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