the OR of all issued capabilities.
149 *
150 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
151 * session capability lists.
152 */
153 struct ceph_cap {
154 struct ceph_inode_info *ci;
155 struct rb_node ci_node; /* per-ci cap tree */
156 struct ceph_mds_session *session;
157 struct list_head session_caps; /* per-session caplist */
158 u64 cap_id; /* unique cap id (mds provided) */
159 union {
160 /* in-use caps */
161 struct {
162 int issued; /* latest, from the mds */
163 int implemented; /* implemented superset of
164 issued (for revocation) */
165 int mds; /* mds index for this cap */
166 int mds_wanted; /* caps wanted from this mds */
167 };
168 /* caps to release */
169 struct {
170 u64 cap_ino;
171 int queue_release;
172 };
173 };
174 u32 seq, issue_seq, mseq;
175 u32 cap_gen; /* active/stale cycle */
176 unsigned long last_used;
177 struct list_head caps_item;
178 };
179
180 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */
181 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */
182 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */
183
184 struct ceph_cap_flush {
185 u64 tid;
186 int caps;
187 bool wake; /* wake up flush waiters when finish ? */
188 bool is_capsnap; /* true means capsnap */
189 struct list_head g_list; // global
190 struct list_head i_list; // per inode
191 };
192
193 /*
194 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
195 * we first complete any in-process sync writes and writeback any dirty
196 * data before flushing the snapped state (tracked here) back to the MDS.
197 */
198 struct ceph_cap_snap {
199 refcount_t nref;
200 struct list_head ci_item;
201
202 struct ceph_cap_flush cap_flush;
203
204 u64 follows;
205 int issued, dirty;
206 struct ceph_snap_context *context;
207
208 umode_t mode;
209 kuid_t uid;
210 kgid_t gid;
211
212 struct ceph_buffer *xattr_blob;
213 u64 xattr_version;
214
215 u64 size;
216 u64 change_attr;
217 struct timespec64 mtime, atime, ctime, btime;
218 u64 time_warp_seq;
219 u64 truncate_size;
220 u32 truncate_seq;
221 int writing; /* a sync write is still in progress */
222 int dirty_pages; /* dirty pages awaiting writeback */
223 bool inline_data;
224 bool need_flush;
225 };
226
ceph_put_cap_snap(struct ceph_cap_snap * capsnap)227 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
228 {
229 if (refcount_dec_and_test(&capsnap->nref)) {
230 if (capsnap->xattr_blob)
231 ceph_buffer_put(capsnap->xattr_blob);
232 kfree(capsnap);
233 }
234 }
235
236 /*
237 * The frag tree describes how a directory is fragmented, potentially across
238 * multiple metadata servers. It is also used to indicate points where
239 * metadata authority is delegated, and whether/where metadata is replicated.
240 *
241 * A _leaf_ frag will be present in the i_fragtree IFF there is
242 * delegation info. That is, if mds >= 0 || ndist > 0.
243 */
244 #define CEPH_MAX_DIRFRAG_REP 4
245
246 struct ceph_inode_frag {
247 struct rb_node node;
248
249 /* fragtree state */
250 u32 frag;
251 int split_by; /* i.e. 2^(split_by) children */
252
253 /* delegation and replication info */
254 int mds; /* -1 if same authority as parent */
255 int ndist; /* >0 if replicated */
256 int dist[CEPH_MAX_DIRFRAG_REP];
257 };
258
259 /*
260 * We cache inode xattrs as an encoded blob until they are first used,
261 * at which point we parse them into an rbtree.
262 */
263 struct ceph_inode_xattr {
264 struct rb_node node;
265
266 const char *name;
267 int name_len;
268 const char *val;
269 int val_len;
270 int dirty;
271
272 int should_free_name;
273 int should_free_val;
274 };
275
276 /*
277 * Ceph dentry state
278 */
279 struct ceph_dentry_info {
280 struct dentry *dentry;
281 struct ceph_mds_session *lease_session;
282 struct list_head lease_list;
283 unsigned flags;
284 int lease_shared_gen;
285 u32 lease_gen;
286 u32 lease_seq;
287 unsigned long lease_renew_after, lease_renew_from;
288 unsigned long time;
289 u64 offset;
290 };
291
292 #define CEPH_DENTRY_REFERENCED 1
293 #define CEPH_DENTRY_LEASE_LIST 2
294 #define CEPH_DENTRY_SHRINK_LIST 4
295 #define CEPH_DENTRY_PRIMARY_LINK 8
296
297 struct ceph_inode_xattrs_info {
298 /*
299 * (still encoded) xattr blob. we avoid the overhead of parsing
300 * this until someone actually calls getxattr, etc.
301 *
302 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
303 * NULL means we don't know.
304 */
305 struct ceph_buffer *blob, *prealloc_blob;
306
307 struct rb_root index;
308 bool dirty;
309 int count;
310 int names_size;
311 int vals_size;
312 u64 version, index_version;
313 };
314
315 /*
316 * Ceph inode.
317 */
318 struct ceph_inode_info {
319 struct ceph_vino i_vino; /* ceph ino + snap */
320
321 spinlock_t i_ceph_lock;
322
323 u64 i_version;
324 u64 i_inline_version;
325 u32 i_time_warp_seq;
326
327 unsigned long i_ceph_flags;
328 atomic64_t i_release_count;
329 atomic64_t i_ordered_count;
330 atomic64_t i_complete_seq[2];
331
332 struct ceph_dir_layout i_dir_layout;
333 struct ceph_file_layout i_layout;
334 struct ceph_file_layout i_cached_layout; // for async creates
335 char *i_symlink;
336
337 /* for dirs */
338 struct timespec64 i_rctime;
339 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
340 u64 i_files, i_subdirs;
341
342 /* quotas */
343 u64 i_max_bytes, i_max_files;
344
345 s32 i_dir_pin;
346
347 struct rb_root i_fragtree;
348 int i_fragtree_nsplits;
349 struct mutex i_fragtree_mutex;
350
351 struct ceph_inode_xattrs_info i_xattrs;
352
353 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
354 * s_mutex. */
355 struct rb_root i_caps; /* cap list */
356 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
357 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
358
359 /*
360 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
361 * is protected by the mdsc->cap_dirty_lock, but each individual item
362 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
363 * requires the mdsc->cap_dirty_lock. List presence for an item can
364 * be tested under the i_ceph_lock. Changing anything requires both.
365 */
366 struct list_head i_dirty_item;
367
368 /*
369 * Link to session's s_cap_flushing list. Protected in a similar
370 * fashion to i_dirty_item, but also by the s_mutex for changes. The
371 * s_cap_flushing list can be walked while holding either the s_mutex
372 * or msdc->cap_dirty_lock. List presence can also be checked while
373 * holding the i_ceph_lock for this inode.
374 */
375 struct list_head i_flushing_item;
376
377 /* we need to track cap writeback on a per-cap-bit basis, to allow
378 * overlapping, pipelined cap flushes to the mds. we can probably
379 * reduce the tid to 8 bits if we're concerned about inode size. */
380 struct ceph_cap_flush *i_prealloc_cap_flush;
381 struct list_head i_cap_flush_list;
382 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
383 unsigned long i_hold_caps_max; /* jiffies */
384 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
385 struct ceph_cap_reservation i_cap_migration_resv;
386 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
387 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
388 dirty|flushing caps */
389 unsigned i_snap_caps; /* cap bits for snapped files */
390
391 unsigned long i_last_rd;
392 unsigned long i_last_wr;
393 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
394
395 struct mutex i_truncate_mutex;
396 u32 i_truncate_seq; /* last truncate to smaller size */
397 u64 i_truncate_size; /* and the size we last truncated down to */
398 int i_truncate_pending; /* still need to call vmtruncate */
399
400 u64 i_max_size; /* max file size authorized by mds */
401 u64 i_reported_size; /* (max_)size reported to or requested of mds */
402 u64 i_wanted_max_size; /* offset we'd like to write too */
403 u64 i_requested_max_size; /* max_size we've requested */
404
405 /* held references to caps */
406 int i_pin_ref;
407 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
408 int i_wrbuffer_ref, i_wrbuffer_ref_head;
409 atomic_t i_filelock_ref;
410 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
411 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
412 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
413
414 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
415 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
416 spinlock_t i_unsafe_lock;
417
418 union {
419 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
420 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
421 };
422 struct list_head i_snap_realm_item;
423 struct list_head i_snap_flush_item;
424 struct timespec64 i_btime;
425 struct timespec64 i_snap_btime;
426
427 struct work_struct i_work;
428 unsigned long i_work_mask;
429
430 #ifdef CONFIG_CEPH_FSCACHE
431 struct fscache_cookie *fscache;
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 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */
585
586 /*
587 * Masks of ceph inode work.
588 */
589 #define CEPH_I_WORK_WRITEBACK 0
590 #define CEPH_I_WORK_INVALIDATE_PAGES 1
591 #define CEPH_I_WORK_VMTRUNCATE 2
592 #define CEPH_I_WORK_CHECK_CAPS 3
593 #define CEPH_I_WORK_FLUSH_SNAPS 4
594
595 /*
596 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
597 * and then clear it when they start succeeding. Note that we do a lockless
598 * check first, and only take the lock if it looks like it needs to be changed.
599 * The write submission code just takes this as a hint, so we're not too
600 * worried if a few slip through in either direction.
601 */
ceph_set_error_write(struct ceph_inode_info * ci)602 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
603 {
604 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
605 spin_lock(&ci->i_ceph_lock);
606 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
607 spin_unlock(&ci->i_ceph_lock);
608 }
609 }
610
ceph_clear_error_write(struct ceph_inode_info * ci)611 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
612 {
613 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
614 spin_lock(&ci->i_ceph_lock);
615 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
616 spin_unlock(&ci->i_ceph_lock);
617 }
618 }
619
__ceph_dir_set_complete(struct ceph_inode_info * ci,long long release_count,long long ordered_count)620 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
621 long long release_count,
622 long long ordered_count)
623 {
624 /*
625 * Makes sure operations that setup readdir cache (update page
626 * cache and i_size) are strongly ordered w.r.t. the following
627 * atomic64_set() operations.
628 */
629 smp_mb();
630 atomic64_set(&ci->i_complete_seq[0], release_count);
631 atomic64_set(&ci->i_complete_seq[1], ordered_count);
632 }
633
__ceph_dir_clear_complete(struct ceph_inode_info * ci)634 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
635 {
636 atomic64_inc(&ci->i_release_count);
637 }
638
__ceph_dir_clear_ordered(struct ceph_inode_info * ci)639 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
640 {
641 atomic64_inc(&ci->i_ordered_count);
642 }
643
__ceph_dir_is_complete(struct ceph_inode_info * ci)644 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
645 {
646 return atomic64_read(&ci->i_complete_seq[0]) ==
647 atomic64_read(&ci->i_release_count);
648 }
649
__ceph_dir_is_complete_ordered(struct ceph_inode_info * ci)650 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
651 {
652 return atomic64_read(&ci->i_complete_seq[0]) ==
653 atomic64_read(&ci->i_release_count) &&
654 atomic64_read(&ci->i_complete_seq[1]) ==
655 atomic64_read(&ci->i_ordered_count);
656 }
657
ceph_dir_clear_complete(struct inode * inode)658 static inline void ceph_dir_clear_complete(struct inode *inode)
659 {
660 __ceph_dir_clear_complete(ceph_inode(inode));
661 }
662
ceph_dir_clear_ordered(struct inode * inode)663 static inline void ceph_dir_clear_ordered(struct inode *inode)
664 {
665 __ceph_dir_clear_ordered(ceph_inode(inode));
666 }
667
ceph_dir_is_complete_ordered(struct inode * inode)668 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
669 {
670 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
671 smp_rmb();
672 return ret;
673 }
674
675 /* find a specific frag @f */
676 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
677 u32 f);
678
679 /*
680 * choose fragment for value @v. copy frag content to pfrag, if leaf
681 * exists
682 */
683 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
684 struct ceph_inode_frag *pfrag,
685 int *found);
686
ceph_dentry(const struct dentry * dentry)687 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
688 {
689 return (struct ceph_dentry_info *)dentry->d_fsdata;
690 }
691
692 /*
693 * caps helpers
694 */
__ceph_is_any_real_caps(struct ceph_inode_info * ci)695 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
696 {
697 return !RB_EMPTY_ROOT(&ci->i_caps);
698 }
699
700 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
701 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
702 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
703 int t);
704 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
705 struct ceph_cap *cap);
706
ceph_caps_issued(struct ceph_inode_info * ci)707 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
708 {
709 int issued;
710 spin_lock(&ci->i_ceph_lock);
711 issued = __ceph_caps_issued(ci, NULL);
712 spin_unlock(&ci->i_ceph_lock);
713 return issued;
714 }
715
ceph_caps_issued_mask_metric(struct ceph_inode_info * ci,int mask,int touch)716 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
717 int mask, int touch)
718 {
719 int r;
720 spin_lock(&ci->i_ceph_lock);
721 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
722 spin_unlock(&ci->i_ceph_lock);
723 return r;
724 }
725
__ceph_caps_dirty(struct ceph_inode_info * ci)726 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
727 {
728 return ci->i_dirty_caps | ci->i_flushing_caps;
729 }
730 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
731 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
732 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
733 struct ceph_cap_flush **pcf);
734
735 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
736 struct ceph_cap *ocap, int mask);
737 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
738 extern int __ceph_caps_used(struct ceph_inode_info *ci);
739
__ceph_is_file_opened(struct ceph_inode_info * ci)740 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
741 {
742 return ci->i_nr_by_mode[0];
743 }
744 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
745 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
746
747 /* what the mds thinks we want */
748 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
749
750 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
751 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
752 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
753 struct ceph_mount_options *fsopt);
754 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
755 struct ceph_cap_reservation *ctx, int need);
756 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
757 struct ceph_cap_reservation *ctx);
758 extern void ceph_reservation_status(struct ceph_fs_client *client,
759 int *total, int *avail, int *used,
760 int *reserved, int *min);
761
762
763
764 /*
765 * we keep buffered readdir results attached to file->private_data
766 */
767 #define CEPH_F_SYNC 1
768 #define CEPH_F_ATEND 2
769
770 struct ceph_file_info {
771 short fmode; /* initialized on open */
772 short flags; /* CEPH_F_* */
773
774 spinlock_t rw_contexts_lock;
775 struct list_head rw_contexts;
776
777 u32 filp_gen;
778 atomic_t num_locks;
779 };
780
781 struct ceph_dir_file_info {
782 struct ceph_file_info file_info;
783
784 /* readdir: position within the dir */
785 u32 frag;
786 struct ceph_mds_request *last_readdir;
787
788 /* readdir: position within a frag */
789 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
790 char *last_name; /* last entry in previous chunk */
791 long long dir_release_count;
792 long long dir_ordered_count;
793 int readdir_cache_idx;
794
795 /* used for -o dirstat read() on directory thing */
796 char *dir_info;
797 int dir_info_len;
798 };
799
800 struct ceph_rw_context {
801 struct list_head list;
802 struct task_struct *thread;
803 int caps;
804 };
805
806 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
807 struct ceph_rw_context _name = { \
808 .thread = current, \
809 .caps = _caps, \
810 }
811
ceph_add_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)812 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
813 struct ceph_rw_context *ctx)
814 {
815 spin_lock(&cf->rw_contexts_lock);
816 list_add(&ctx->list, &cf->rw_contexts);
817 spin_unlock(&cf->rw_contexts_lock);
818 }
819
ceph_del_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)820 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
821 struct ceph_rw_context *ctx)
822 {
823 spin_lock(&cf->rw_contexts_lock);
824 list_del(&ctx->list);
825 spin_unlock(&cf->rw_contexts_lock);
826 }
827
828 static inline struct ceph_rw_context*
ceph_find_rw_context(struct ceph_file_info * cf)829 ceph_find_rw_context(struct ceph_file_info *cf)
830 {
831 struct ceph_rw_context *ctx, *found = NULL;
832 spin_lock(&cf->rw_contexts_lock);
833 list_for_each_entry(ctx, &cf->rw_contexts, list) {
834 if (ctx->thread == current) {
835 found = ctx;
836 break;
837 }
838 }
839 spin_unlock(&cf->rw_contexts_lock);
840 return found;
841 }
842
843 struct ceph_readdir_cache_control {
844 struct page *page;
845 struct dentry **dentries;
846 int index;
847 };
848
849 /*
850 * A "snap realm" describes a subset of the file hierarchy sharing
851 * the same set of snapshots that apply to it. The realms themselves
852 * are organized into a hierarchy, such that children inherit (some of)
853 * the snapshots of their parents.
854 *
855 * All inodes within the realm that have capabilities are linked into a
856 * per-realm list.
857 */
858 struct ceph_snap_realm {
859 u64 ino;
860 struct inode *inode;
861 atomic_t nref;
862 struct rb_node node;
863
864 u64 created, seq;
865 u64 parent_ino;
866 u64 parent_since; /* snapid when our current parent became so */
867
868 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
869 u32 num_prior_parent_snaps; /* had prior to parent_since */
870 u64 *snaps; /* snaps specific to this realm */
871 u32 num_snaps;
872
873 struct ceph_snap_realm *parent;
874 struct list_head children; /* list of child realms */
875 struct list_head child_item;
876
877 struct list_head empty_item; /* if i have ref==0 */
878
879 struct list_head dirty_item; /* if realm needs new context */
880
881 /* the current set of snaps for this realm */
882 struct ceph_snap_context *cached_context;
883
884 struct list_head inodes_with_caps;
885 spinlock_t inodes_with_caps_lock;
886 };
887
default_congestion_kb(void)888 static inline int default_congestion_kb(void)
889 {
890 int congestion_kb;
891
892 /*
893 * Copied from NFS
894 *
895 * congestion size, scale with available memory.
896 *
897 * 64MB: 8192k
898 * 128MB: 11585k
899 * 256MB: 16384k
900 * 512MB: 23170k
901 * 1GB: 32768k
902 * 2GB: 46340k
903 * 4GB: 65536k
904 * 8GB: 92681k
905 * 16GB: 131072k
906 *
907 * This allows larger machines to have larger/more transfers.
908 * Limit the default to 256M
909 */
910 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
911 if (congestion_kb > 256*1024)
912 congestion_kb = 256*1024;
913
914 return congestion_kb;
915 }
916
917
918 /* super.c */
919 extern int ceph_force_reconnect(struct super_block *sb);
920 /* snap.c */
921 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
922 u64 ino);
923 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
924 struct ceph_snap_realm *realm);
925 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
926 struct ceph_snap_realm *realm);
927 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
928 void *p, void *e, bool deletion,
929 struct ceph_snap_realm **realm_ret);
930 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
931 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
932 struct ceph_mds_session *session,
933 struct ceph_msg *msg);
934 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
935 struct ceph_cap_snap *capsnap);
936 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
937
938 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
939 u64 snap);
940 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
941 struct ceph_snapid_map *sm);
942 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
943 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
944 void ceph_umount_begin(struct super_block *sb);
945
946
947 /*
948 * a cap_snap is "pending" if it is still awaiting an in-progress
949 * sync write (that may/may not still update size, mtime, etc.).
950 */
__ceph_have_pending_cap_snap(struct ceph_inode_info * ci)951 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
952 {
953 return !list_empty(&ci->i_cap_snaps) &&
954 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
955 ci_item)->writing;
956 }
957
958 /* inode.c */
959 struct ceph_mds_reply_info_in;
960 struct ceph_mds_reply_dirfrag;
961
962 extern const struct inode_operations ceph_file_iops;
963
964 extern struct inode *ceph_alloc_inode(struct super_block *sb);
965 extern void ceph_evict_inode(struct inode *inode);
966 extern void ceph_free_inode(struct inode *inode);
967
968 extern struct inode *ceph_get_inode(struct super_block *sb,
969 struct ceph_vino vino);
970 extern struct inode *ceph_get_snapdir(struct inode *parent);
971 extern int ceph_fill_file_size(struct inode *inode, int issued,
972 u32 truncate_seq, u64 truncate_size, u64 size);
973 extern void ceph_fill_file_time(struct inode *inode, int issued,
974 u64 time_warp_seq, struct timespec64 *ctime,
975 struct timespec64 *mtime,
976 struct timespec64 *atime);
977 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
978 struct ceph_mds_reply_info_in *iinfo,
979 struct ceph_mds_reply_dirfrag *dirinfo,
980 struct ceph_mds_session *session, int cap_fmode,
981 struct ceph_cap_reservation *caps_reservation);
982 extern int ceph_fill_trace(struct super_block *sb,
983 struct ceph_mds_request *req);
984 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
985 struct ceph_mds_session *session);
986
987 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
988
989 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
990 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
991
992 void ceph_queue_inode_work(struct inode *inode, int work_bit);
993
ceph_queue_vmtruncate(struct inode * inode)994 static inline void ceph_queue_vmtruncate(struct inode *inode)
995 {
996 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
997 }
998
ceph_queue_invalidate(struct inode * inode)999 static inline void ceph_queue_invalidate(struct inode *inode)
1000 {
1001 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1002 }
1003
ceph_queue_writeback(struct inode * inode)1004 static inline void ceph_queue_writeback(struct inode *inode)
1005 {
1006 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1007 }
1008
ceph_queue_check_caps(struct inode * inode)1009 static inline void ceph_queue_check_caps(struct inode *inode)
1010 {
1011 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1012 }
1013
ceph_queue_flush_snaps(struct inode * inode)1014 static inline void ceph_queue_flush_snaps(struct inode *inode)
1015 {
1016 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1017 }
1018
1019 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1020 int mask, bool force);
ceph_do_getattr(struct inode * inode,int mask,bool force)1021 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1022 {
1023 return __ceph_do_getattr(inode, NULL, mask, force);
1024 }
1025 extern int ceph_permission(struct user_namespace *mnt_userns,
1026 struct inode *inode, int mask);
1027 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1028 extern int ceph_setattr(struct user_namespace *mnt_userns,
1029 struct dentry *dentry, struct iattr *attr);
1030 extern int ceph_getattr(struct user_namespace *mnt_userns,
1031 const struct path *path, struct kstat *stat,
1032 u32 request_mask, unsigned int flags);
1033 void ceph_inode_shutdown(struct inode *inode);
1034
ceph_inode_is_shutdown(struct inode * inode)1035 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1036 {
1037 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1038 struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1039 int state = READ_ONCE(fsc->mount_state);
1040
1041 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1042 }
1043
1044 /* xattr.c */
1045 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1046 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1047 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1048 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1049 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1050 extern const struct xattr_handler *ceph_xattr_handlers[];
1051
1052 struct ceph_acl_sec_ctx {
1053 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1054 void *default_acl;
1055 void *acl;
1056 #endif
1057 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1058 void *sec_ctx;
1059 u32 sec_ctxlen;
1060 #endif
1061 struct ceph_pagelist *pagelist;
1062 };
1063
1064 #ifdef CONFIG_SECURITY
1065 extern bool ceph_security_xattr_deadlock(struct inode *in);
1066 extern bool ceph_security_xattr_wanted(struct inode *in);
1067 #else
ceph_security_xattr_deadlock(struct inode * in)1068 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1069 {
1070 return false;
1071 }
ceph_security_xattr_wanted(struct inode * in)1072 static inline bool ceph_security_xattr_wanted(struct inode *in)
1073 {
1074 return false;
1075 }
1076 #endif
1077
1078 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1079 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1080 struct ceph_acl_sec_ctx *ctx);
ceph_security_invalidate_secctx(struct inode * inode)1081 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1082 {
1083 security_inode_invalidate_secctx(inode);
1084 }
1085 #else
ceph_security_init_secctx(struct dentry * dentry,umode_t mode,struct ceph_acl_sec_ctx * ctx)1086 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1087 struct ceph_acl_sec_ctx *ctx)
1088 {
1089 return 0;
1090 }
ceph_security_invalidate_secctx(struct inode * inode)1091 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1092 {
1093 }
1094 #endif
1095
1096 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1097
1098 /* acl.c */
1099 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1100
1101 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1102 int ceph_set_acl(struct user_namespace *mnt_userns,
1103 struct inode *inode, struct posix_acl *acl, int type);
1104 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1105 struct ceph_acl_sec_ctx *as_ctx);
1106 void ceph_init_inode_acls(struct inode *inode,
1107 struct ceph_acl_sec_ctx *as_ctx);
1108
ceph_forget_all_cached_acls(struct inode * inode)1109 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1110 {
1111 forget_all_cached_acls(inode);
1112 }
1113
1114 #else
1115
1116 #define ceph_get_acl NULL
1117 #define ceph_set_acl NULL
1118
ceph_pre_init_acls(struct inode * dir,umode_t * mode,struct ceph_acl_sec_ctx * as_ctx)1119 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1120 struct ceph_acl_sec_ctx *as_ctx)
1121 {
1122 return 0;
1123 }
ceph_init_inode_acls(struct inode * inode,struct ceph_acl_sec_ctx * as_ctx)1124 static inline void ceph_init_inode_acls(struct inode *inode,
1125 struct ceph_acl_sec_ctx *as_ctx)
1126 {
1127 }
ceph_acl_chmod(struct dentry * dentry,struct inode * inode)1128 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1129 {
1130 return 0;
1131 }
1132
ceph_forget_all_cached_acls(struct inode * inode)1133 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1134 {
1135 }
1136
1137 #endif
1138
1139 /* caps.c */
1140 extern const char *ceph_cap_string(int c);
1141 extern void ceph_handle_caps(struct ceph_mds_session *session,
1142 struct ceph_msg *msg);
1143 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1144 struct ceph_cap_reservation *ctx);
1145 extern void ceph_add_cap(struct inode *inode,
1146 struct ceph_mds_session *session, u64 cap_id,
1147 unsigned issued, unsigned wanted,
1148 unsigned cap, unsigned seq, u64 realmino, int flags,
1149 struct ceph_cap **new_cap);
1150 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1151 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1152 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1153 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1154 struct ceph_cap *cap);
1155 extern int ceph_is_any_caps(struct inode *inode);
1156
1157 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1158 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1159 int datasync);
1160 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1161 struct ceph_mds_session *session);
1162 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1163 struct ceph_mds_session *session);
1164 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1165 struct ceph_inode_info *ci);
1166 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1167 int mds);
1168 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1169 bool snap_rwsem_locked);
1170 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1171 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1172 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1173 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1174 int had);
1175 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1176 struct ceph_snap_context *snapc);
1177 extern void __ceph_remove_capsnap(struct inode *inode,
1178 struct ceph_cap_snap *capsnap,
1179 bool *wake_ci, bool *wake_mdsc);
1180 extern void ceph_remove_capsnap(struct inode *inode,
1181 struct ceph_cap_snap *capsnap,
1182 bool *wake_ci, bool *wake_mdsc);
1183 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1184 struct ceph_mds_session **psession);
1185 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1186 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1187 struct ceph_mds_session *session);
1188 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1189 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1190 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1191 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1192 int mds, int drop, int unless, int force);
1193 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1194 struct inode *dir,
1195 int mds, int drop, int unless);
1196
1197 extern int ceph_get_caps(struct file *filp, int need, int want,
1198 loff_t endoff, int *got);
1199 extern int ceph_try_get_caps(struct inode *inode,
1200 int need, int want, bool nonblock, int *got);
1201
1202 /* for counting open files by mode */
1203 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1204 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1205 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1206 struct ceph_mds_client *mdsc, int fmode);
1207
1208 /* addr.c */
1209 extern const struct address_space_operations ceph_aops;
1210 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1211 extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1212 extern int ceph_pool_perm_check(struct inode *inode, int need);
1213 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1214 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1215
1216 /* file.c */
1217 extern const struct file_operations ceph_file_fops;
1218
1219 extern int ceph_renew_caps(struct inode *inode, int fmode);
1220 extern int ceph_open(struct inode *inode, struct file *file);
1221 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1222 struct file *file, unsigned flags, umode_t mode);
1223 extern int ceph_release(struct inode *inode, struct file *filp);
1224 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1225 char *data, size_t len);
1226
1227 /* dir.c */
1228 extern const struct file_operations ceph_dir_fops;
1229 extern const struct file_operations ceph_snapdir_fops;
1230 extern const struct inode_operations ceph_dir_iops;
1231 extern const struct inode_operations ceph_snapdir_iops;
1232 extern const struct dentry_operations ceph_dentry_ops;
1233
1234 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1235 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1236 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1237 struct dentry *dentry);
1238 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1239 struct dentry *dentry, int err);
1240
1241 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1242 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1243 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1244 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1245 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1246 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1247
1248 /* ioctl.c */
1249 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1250
1251 /* export.c */
1252 extern const struct export_operations ceph_export_ops;
1253 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1254
1255 /* locks.c */
1256 extern __init void ceph_flock_init(void);
1257 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1258 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1259 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1260 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1261 struct ceph_filelock *flocks,
1262 int num_fcntl_locks,
1263 int num_flock_locks);
1264 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1265 struct ceph_pagelist *pagelist,
1266 int num_fcntl_locks, int num_flock_locks);
1267
1268 /* debugfs.c */
1269 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1270 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1271
1272 /* quota.c */
__ceph_has_any_quota(struct ceph_inode_info * ci)1273 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1274 {
1275 return ci->i_max_files || ci->i_max_bytes;
1276 }
1277
1278 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1279
__ceph_update_quota(struct ceph_inode_info * ci,u64 max_bytes,u64 max_files)1280 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1281 u64 max_bytes, u64 max_files)
1282 {
1283 bool had_quota, has_quota;
1284 had_quota = __ceph_has_any_quota(ci);
1285 ci->i_max_bytes = max_bytes;
1286 ci->i_max_files = max_files;
1287 has_quota = __ceph_has_any_quota(ci);
1288
1289 if (had_quota != has_quota)
1290 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1291 }
1292
1293 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1294 struct ceph_mds_session *session,
1295 struct ceph_msg *msg);
1296 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1297 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1298 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1299 loff_t newlen);
1300 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1301 loff_t newlen);
1302 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1303 struct kstatfs *buf);
1304 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1305
1306 #endif /* _FS_CEPH_SUPER_H */
1307