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 #include <linux/netfs.h>
21 #include <linux/fscache.h>
22 #include <linux/hashtable.h>
23
24 #include <linux/ceph/libceph.h>
25
26 /* large granularity for statfs utilization stats to facilitate
27 * large volume sizes on 32-bit machines. */
28 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */
29 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT)
30 #define CEPH_4K_BLOCK_SHIFT 12 /* 4 KB */
31
32 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */
33 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */
34 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */
35 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */
36 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */
37 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */
38 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */
39 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */
40 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */
41 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */
42 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */
43 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */
44 #define CEPH_MOUNT_OPT_NOPAGECACHE (1<<16) /* bypass pagecache altogether */
45
46 #define CEPH_MOUNT_OPT_DEFAULT \
47 (CEPH_MOUNT_OPT_DCACHE | \
48 CEPH_MOUNT_OPT_NOCOPYFROM | \
49 CEPH_MOUNT_OPT_ASYNC_DIROPS)
50
51 #define ceph_set_mount_opt(fsc, opt) \
52 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
53 #define ceph_clear_mount_opt(fsc, opt) \
54 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
55 #define ceph_test_mount_opt(fsc, opt) \
56 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
57
58 /* max size of osd read request, limited by libceph */
59 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN
60 /* osd has a configurable limitaion of max write size.
61 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
62 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN
63 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */
64 #define CEPH_MAX_READDIR_DEFAULT 1024
65 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024)
66 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
67
68 /*
69 * Delay telling the MDS we no longer want caps, in case we reopen
70 * the file. Delay a minimum amount of time, even if we send a cap
71 * message for some other reason. Otherwise, take the oppotunity to
72 * update the mds to avoid sending another message later.
73 */
74 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */
75 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */
76
77 struct ceph_mount_options {
78 unsigned int flags;
79
80 unsigned int wsize; /* max write size */
81 unsigned int rsize; /* max read size */
82 unsigned int rasize; /* max readahead */
83 unsigned int congestion_kb; /* max writeback in flight */
84 unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
85 int caps_max;
86 unsigned int max_readdir; /* max readdir result (entries) */
87 unsigned int max_readdir_bytes; /* max readdir result (bytes) */
88
89 bool new_dev_syntax;
90
91 /*
92 * everything above this point can be memcmp'd; everything below
93 * is handled in compare_mount_options()
94 */
95
96 char *snapdir_name; /* default ".snap" */
97 char *mds_namespace; /* default NULL */
98 char *server_path; /* default NULL (means "/") */
99 char *fscache_uniq; /* default NULL */
100 char *mon_addr;
101 };
102
103 /* mount state */
104 enum {
105 CEPH_MOUNT_MOUNTING,
106 CEPH_MOUNT_MOUNTED,
107 CEPH_MOUNT_UNMOUNTING,
108 CEPH_MOUNT_UNMOUNTED,
109 CEPH_MOUNT_SHUTDOWN,
110 CEPH_MOUNT_RECOVER,
111 CEPH_MOUNT_FENCE_IO,
112 };
113
114 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8
115
116 struct ceph_fs_client {
117 struct super_block *sb;
118
119 struct list_head metric_wakeup;
120
121 struct ceph_mount_options *mount_options;
122 struct ceph_client *client;
123
124 int mount_state;
125
126 bool blocklisted;
127
128 bool have_copy_from2;
129
130 u32 filp_gen;
131 loff_t max_file_size;
132
133 struct ceph_mds_client *mdsc;
134
135 atomic_long_t writeback_count;
136 bool write_congested;
137
138 struct workqueue_struct *inode_wq;
139 struct workqueue_struct *cap_wq;
140
141 DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS);
142 spinlock_t async_unlink_conflict_lock;
143
144 #ifdef CONFIG_DEBUG_FS
145 struct dentry *debugfs_dentry_lru, *debugfs_caps;
146 struct dentry *debugfs_congestion_kb;
147 struct dentry *debugfs_bdi;
148 struct dentry *debugfs_mdsc, *debugfs_mdsmap;
149 struct dentry *debugfs_status;
150 struct dentry *debugfs_mds_sessions;
151 struct dentry *debugfs_metrics_dir;
152 #endif
153
154 #ifdef CONFIG_CEPH_FSCACHE
155 struct fscache_volume *fscache;
156 #endif
157 };
158
159
160 /*
161 * File i/o capability. This tracks shared state with the metadata
162 * server that allows us to cache or writeback attributes or to read
163 * and write data. For any given inode, we should have one or more
164 * capabilities, one issued by each metadata server, and our
165 * cumulative access is the OR of all issued capabilities.
166 *
167 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
168 * session capability lists.
169 */
170 struct ceph_cap {
171 struct ceph_inode_info *ci;
172 struct rb_node ci_node; /* per-ci cap tree */
173 struct ceph_mds_session *session;
174 struct list_head session_caps; /* per-session caplist */
175 u64 cap_id; /* unique cap id (mds provided) */
176 union {
177 /* in-use caps */
178 struct {
179 int issued; /* latest, from the mds */
180 int implemented; /* implemented superset of
181 issued (for revocation) */
182 int mds; /* mds index for this cap */
183 int mds_wanted; /* caps wanted from this mds */
184 };
185 /* caps to release */
186 struct {
187 u64 cap_ino;
188 int queue_release;
189 };
190 };
191 u32 seq, issue_seq, mseq;
192 u32 cap_gen; /* active/stale cycle */
193 unsigned long last_used;
194 struct list_head caps_item;
195 };
196
197 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */
198 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */
199 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */
200
201 struct ceph_cap_flush {
202 u64 tid;
203 int caps;
204 bool wake; /* wake up flush waiters when finish ? */
205 bool is_capsnap; /* true means capsnap */
206 struct list_head g_list; // global
207 struct list_head i_list; // per inode
208 };
209
210 /*
211 * Snapped cap state that is pending flush to mds. When a snapshot occurs,
212 * we first complete any in-process sync writes and writeback any dirty
213 * data before flushing the snapped state (tracked here) back to the MDS.
214 */
215 struct ceph_cap_snap {
216 refcount_t nref;
217 struct list_head ci_item;
218
219 struct ceph_cap_flush cap_flush;
220
221 u64 follows;
222 int issued, dirty;
223 struct ceph_snap_context *context;
224
225 umode_t mode;
226 kuid_t uid;
227 kgid_t gid;
228
229 struct ceph_buffer *xattr_blob;
230 u64 xattr_version;
231
232 u64 size;
233 u64 change_attr;
234 struct timespec64 mtime, atime, ctime, btime;
235 u64 time_warp_seq;
236 u64 truncate_size;
237 u32 truncate_seq;
238 int writing; /* a sync write is still in progress */
239 int dirty_pages; /* dirty pages awaiting writeback */
240 bool inline_data;
241 bool need_flush;
242 };
243
ceph_put_cap_snap(struct ceph_cap_snap * capsnap)244 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
245 {
246 if (refcount_dec_and_test(&capsnap->nref)) {
247 if (capsnap->xattr_blob)
248 ceph_buffer_put(capsnap->xattr_blob);
249 kmem_cache_free(ceph_cap_snap_cachep, capsnap);
250 }
251 }
252
253 /*
254 * The frag tree describes how a directory is fragmented, potentially across
255 * multiple metadata servers. It is also used to indicate points where
256 * metadata authority is delegated, and whether/where metadata is replicated.
257 *
258 * A _leaf_ frag will be present in the i_fragtree IFF there is
259 * delegation info. That is, if mds >= 0 || ndist > 0.
260 */
261 #define CEPH_MAX_DIRFRAG_REP 4
262
263 struct ceph_inode_frag {
264 struct rb_node node;
265
266 /* fragtree state */
267 u32 frag;
268 int split_by; /* i.e. 2^(split_by) children */
269
270 /* delegation and replication info */
271 int mds; /* -1 if same authority as parent */
272 int ndist; /* >0 if replicated */
273 int dist[CEPH_MAX_DIRFRAG_REP];
274 };
275
276 /*
277 * We cache inode xattrs as an encoded blob until they are first used,
278 * at which point we parse them into an rbtree.
279 */
280 struct ceph_inode_xattr {
281 struct rb_node node;
282
283 const char *name;
284 int name_len;
285 const char *val;
286 int val_len;
287 int dirty;
288
289 int should_free_name;
290 int should_free_val;
291 };
292
293 /*
294 * Ceph dentry state
295 */
296 struct ceph_dentry_info {
297 struct dentry *dentry;
298 struct ceph_mds_session *lease_session;
299 struct list_head lease_list;
300 struct hlist_node hnode;
301 unsigned long flags;
302 int lease_shared_gen;
303 u32 lease_gen;
304 u32 lease_seq;
305 unsigned long lease_renew_after, lease_renew_from;
306 unsigned long time;
307 u64 offset;
308 };
309
310 #define CEPH_DENTRY_REFERENCED (1 << 0)
311 #define CEPH_DENTRY_LEASE_LIST (1 << 1)
312 #define CEPH_DENTRY_SHRINK_LIST (1 << 2)
313 #define CEPH_DENTRY_PRIMARY_LINK (1 << 3)
314 #define CEPH_DENTRY_ASYNC_UNLINK_BIT (4)
315 #define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT)
316 #define CEPH_DENTRY_ASYNC_CREATE_BIT (5)
317 #define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT)
318
319 struct ceph_inode_xattrs_info {
320 /*
321 * (still encoded) xattr blob. we avoid the overhead of parsing
322 * this until someone actually calls getxattr, etc.
323 *
324 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
325 * NULL means we don't know.
326 */
327 struct ceph_buffer *blob, *prealloc_blob;
328
329 struct rb_root index;
330 bool dirty;
331 int count;
332 int names_size;
333 int vals_size;
334 u64 version, index_version;
335 };
336
337 /*
338 * Ceph inode.
339 */
340 struct ceph_inode_info {
341 struct netfs_inode netfs; /* Netfslib context and vfs inode */
342 struct ceph_vino i_vino; /* ceph ino + snap */
343
344 spinlock_t i_ceph_lock;
345
346 u64 i_version;
347 u64 i_inline_version;
348 u32 i_time_warp_seq;
349
350 unsigned long i_ceph_flags;
351 atomic64_t i_release_count;
352 atomic64_t i_ordered_count;
353 atomic64_t i_complete_seq[2];
354
355 struct ceph_dir_layout i_dir_layout;
356 struct ceph_file_layout i_layout;
357 struct ceph_file_layout i_cached_layout; // for async creates
358 char *i_symlink;
359
360 /* for dirs */
361 struct timespec64 i_rctime;
362 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
363 u64 i_files, i_subdirs;
364
365 /* quotas */
366 u64 i_max_bytes, i_max_files;
367
368 s32 i_dir_pin;
369
370 struct rb_root i_fragtree;
371 int i_fragtree_nsplits;
372 struct mutex i_fragtree_mutex;
373
374 struct ceph_inode_xattrs_info i_xattrs;
375
376 /* capabilities. protected _both_ by i_ceph_lock and cap->session's
377 * s_mutex. */
378 struct rb_root i_caps; /* cap list */
379 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */
380 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */
381
382 /*
383 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
384 * is protected by the mdsc->cap_dirty_lock, but each individual item
385 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
386 * requires the mdsc->cap_dirty_lock. List presence for an item can
387 * be tested under the i_ceph_lock. Changing anything requires both.
388 */
389 struct list_head i_dirty_item;
390
391 /*
392 * Link to session's s_cap_flushing list. Protected in a similar
393 * fashion to i_dirty_item, but also by the s_mutex for changes. The
394 * s_cap_flushing list can be walked while holding either the s_mutex
395 * or msdc->cap_dirty_lock. List presence can also be checked while
396 * holding the i_ceph_lock for this inode.
397 */
398 struct list_head i_flushing_item;
399
400 /* we need to track cap writeback on a per-cap-bit basis, to allow
401 * overlapping, pipelined cap flushes to the mds. we can probably
402 * reduce the tid to 8 bits if we're concerned about inode size. */
403 struct ceph_cap_flush *i_prealloc_cap_flush;
404 struct list_head i_cap_flush_list;
405 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */
406 unsigned long i_hold_caps_max; /* jiffies */
407 struct list_head i_cap_delay_list; /* for delayed cap release to mds */
408 struct ceph_cap_reservation i_cap_migration_resv;
409 struct list_head i_cap_snaps; /* snapped state pending flush to mds */
410 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or
411 dirty|flushing caps */
412 unsigned i_snap_caps; /* cap bits for snapped files */
413
414 unsigned long i_last_rd;
415 unsigned long i_last_wr;
416 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */
417
418 struct mutex i_truncate_mutex;
419 u32 i_truncate_seq; /* last truncate to smaller size */
420 u64 i_truncate_size; /* and the size we last truncated down to */
421 int i_truncate_pending; /* still need to call vmtruncate */
422
423 u64 i_max_size; /* max file size authorized by mds */
424 u64 i_reported_size; /* (max_)size reported to or requested of mds */
425 u64 i_wanted_max_size; /* offset we'd like to write too */
426 u64 i_requested_max_size; /* max_size we've requested */
427
428 /* held references to caps */
429 int i_pin_ref;
430 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
431 int i_wrbuffer_ref, i_wrbuffer_ref_head;
432 atomic_t i_filelock_ref;
433 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */
434 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */
435 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
436
437 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
438 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */
439 spinlock_t i_unsafe_lock;
440
441 union {
442 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
443 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
444 };
445 struct list_head i_snap_realm_item;
446 struct list_head i_snap_flush_item;
447 struct timespec64 i_btime;
448 struct timespec64 i_snap_btime;
449
450 struct work_struct i_work;
451 unsigned long i_work_mask;
452 };
453
454 struct ceph_netfs_request_data {
455 int caps;
456
457 /*
458 * Maximum size of a file readahead request.
459 * The fadvise could update the bdi's default ra_pages.
460 */
461 unsigned int file_ra_pages;
462
463 /* Set it if fadvise disables file readahead entirely */
464 bool file_ra_disabled;
465 };
466
467 static inline struct ceph_inode_info *
ceph_inode(const struct inode * inode)468 ceph_inode(const struct inode *inode)
469 {
470 return container_of(inode, struct ceph_inode_info, netfs.inode);
471 }
472
473 static inline struct ceph_fs_client *
ceph_inode_to_client(const struct inode * inode)474 ceph_inode_to_client(const struct inode *inode)
475 {
476 return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
477 }
478
479 static inline struct ceph_fs_client *
ceph_sb_to_client(const struct super_block * sb)480 ceph_sb_to_client(const struct super_block *sb)
481 {
482 return (struct ceph_fs_client *)sb->s_fs_info;
483 }
484
485 static inline struct ceph_mds_client *
ceph_sb_to_mdsc(const struct super_block * sb)486 ceph_sb_to_mdsc(const struct super_block *sb)
487 {
488 return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
489 }
490
491 static inline struct ceph_vino
ceph_vino(const struct inode * inode)492 ceph_vino(const struct inode *inode)
493 {
494 return ceph_inode(inode)->i_vino;
495 }
496
ceph_ino_to_ino32(u64 vino)497 static inline u32 ceph_ino_to_ino32(u64 vino)
498 {
499 u32 ino = vino & 0xffffffff;
500 ino ^= vino >> 32;
501 if (!ino)
502 ino = 2;
503 return ino;
504 }
505
506 /*
507 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
508 * some arches. We generally do not use this value inside the ceph driver, but
509 * we do want to set it to something, so that generic vfs code has an
510 * appropriate value for tracepoints and the like.
511 */
ceph_vino_to_ino_t(struct ceph_vino vino)512 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
513 {
514 if (sizeof(ino_t) == sizeof(u32))
515 return ceph_ino_to_ino32(vino.ino);
516 return (ino_t)vino.ino;
517 }
518
519 /* for printf-style formatting */
520 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
521
ceph_ino(struct inode * inode)522 static inline u64 ceph_ino(struct inode *inode)
523 {
524 return ceph_inode(inode)->i_vino.ino;
525 }
526
ceph_snap(struct inode * inode)527 static inline u64 ceph_snap(struct inode *inode)
528 {
529 return ceph_inode(inode)->i_vino.snap;
530 }
531
532 /**
533 * ceph_present_ino - format an inode number for presentation to userland
534 * @sb: superblock where the inode lives
535 * @ino: inode number to (possibly) convert
536 *
537 * If the user mounted with the ino32 option, then the 64-bit value needs
538 * to be converted to something that can fit inside 32 bits. Note that
539 * internal kernel code never uses this value, so this is entirely for
540 * userland consumption.
541 */
ceph_present_ino(struct super_block * sb,u64 ino)542 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
543 {
544 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
545 return ceph_ino_to_ino32(ino);
546 return ino;
547 }
548
ceph_present_inode(struct inode * inode)549 static inline u64 ceph_present_inode(struct inode *inode)
550 {
551 return ceph_present_ino(inode->i_sb, ceph_ino(inode));
552 }
553
ceph_ino_compare(struct inode * inode,void * data)554 static inline int ceph_ino_compare(struct inode *inode, void *data)
555 {
556 struct ceph_vino *pvino = (struct ceph_vino *)data;
557 struct ceph_inode_info *ci = ceph_inode(inode);
558 return ci->i_vino.ino == pvino->ino &&
559 ci->i_vino.snap == pvino->snap;
560 }
561
562 /*
563 * The MDS reserves a set of inodes for its own usage. These should never
564 * be accessible by clients, and so the MDS has no reason to ever hand these
565 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
566 *
567 * These come from src/mds/mdstypes.h in the ceph sources.
568 */
569 #define CEPH_MAX_MDS 0x100
570 #define CEPH_NUM_STRAY 10
571 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS)
572 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS)
573 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
574
ceph_vino_is_reserved(const struct ceph_vino vino)575 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
576 {
577 if (vino.ino >= CEPH_INO_SYSTEM_BASE ||
578 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET)
579 return false;
580
581 /* Don't warn on mdsdirs */
582 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET,
583 "Attempt to access reserved inode number 0x%llx",
584 vino.ino);
585 return true;
586 }
587
ceph_find_inode(struct super_block * sb,struct ceph_vino vino)588 static inline struct inode *ceph_find_inode(struct super_block *sb,
589 struct ceph_vino vino)
590 {
591 if (ceph_vino_is_reserved(vino))
592 return NULL;
593
594 /*
595 * NB: The hashval will be run through the fs/inode.c hash function
596 * anyway, so there is no need to squash the inode number down to
597 * 32-bits first. Just use low-order bits on arches with 32-bit long.
598 */
599 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
600 }
601
602
603 /*
604 * Ceph inode.
605 */
606 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */
607 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */
608 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */
609 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */
610 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */
611 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */
612 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */
613 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */
614 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */
615 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */
616 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */
617 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */
618 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT)
619 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */
620
621 /*
622 * Masks of ceph inode work.
623 */
624 #define CEPH_I_WORK_WRITEBACK 0
625 #define CEPH_I_WORK_INVALIDATE_PAGES 1
626 #define CEPH_I_WORK_VMTRUNCATE 2
627 #define CEPH_I_WORK_CHECK_CAPS 3
628 #define CEPH_I_WORK_FLUSH_SNAPS 4
629
630 /*
631 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
632 * and then clear it when they start succeeding. Note that we do a lockless
633 * check first, and only take the lock if it looks like it needs to be changed.
634 * The write submission code just takes this as a hint, so we're not too
635 * worried if a few slip through in either direction.
636 */
ceph_set_error_write(struct ceph_inode_info * ci)637 static inline void ceph_set_error_write(struct ceph_inode_info *ci)
638 {
639 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
640 spin_lock(&ci->i_ceph_lock);
641 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
642 spin_unlock(&ci->i_ceph_lock);
643 }
644 }
645
ceph_clear_error_write(struct ceph_inode_info * ci)646 static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
647 {
648 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
649 spin_lock(&ci->i_ceph_lock);
650 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
651 spin_unlock(&ci->i_ceph_lock);
652 }
653 }
654
__ceph_dir_set_complete(struct ceph_inode_info * ci,long long release_count,long long ordered_count)655 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
656 long long release_count,
657 long long ordered_count)
658 {
659 /*
660 * Makes sure operations that setup readdir cache (update page
661 * cache and i_size) are strongly ordered w.r.t. the following
662 * atomic64_set() operations.
663 */
664 smp_mb();
665 atomic64_set(&ci->i_complete_seq[0], release_count);
666 atomic64_set(&ci->i_complete_seq[1], ordered_count);
667 }
668
__ceph_dir_clear_complete(struct ceph_inode_info * ci)669 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
670 {
671 atomic64_inc(&ci->i_release_count);
672 }
673
__ceph_dir_clear_ordered(struct ceph_inode_info * ci)674 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
675 {
676 atomic64_inc(&ci->i_ordered_count);
677 }
678
__ceph_dir_is_complete(struct ceph_inode_info * ci)679 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
680 {
681 return atomic64_read(&ci->i_complete_seq[0]) ==
682 atomic64_read(&ci->i_release_count);
683 }
684
__ceph_dir_is_complete_ordered(struct ceph_inode_info * ci)685 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
686 {
687 return atomic64_read(&ci->i_complete_seq[0]) ==
688 atomic64_read(&ci->i_release_count) &&
689 atomic64_read(&ci->i_complete_seq[1]) ==
690 atomic64_read(&ci->i_ordered_count);
691 }
692
ceph_dir_clear_complete(struct inode * inode)693 static inline void ceph_dir_clear_complete(struct inode *inode)
694 {
695 __ceph_dir_clear_complete(ceph_inode(inode));
696 }
697
ceph_dir_clear_ordered(struct inode * inode)698 static inline void ceph_dir_clear_ordered(struct inode *inode)
699 {
700 __ceph_dir_clear_ordered(ceph_inode(inode));
701 }
702
ceph_dir_is_complete_ordered(struct inode * inode)703 static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
704 {
705 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
706 smp_rmb();
707 return ret;
708 }
709
710 /* find a specific frag @f */
711 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
712 u32 f);
713
714 /*
715 * choose fragment for value @v. copy frag content to pfrag, if leaf
716 * exists
717 */
718 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
719 struct ceph_inode_frag *pfrag,
720 int *found);
721
ceph_dentry(const struct dentry * dentry)722 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
723 {
724 return (struct ceph_dentry_info *)dentry->d_fsdata;
725 }
726
727 /*
728 * caps helpers
729 */
__ceph_is_any_real_caps(struct ceph_inode_info * ci)730 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
731 {
732 return !RB_EMPTY_ROOT(&ci->i_caps);
733 }
734
735 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
736 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
737 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
738 int t);
739 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
740 struct ceph_cap *cap);
741
ceph_caps_issued(struct ceph_inode_info * ci)742 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
743 {
744 int issued;
745 spin_lock(&ci->i_ceph_lock);
746 issued = __ceph_caps_issued(ci, NULL);
747 spin_unlock(&ci->i_ceph_lock);
748 return issued;
749 }
750
ceph_caps_issued_mask_metric(struct ceph_inode_info * ci,int mask,int touch)751 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
752 int mask, int touch)
753 {
754 int r;
755 spin_lock(&ci->i_ceph_lock);
756 r = __ceph_caps_issued_mask_metric(ci, mask, touch);
757 spin_unlock(&ci->i_ceph_lock);
758 return r;
759 }
760
__ceph_caps_dirty(struct ceph_inode_info * ci)761 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
762 {
763 return ci->i_dirty_caps | ci->i_flushing_caps;
764 }
765 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
766 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
767 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
768 struct ceph_cap_flush **pcf);
769
770 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
771 struct ceph_cap *ocap, int mask);
772 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
773 extern int __ceph_caps_used(struct ceph_inode_info *ci);
774
__ceph_is_file_opened(struct ceph_inode_info * ci)775 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
776 {
777 return ci->i_nr_by_mode[0];
778 }
779 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
780 extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
781
782 /* what the mds thinks we want */
783 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
784
785 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
786 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
787 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
788 struct ceph_mount_options *fsopt);
789 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
790 struct ceph_cap_reservation *ctx, int need);
791 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
792 struct ceph_cap_reservation *ctx);
793 extern void ceph_reservation_status(struct ceph_fs_client *client,
794 int *total, int *avail, int *used,
795 int *reserved, int *min);
796 extern void change_auth_cap_ses(struct ceph_inode_info *ci,
797 struct ceph_mds_session *session);
798
799
800
801 /*
802 * we keep buffered readdir results attached to file->private_data
803 */
804 #define CEPH_F_SYNC 1
805 #define CEPH_F_ATEND 2
806
807 struct ceph_file_info {
808 short fmode; /* initialized on open */
809 short flags; /* CEPH_F_* */
810
811 spinlock_t rw_contexts_lock;
812 struct list_head rw_contexts;
813
814 u32 filp_gen;
815 };
816
817 struct ceph_dir_file_info {
818 struct ceph_file_info file_info;
819
820 /* readdir: position within the dir */
821 u32 frag;
822 struct ceph_mds_request *last_readdir;
823
824 /* readdir: position within a frag */
825 unsigned next_offset; /* offset of next chunk (last_name's + 1) */
826 char *last_name; /* last entry in previous chunk */
827 long long dir_release_count;
828 long long dir_ordered_count;
829 int readdir_cache_idx;
830
831 /* used for -o dirstat read() on directory thing */
832 char *dir_info;
833 int dir_info_len;
834 };
835
836 struct ceph_rw_context {
837 struct list_head list;
838 struct task_struct *thread;
839 int caps;
840 };
841
842 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \
843 struct ceph_rw_context _name = { \
844 .thread = current, \
845 .caps = _caps, \
846 }
847
ceph_add_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)848 static inline void ceph_add_rw_context(struct ceph_file_info *cf,
849 struct ceph_rw_context *ctx)
850 {
851 spin_lock(&cf->rw_contexts_lock);
852 list_add(&ctx->list, &cf->rw_contexts);
853 spin_unlock(&cf->rw_contexts_lock);
854 }
855
ceph_del_rw_context(struct ceph_file_info * cf,struct ceph_rw_context * ctx)856 static inline void ceph_del_rw_context(struct ceph_file_info *cf,
857 struct ceph_rw_context *ctx)
858 {
859 spin_lock(&cf->rw_contexts_lock);
860 list_del(&ctx->list);
861 spin_unlock(&cf->rw_contexts_lock);
862 }
863
864 static inline struct ceph_rw_context*
ceph_find_rw_context(struct ceph_file_info * cf)865 ceph_find_rw_context(struct ceph_file_info *cf)
866 {
867 struct ceph_rw_context *ctx, *found = NULL;
868 spin_lock(&cf->rw_contexts_lock);
869 list_for_each_entry(ctx, &cf->rw_contexts, list) {
870 if (ctx->thread == current) {
871 found = ctx;
872 break;
873 }
874 }
875 spin_unlock(&cf->rw_contexts_lock);
876 return found;
877 }
878
879 struct ceph_readdir_cache_control {
880 struct page *page;
881 struct dentry **dentries;
882 int index;
883 };
884
885 /*
886 * A "snap realm" describes a subset of the file hierarchy sharing
887 * the same set of snapshots that apply to it. The realms themselves
888 * are organized into a hierarchy, such that children inherit (some of)
889 * the snapshots of their parents.
890 *
891 * All inodes within the realm that have capabilities are linked into a
892 * per-realm list.
893 */
894 struct ceph_snap_realm {
895 u64 ino;
896 struct inode *inode;
897 atomic_t nref;
898 struct rb_node node;
899
900 u64 created, seq;
901 u64 parent_ino;
902 u64 parent_since; /* snapid when our current parent became so */
903
904 u64 *prior_parent_snaps; /* snaps inherited from any parents we */
905 u32 num_prior_parent_snaps; /* had prior to parent_since */
906 u64 *snaps; /* snaps specific to this realm */
907 u32 num_snaps;
908
909 struct ceph_snap_realm *parent;
910 struct list_head children; /* list of child realms */
911 struct list_head child_item;
912
913 struct list_head empty_item; /* if i have ref==0 */
914
915 struct list_head dirty_item; /* if realm needs new context */
916
917 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */
918
919 /* the current set of snaps for this realm */
920 struct ceph_snap_context *cached_context;
921
922 struct list_head inodes_with_caps;
923 spinlock_t inodes_with_caps_lock;
924 };
925
default_congestion_kb(void)926 static inline int default_congestion_kb(void)
927 {
928 int congestion_kb;
929
930 /*
931 * Copied from NFS
932 *
933 * congestion size, scale with available memory.
934 *
935 * 64MB: 8192k
936 * 128MB: 11585k
937 * 256MB: 16384k
938 * 512MB: 23170k
939 * 1GB: 32768k
940 * 2GB: 46340k
941 * 4GB: 65536k
942 * 8GB: 92681k
943 * 16GB: 131072k
944 *
945 * This allows larger machines to have larger/more transfers.
946 * Limit the default to 256M
947 */
948 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
949 if (congestion_kb > 256*1024)
950 congestion_kb = 256*1024;
951
952 return congestion_kb;
953 }
954
955
956 /* super.c */
957 extern int ceph_force_reconnect(struct super_block *sb);
958 /* snap.c */
959 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
960 u64 ino);
961 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
962 struct ceph_snap_realm *realm);
963 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
964 struct ceph_snap_realm *realm);
965 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
966 void *p, void *e, bool deletion,
967 struct ceph_snap_realm **realm_ret);
968 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
969 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
970 struct ceph_mds_session *session,
971 struct ceph_msg *msg);
972 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
973 struct ceph_cap_snap *capsnap);
974 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc);
975
976 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
977 u64 snap);
978 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
979 struct ceph_snapid_map *sm);
980 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
981 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
982 void ceph_umount_begin(struct super_block *sb);
983
984
985 /*
986 * a cap_snap is "pending" if it is still awaiting an in-progress
987 * sync write (that may/may not still update size, mtime, etc.).
988 */
__ceph_have_pending_cap_snap(struct ceph_inode_info * ci)989 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
990 {
991 return !list_empty(&ci->i_cap_snaps) &&
992 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
993 ci_item)->writing;
994 }
995
996 /* inode.c */
997 struct ceph_mds_reply_info_in;
998 struct ceph_mds_reply_dirfrag;
999
1000 extern const struct inode_operations ceph_file_iops;
1001
1002 extern struct inode *ceph_alloc_inode(struct super_block *sb);
1003 extern void ceph_evict_inode(struct inode *inode);
1004 extern void ceph_free_inode(struct inode *inode);
1005
1006 extern struct inode *ceph_get_inode(struct super_block *sb,
1007 struct ceph_vino vino);
1008 extern struct inode *ceph_get_snapdir(struct inode *parent);
1009 extern int ceph_fill_file_size(struct inode *inode, int issued,
1010 u32 truncate_seq, u64 truncate_size, u64 size);
1011 extern void ceph_fill_file_time(struct inode *inode, int issued,
1012 u64 time_warp_seq, struct timespec64 *ctime,
1013 struct timespec64 *mtime,
1014 struct timespec64 *atime);
1015 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
1016 struct ceph_mds_reply_info_in *iinfo,
1017 struct ceph_mds_reply_dirfrag *dirinfo,
1018 struct ceph_mds_session *session, int cap_fmode,
1019 struct ceph_cap_reservation *caps_reservation);
1020 extern int ceph_fill_trace(struct super_block *sb,
1021 struct ceph_mds_request *req);
1022 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
1023 struct ceph_mds_session *session);
1024
1025 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
1026
1027 extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
1028 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
1029
1030 void ceph_queue_inode_work(struct inode *inode, int work_bit);
1031
ceph_queue_vmtruncate(struct inode * inode)1032 static inline void ceph_queue_vmtruncate(struct inode *inode)
1033 {
1034 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
1035 }
1036
ceph_queue_invalidate(struct inode * inode)1037 static inline void ceph_queue_invalidate(struct inode *inode)
1038 {
1039 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1040 }
1041
ceph_queue_writeback(struct inode * inode)1042 static inline void ceph_queue_writeback(struct inode *inode)
1043 {
1044 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1045 }
1046
ceph_queue_check_caps(struct inode * inode)1047 static inline void ceph_queue_check_caps(struct inode *inode)
1048 {
1049 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1050 }
1051
ceph_queue_flush_snaps(struct inode * inode)1052 static inline void ceph_queue_flush_snaps(struct inode *inode)
1053 {
1054 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1055 }
1056
1057 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask);
1058 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1059 int mask, bool force);
ceph_do_getattr(struct inode * inode,int mask,bool force)1060 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1061 {
1062 return __ceph_do_getattr(inode, NULL, mask, force);
1063 }
1064 extern int ceph_permission(struct user_namespace *mnt_userns,
1065 struct inode *inode, int mask);
1066 extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1067 extern int ceph_setattr(struct user_namespace *mnt_userns,
1068 struct dentry *dentry, struct iattr *attr);
1069 extern int ceph_getattr(struct user_namespace *mnt_userns,
1070 const struct path *path, struct kstat *stat,
1071 u32 request_mask, unsigned int flags);
1072 void ceph_inode_shutdown(struct inode *inode);
1073
ceph_inode_is_shutdown(struct inode * inode)1074 static inline bool ceph_inode_is_shutdown(struct inode *inode)
1075 {
1076 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);
1077 struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1078 int state = READ_ONCE(fsc->mount_state);
1079
1080 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN;
1081 }
1082
1083 /* xattr.c */
1084 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1085 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size);
1086 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1087 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1088 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1089 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1090 extern const struct xattr_handler *ceph_xattr_handlers[];
1091
1092 struct ceph_acl_sec_ctx {
1093 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1094 void *default_acl;
1095 void *acl;
1096 #endif
1097 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1098 void *sec_ctx;
1099 u32 sec_ctxlen;
1100 #endif
1101 struct ceph_pagelist *pagelist;
1102 };
1103
1104 #ifdef CONFIG_SECURITY
1105 extern bool ceph_security_xattr_deadlock(struct inode *in);
1106 extern bool ceph_security_xattr_wanted(struct inode *in);
1107 #else
ceph_security_xattr_deadlock(struct inode * in)1108 static inline bool ceph_security_xattr_deadlock(struct inode *in)
1109 {
1110 return false;
1111 }
ceph_security_xattr_wanted(struct inode * in)1112 static inline bool ceph_security_xattr_wanted(struct inode *in)
1113 {
1114 return false;
1115 }
1116 #endif
1117
1118 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1119 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1120 struct ceph_acl_sec_ctx *ctx);
ceph_security_invalidate_secctx(struct inode * inode)1121 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1122 {
1123 security_inode_invalidate_secctx(inode);
1124 }
1125 #else
ceph_security_init_secctx(struct dentry * dentry,umode_t mode,struct ceph_acl_sec_ctx * ctx)1126 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1127 struct ceph_acl_sec_ctx *ctx)
1128 {
1129 return 0;
1130 }
ceph_security_invalidate_secctx(struct inode * inode)1131 static inline void ceph_security_invalidate_secctx(struct inode *inode)
1132 {
1133 }
1134 #endif
1135
1136 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1137
1138 /* acl.c */
1139 #ifdef CONFIG_CEPH_FS_POSIX_ACL
1140
1141 struct posix_acl *ceph_get_acl(struct inode *, int, bool);
1142 int ceph_set_acl(struct user_namespace *mnt_userns,
1143 struct inode *inode, struct posix_acl *acl, int type);
1144 int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1145 struct ceph_acl_sec_ctx *as_ctx);
1146 void ceph_init_inode_acls(struct inode *inode,
1147 struct ceph_acl_sec_ctx *as_ctx);
1148
ceph_forget_all_cached_acls(struct inode * inode)1149 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1150 {
1151 forget_all_cached_acls(inode);
1152 }
1153
1154 #else
1155
1156 #define ceph_get_acl NULL
1157 #define ceph_set_acl NULL
1158
ceph_pre_init_acls(struct inode * dir,umode_t * mode,struct ceph_acl_sec_ctx * as_ctx)1159 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1160 struct ceph_acl_sec_ctx *as_ctx)
1161 {
1162 return 0;
1163 }
ceph_init_inode_acls(struct inode * inode,struct ceph_acl_sec_ctx * as_ctx)1164 static inline void ceph_init_inode_acls(struct inode *inode,
1165 struct ceph_acl_sec_ctx *as_ctx)
1166 {
1167 }
ceph_acl_chmod(struct dentry * dentry,struct inode * inode)1168 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1169 {
1170 return 0;
1171 }
1172
ceph_forget_all_cached_acls(struct inode * inode)1173 static inline void ceph_forget_all_cached_acls(struct inode *inode)
1174 {
1175 }
1176
1177 #endif
1178
1179 /* caps.c */
1180 extern const char *ceph_cap_string(int c);
1181 extern void ceph_handle_caps(struct ceph_mds_session *session,
1182 struct ceph_msg *msg);
1183 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1184 struct ceph_cap_reservation *ctx);
1185 extern void ceph_add_cap(struct inode *inode,
1186 struct ceph_mds_session *session, u64 cap_id,
1187 unsigned issued, unsigned wanted,
1188 unsigned cap, unsigned seq, u64 realmino, int flags,
1189 struct ceph_cap **new_cap);
1190 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1191 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1192 extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1193 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1194 struct ceph_cap *cap);
1195 extern int ceph_is_any_caps(struct inode *inode);
1196
1197 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1198 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1199 int datasync);
1200 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1201 struct ceph_mds_session *session);
1202 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1203 struct ceph_mds_session *session);
1204 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1205 struct ceph_inode_info *ci);
1206 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci,
1207 int mds);
1208 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1209 int mds);
1210 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1211 bool snap_rwsem_locked);
1212 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1213 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1214 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1215 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1216 int had);
1217 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1218 struct ceph_snap_context *snapc);
1219 extern void __ceph_remove_capsnap(struct inode *inode,
1220 struct ceph_cap_snap *capsnap,
1221 bool *wake_ci, bool *wake_mdsc);
1222 extern void ceph_remove_capsnap(struct inode *inode,
1223 struct ceph_cap_snap *capsnap,
1224 bool *wake_ci, bool *wake_mdsc);
1225 extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1226 struct ceph_mds_session **psession);
1227 extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1228 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1229 struct ceph_mds_session *session);
1230 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1231 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1232 extern int ceph_drop_caps_for_unlink(struct inode *inode);
1233 extern int ceph_encode_inode_release(void **p, struct inode *inode,
1234 int mds, int drop, int unless, int force);
1235 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1236 struct inode *dir,
1237 int mds, int drop, int unless);
1238
1239 extern int ceph_get_caps(struct file *filp, int need, int want,
1240 loff_t endoff, int *got);
1241 extern int ceph_try_get_caps(struct inode *inode,
1242 int need, int want, bool nonblock, int *got);
1243
1244 /* for counting open files by mode */
1245 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1246 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1247 extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1248 struct ceph_mds_client *mdsc, int fmode);
1249
1250 /* addr.c */
1251 extern const struct address_space_operations ceph_aops;
1252 extern const struct netfs_request_ops ceph_netfs_ops;
1253 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1254 extern int ceph_uninline_data(struct file *file);
1255 extern int ceph_pool_perm_check(struct inode *inode, int need);
1256 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1257 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate);
1258
ceph_has_inline_data(struct ceph_inode_info * ci)1259 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci)
1260 {
1261 if (ci->i_inline_version == CEPH_INLINE_NONE ||
1262 ci->i_inline_version == 1) /* initial version, no data */
1263 return false;
1264 return true;
1265 }
1266
1267 /* file.c */
1268 extern const struct file_operations ceph_file_fops;
1269
1270 extern int ceph_renew_caps(struct inode *inode, int fmode);
1271 extern int ceph_open(struct inode *inode, struct file *file);
1272 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1273 struct file *file, unsigned flags, umode_t mode);
1274 extern int ceph_release(struct inode *inode, struct file *filp);
1275 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1276 char *data, size_t len);
1277
1278 /* dir.c */
1279 extern const struct file_operations ceph_dir_fops;
1280 extern const struct file_operations ceph_snapdir_fops;
1281 extern const struct inode_operations ceph_dir_iops;
1282 extern const struct inode_operations ceph_snapdir_iops;
1283 extern const struct dentry_operations ceph_dentry_ops;
1284
1285 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1286 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1287 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1288 struct dentry *dentry);
1289 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1290 struct dentry *dentry, int err);
1291
1292 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1293 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1294 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1295 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1296 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1297 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1298
1299 /* ioctl.c */
1300 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1301
1302 /* export.c */
1303 extern const struct export_operations ceph_export_ops;
1304 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1305
1306 /* locks.c */
1307 extern __init void ceph_flock_init(void);
1308 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1309 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1310 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1311 extern int ceph_encode_locks_to_buffer(struct inode *inode,
1312 struct ceph_filelock *flocks,
1313 int num_fcntl_locks,
1314 int num_flock_locks);
1315 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1316 struct ceph_pagelist *pagelist,
1317 int num_fcntl_locks, int num_flock_locks);
1318
1319 /* debugfs.c */
1320 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1321 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1322
1323 /* quota.c */
1324
1325 enum quota_get_realm {
1326 QUOTA_GET_MAX_FILES,
1327 QUOTA_GET_MAX_BYTES,
1328 QUOTA_GET_ANY
1329 };
1330
__ceph_has_quota(struct ceph_inode_info * ci,enum quota_get_realm which)1331 static inline bool __ceph_has_quota(struct ceph_inode_info *ci,
1332 enum quota_get_realm which)
1333 {
1334 bool has_quota = false;
1335
1336 switch (which) {
1337 case QUOTA_GET_MAX_BYTES:
1338 has_quota = !!ci->i_max_bytes;
1339 break;
1340 case QUOTA_GET_MAX_FILES:
1341 has_quota = !!ci->i_max_files;
1342 break;
1343 default:
1344 has_quota = !!(ci->i_max_files || ci->i_max_bytes);
1345 }
1346 return has_quota;
1347 }
1348
1349 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1350
__ceph_update_quota(struct ceph_inode_info * ci,u64 max_bytes,u64 max_files)1351 static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1352 u64 max_bytes, u64 max_files)
1353 {
1354 bool had_quota, has_quota;
1355 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1356 ci->i_max_bytes = max_bytes;
1357 ci->i_max_files = max_files;
1358 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY);
1359
1360 if (had_quota != has_quota)
1361 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);
1362 }
1363
1364 extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1365 struct ceph_mds_session *session,
1366 struct ceph_msg *msg);
1367 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1368 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1369 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1370 loff_t newlen);
1371 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1372 loff_t newlen);
1373 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1374 struct kstatfs *buf);
1375 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1376
1377 bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc,
1378 struct ceph_mds_session *session);
1379 void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc);
1380 #endif /* _FS_CEPH_SUPER_H */
1381