Commit | Line | Data |
---|---|---|
de57606c SW |
1 | #ifndef _FS_CEPH_SUPER_H |
2 | #define _FS_CEPH_SUPER_H | |
3 | ||
4 | #include "ceph_debug.h" | |
5 | ||
6 | #include <asm/unaligned.h> | |
7 | #include <linux/backing-dev.h> | |
8 | #include <linux/completion.h> | |
9 | #include <linux/exportfs.h> | |
10 | #include <linux/fs.h> | |
11 | #include <linux/mempool.h> | |
12 | #include <linux/pagemap.h> | |
13 | #include <linux/wait.h> | |
14 | ||
15 | #include "types.h" | |
16 | #include "messenger.h" | |
17 | #include "msgpool.h" | |
18 | #include "mon_client.h" | |
19 | #include "mds_client.h" | |
20 | #include "osd_client.h" | |
21 | #include "ceph_fs.h" | |
22 | ||
23 | /* f_type in struct statfs */ | |
24 | #define CEPH_SUPER_MAGIC 0x00c36400 | |
25 | ||
26 | /* large granularity for statfs utilization stats to facilitate | |
27 | * large volume sizes on 32-bit machines. */ | |
28 | #define CEPH_BLOCK_SHIFT 20 /* 1 MB */ | |
29 | #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) | |
30 | ||
31 | /* | |
32 | * mount options | |
33 | */ | |
34 | #define CEPH_OPT_FSID (1<<0) | |
35 | #define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */ | |
36 | #define CEPH_OPT_MYIP (1<<2) /* specified my ip */ | |
37 | #define CEPH_OPT_DIRSTAT (1<<4) /* funky `cat dirname` for stats */ | |
38 | #define CEPH_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ | |
39 | #define CEPH_OPT_NOCRC (1<<6) /* no data crc on writes */ | |
40 | #define CEPH_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ | |
41 | ||
42 | #define CEPH_OPT_DEFAULT (CEPH_OPT_RBYTES) | |
43 | ||
44 | #define ceph_set_opt(client, opt) \ | |
6b805185 | 45 | (client)->mount_args->flags |= CEPH_OPT_##opt; |
de57606c | 46 | #define ceph_test_opt(client, opt) \ |
6b805185 | 47 | (!!((client)->mount_args->flags & CEPH_OPT_##opt)) |
de57606c SW |
48 | |
49 | ||
de57606c SW |
50 | struct ceph_mount_args { |
51 | int sb_flags; | |
6b805185 SW |
52 | int num_mon; |
53 | struct ceph_entity_addr *mon_addr; | |
de57606c SW |
54 | int flags; |
55 | int mount_timeout; | |
f5a2041b | 56 | int osd_idle_ttl; |
de57606c SW |
57 | int caps_wanted_delay_min, caps_wanted_delay_max; |
58 | struct ceph_fsid fsid; | |
59 | struct ceph_entity_addr my_addr; | |
60 | int wsize; | |
61 | int rsize; /* max readahead */ | |
62 | int max_readdir; /* max readdir size */ | |
2baba250 | 63 | int congestion_kb; /* max readdir size */ |
de57606c SW |
64 | int osd_timeout; |
65 | char *snapdir_name; /* default ".snap" */ | |
4e7a5dcd | 66 | char *name; |
de57606c SW |
67 | char *secret; |
68 | int cap_release_safety; | |
69 | }; | |
70 | ||
71 | /* | |
72 | * defaults | |
73 | */ | |
74 | #define CEPH_MOUNT_TIMEOUT_DEFAULT 60 | |
f5a2041b | 75 | #define CEPH_OSD_IDLE_TTL_DEFAULT 60 |
8fa97655 | 76 | #define CEPH_MOUNT_RSIZE_DEFAULT (512*1024) /* readahead */ |
de57606c SW |
77 | |
78 | #define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024) | |
79 | #define CEPH_MSG_MAX_DATA_LEN (16*1024*1024) | |
80 | ||
81 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" | |
4e7a5dcd | 82 | #define CEPH_AUTH_NAME_DEFAULT "guest" |
de57606c SW |
83 | |
84 | /* | |
85 | * Delay telling the MDS we no longer want caps, in case we reopen | |
86 | * the file. Delay a minimum amount of time, even if we send a cap | |
87 | * message for some other reason. Otherwise, take the oppotunity to | |
88 | * update the mds to avoid sending another message later. | |
89 | */ | |
90 | #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ | |
91 | #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ | |
92 | ||
93 | ||
94 | /* mount state */ | |
95 | enum { | |
96 | CEPH_MOUNT_MOUNTING, | |
97 | CEPH_MOUNT_MOUNTED, | |
98 | CEPH_MOUNT_UNMOUNTING, | |
99 | CEPH_MOUNT_UNMOUNTED, | |
100 | CEPH_MOUNT_SHUTDOWN, | |
101 | }; | |
102 | ||
103 | /* | |
104 | * subtract jiffies | |
105 | */ | |
106 | static inline unsigned long time_sub(unsigned long a, unsigned long b) | |
107 | { | |
108 | BUG_ON(time_after(b, a)); | |
109 | return (long)a - (long)b; | |
110 | } | |
111 | ||
112 | /* | |
113 | * per-filesystem client state | |
114 | * | |
115 | * possibly shared by multiple mount points, if they are | |
116 | * mounting the same ceph filesystem/cluster. | |
117 | */ | |
118 | struct ceph_client { | |
0743304d SW |
119 | struct ceph_fsid fsid; |
120 | bool have_fsid; | |
de57606c SW |
121 | |
122 | struct mutex mount_mutex; /* serialize mount attempts */ | |
6b805185 | 123 | struct ceph_mount_args *mount_args; |
de57606c SW |
124 | |
125 | struct super_block *sb; | |
126 | ||
127 | unsigned long mount_state; | |
9bd2e6f8 | 128 | wait_queue_head_t auth_wq; |
de57606c | 129 | |
9bd2e6f8 | 130 | int auth_err; |
de57606c | 131 | |
85ccce43 SW |
132 | int min_caps; /* min caps i added */ |
133 | ||
de57606c SW |
134 | struct ceph_messenger *msgr; /* messenger instance */ |
135 | struct ceph_mon_client monc; | |
136 | struct ceph_mds_client mdsc; | |
137 | struct ceph_osd_client osdc; | |
138 | ||
139 | /* writeback */ | |
140 | mempool_t *wb_pagevec_pool; | |
141 | struct workqueue_struct *wb_wq; | |
142 | struct workqueue_struct *pg_inv_wq; | |
143 | struct workqueue_struct *trunc_wq; | |
2baba250 | 144 | atomic_long_t writeback_count; |
de57606c SW |
145 | |
146 | struct backing_dev_info backing_dev_info; | |
0743304d SW |
147 | |
148 | #ifdef CONFIG_DEBUG_FS | |
149 | struct dentry *debugfs_monmap; | |
150 | struct dentry *debugfs_mdsmap, *debugfs_osdmap; | |
151 | struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps; | |
2baba250 | 152 | struct dentry *debugfs_congestion_kb; |
06edf046 | 153 | struct dentry *debugfs_bdi; |
0743304d | 154 | #endif |
de57606c SW |
155 | }; |
156 | ||
157 | static inline struct ceph_client *ceph_client(struct super_block *sb) | |
158 | { | |
159 | return sb->s_fs_info; | |
160 | } | |
161 | ||
162 | ||
163 | /* | |
164 | * File i/o capability. This tracks shared state with the metadata | |
165 | * server that allows us to cache or writeback attributes or to read | |
166 | * and write data. For any given inode, we should have one or more | |
167 | * capabilities, one issued by each metadata server, and our | |
168 | * cumulative access is the OR of all issued capabilities. | |
169 | * | |
170 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds | |
171 | * session capability lists. | |
172 | */ | |
173 | struct ceph_cap { | |
174 | struct ceph_inode_info *ci; | |
175 | struct rb_node ci_node; /* per-ci cap tree */ | |
176 | struct ceph_mds_session *session; | |
177 | struct list_head session_caps; /* per-session caplist */ | |
178 | int mds; | |
179 | u64 cap_id; /* unique cap id (mds provided) */ | |
180 | int issued; /* latest, from the mds */ | |
181 | int implemented; /* implemented superset of issued (for revocation) */ | |
182 | int mds_wanted; | |
685f9a5d SW |
183 | u32 seq, issue_seq, mseq; |
184 | u32 cap_gen; /* active/stale cycle */ | |
de57606c SW |
185 | unsigned long last_used; |
186 | struct list_head caps_item; | |
187 | }; | |
188 | ||
189 | #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ | |
190 | #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ | |
191 | #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ | |
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 | atomic_t nref; | |
200 | struct ceph_inode_info *ci; | |
201 | struct list_head ci_item, flushing_item; | |
202 | ||
203 | u64 follows, flush_tid; | |
204 | int issued, dirty; | |
205 | struct ceph_snap_context *context; | |
206 | ||
207 | mode_t mode; | |
208 | uid_t uid; | |
209 | gid_t gid; | |
210 | ||
211 | void *xattr_blob; | |
212 | int xattr_len; | |
213 | u64 xattr_version; | |
214 | ||
215 | u64 size; | |
216 | struct timespec mtime, atime, ctime; | |
217 | u64 time_warp_seq; | |
218 | int writing; /* a sync write is still in progress */ | |
219 | int dirty_pages; /* dirty pages awaiting writeback */ | |
220 | }; | |
221 | ||
222 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) | |
223 | { | |
224 | if (atomic_dec_and_test(&capsnap->nref)) | |
225 | kfree(capsnap); | |
226 | } | |
227 | ||
228 | /* | |
229 | * The frag tree describes how a directory is fragmented, potentially across | |
230 | * multiple metadata servers. It is also used to indicate points where | |
231 | * metadata authority is delegated, and whether/where metadata is replicated. | |
232 | * | |
233 | * A _leaf_ frag will be present in the i_fragtree IFF there is | |
234 | * delegation info. That is, if mds >= 0 || ndist > 0. | |
235 | */ | |
236 | #define CEPH_MAX_DIRFRAG_REP 4 | |
237 | ||
238 | struct ceph_inode_frag { | |
239 | struct rb_node node; | |
240 | ||
241 | /* fragtree state */ | |
242 | u32 frag; | |
243 | int split_by; /* i.e. 2^(split_by) children */ | |
244 | ||
245 | /* delegation and replication info */ | |
246 | int mds; /* -1 if same authority as parent */ | |
247 | int ndist; /* >0 if replicated */ | |
248 | int dist[CEPH_MAX_DIRFRAG_REP]; | |
249 | }; | |
250 | ||
251 | /* | |
252 | * We cache inode xattrs as an encoded blob until they are first used, | |
253 | * at which point we parse them into an rbtree. | |
254 | */ | |
255 | struct ceph_inode_xattr { | |
256 | struct rb_node node; | |
257 | ||
258 | const char *name; | |
259 | int name_len; | |
260 | const char *val; | |
261 | int val_len; | |
262 | int dirty; | |
263 | ||
264 | int should_free_name; | |
265 | int should_free_val; | |
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 | #define CEPH_I_COMPLETE 1 /* we have complete directory cached */ | |
290 | #define CEPH_I_NODELAY 4 /* do not delay cap release */ | |
291 | #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ | |
292 | ||
293 | struct ceph_inode_info { | |
294 | struct ceph_vino i_vino; /* ceph ino + snap */ | |
295 | ||
296 | u64 i_version; | |
297 | u32 i_time_warp_seq; | |
298 | ||
299 | unsigned i_ceph_flags; | |
300 | unsigned long i_release_count; | |
301 | ||
302 | struct ceph_file_layout i_layout; | |
303 | char *i_symlink; | |
304 | ||
305 | /* for dirs */ | |
306 | struct timespec i_rctime; | |
307 | u64 i_rbytes, i_rfiles, i_rsubdirs; | |
308 | u64 i_files, i_subdirs; | |
309 | u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */ | |
310 | ||
311 | struct rb_root i_fragtree; | |
312 | struct mutex i_fragtree_mutex; | |
313 | ||
314 | struct ceph_inode_xattrs_info i_xattrs; | |
315 | ||
316 | /* capabilities. protected _both_ by i_lock and cap->session's | |
317 | * s_mutex. */ | |
318 | struct rb_root i_caps; /* cap list */ | |
319 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ | |
320 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ | |
321 | struct list_head i_dirty_item, i_flushing_item; | |
322 | u64 i_cap_flush_seq; | |
323 | /* we need to track cap writeback on a per-cap-bit basis, to allow | |
324 | * overlapping, pipelined cap flushes to the mds. we can probably | |
325 | * reduce the tid to 8 bits if we're concerned about inode size. */ | |
326 | u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; | |
327 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ | |
328 | unsigned long i_hold_caps_min; /* jiffies */ | |
329 | unsigned long i_hold_caps_max; /* jiffies */ | |
330 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ | |
331 | int i_cap_exporting_mds; /* to handle cap migration between */ | |
332 | unsigned i_cap_exporting_mseq; /* mds's. */ | |
333 | unsigned i_cap_exporting_issued; | |
334 | struct ceph_cap_reservation i_cap_migration_resv; | |
335 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ | |
336 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */ | |
337 | unsigned i_snap_caps; /* cap bits for snapped files */ | |
338 | ||
339 | int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ | |
340 | ||
341 | u32 i_truncate_seq; /* last truncate to smaller size */ | |
342 | u64 i_truncate_size; /* and the size we last truncated down to */ | |
343 | int i_truncate_pending; /* still need to call vmtruncate */ | |
344 | ||
345 | u64 i_max_size; /* max file size authorized by mds */ | |
346 | u64 i_reported_size; /* (max_)size reported to or requested of mds */ | |
347 | u64 i_wanted_max_size; /* offset we'd like to write too */ | |
348 | u64 i_requested_max_size; /* max_size we've requested */ | |
349 | ||
350 | /* held references to caps */ | |
351 | int i_pin_ref; | |
352 | int i_rd_ref, i_rdcache_ref, i_wr_ref; | |
353 | int i_wrbuffer_ref, i_wrbuffer_ref_head; | |
354 | u32 i_shared_gen; /* increment each time we get FILE_SHARED */ | |
355 | u32 i_rdcache_gen; /* we increment this each time we get | |
356 | FILE_CACHE. If it's non-zero, we | |
357 | _may_ have cached pages. */ | |
358 | u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ | |
359 | ||
360 | struct list_head i_unsafe_writes; /* uncommitted sync writes */ | |
361 | struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ | |
362 | spinlock_t i_unsafe_lock; | |
363 | ||
364 | struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ | |
365 | int i_snap_realm_counter; /* snap realm (if caps) */ | |
366 | struct list_head i_snap_realm_item; | |
367 | struct list_head i_snap_flush_item; | |
368 | ||
369 | struct work_struct i_wb_work; /* writeback work */ | |
370 | struct work_struct i_pg_inv_work; /* page invalidation work */ | |
371 | ||
372 | struct work_struct i_vmtruncate_work; | |
373 | ||
374 | struct inode vfs_inode; /* at end */ | |
375 | }; | |
376 | ||
377 | static inline struct ceph_inode_info *ceph_inode(struct inode *inode) | |
378 | { | |
fbbccec9 | 379 | return container_of(inode, struct ceph_inode_info, vfs_inode); |
de57606c SW |
380 | } |
381 | ||
382 | static inline void ceph_i_clear(struct inode *inode, unsigned mask) | |
383 | { | |
384 | struct ceph_inode_info *ci = ceph_inode(inode); | |
385 | ||
386 | spin_lock(&inode->i_lock); | |
387 | ci->i_ceph_flags &= ~mask; | |
388 | spin_unlock(&inode->i_lock); | |
389 | } | |
390 | ||
391 | static inline void ceph_i_set(struct inode *inode, unsigned mask) | |
392 | { | |
393 | struct ceph_inode_info *ci = ceph_inode(inode); | |
394 | ||
395 | spin_lock(&inode->i_lock); | |
396 | ci->i_ceph_flags |= mask; | |
397 | spin_unlock(&inode->i_lock); | |
398 | } | |
399 | ||
400 | static inline bool ceph_i_test(struct inode *inode, unsigned mask) | |
401 | { | |
402 | struct ceph_inode_info *ci = ceph_inode(inode); | |
403 | bool r; | |
404 | ||
405 | smp_mb(); | |
406 | r = (ci->i_ceph_flags & mask) == mask; | |
407 | return r; | |
408 | } | |
409 | ||
410 | ||
411 | /* find a specific frag @f */ | |
412 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, | |
413 | u32 f); | |
414 | ||
415 | /* | |
416 | * choose fragment for value @v. copy frag content to pfrag, if leaf | |
417 | * exists | |
418 | */ | |
419 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, | |
420 | struct ceph_inode_frag *pfrag, | |
421 | int *found); | |
422 | ||
423 | /* | |
424 | * Ceph dentry state | |
425 | */ | |
426 | struct ceph_dentry_info { | |
427 | struct ceph_mds_session *lease_session; | |
428 | u32 lease_gen, lease_shared_gen; | |
429 | u32 lease_seq; | |
430 | unsigned long lease_renew_after, lease_renew_from; | |
431 | struct list_head lru; | |
432 | struct dentry *dentry; | |
433 | u64 time; | |
434 | u64 offset; | |
435 | }; | |
436 | ||
437 | static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) | |
438 | { | |
439 | return (struct ceph_dentry_info *)dentry->d_fsdata; | |
440 | } | |
441 | ||
442 | static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) | |
443 | { | |
444 | return ((loff_t)frag << 32) | (loff_t)off; | |
445 | } | |
446 | ||
447 | /* | |
448 | * ino_t is <64 bits on many architectures, blech. | |
449 | * | |
450 | * don't include snap in ino hash, at least for now. | |
451 | */ | |
452 | static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) | |
453 | { | |
454 | ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */ | |
455 | #if BITS_PER_LONG == 32 | |
456 | ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8; | |
457 | if (!ino) | |
458 | ino = 1; | |
459 | #endif | |
460 | return ino; | |
461 | } | |
462 | ||
463 | static inline int ceph_set_ino_cb(struct inode *inode, void *data) | |
464 | { | |
465 | ceph_inode(inode)->i_vino = *(struct ceph_vino *)data; | |
466 | inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data); | |
467 | return 0; | |
468 | } | |
469 | ||
470 | static inline struct ceph_vino ceph_vino(struct inode *inode) | |
471 | { | |
472 | return ceph_inode(inode)->i_vino; | |
473 | } | |
474 | ||
475 | /* for printf-style formatting */ | |
476 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap | |
477 | ||
478 | static inline u64 ceph_ino(struct inode *inode) | |
479 | { | |
480 | return ceph_inode(inode)->i_vino.ino; | |
481 | } | |
482 | static inline u64 ceph_snap(struct inode *inode) | |
483 | { | |
484 | return ceph_inode(inode)->i_vino.snap; | |
485 | } | |
486 | ||
487 | static inline int ceph_ino_compare(struct inode *inode, void *data) | |
488 | { | |
489 | struct ceph_vino *pvino = (struct ceph_vino *)data; | |
490 | struct ceph_inode_info *ci = ceph_inode(inode); | |
491 | return ci->i_vino.ino == pvino->ino && | |
492 | ci->i_vino.snap == pvino->snap; | |
493 | } | |
494 | ||
495 | static inline struct inode *ceph_find_inode(struct super_block *sb, | |
496 | struct ceph_vino vino) | |
497 | { | |
498 | ino_t t = ceph_vino_to_ino(vino); | |
499 | return ilookup5(sb, t, ceph_ino_compare, &vino); | |
500 | } | |
501 | ||
502 | ||
503 | /* | |
504 | * caps helpers | |
505 | */ | |
506 | static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) | |
507 | { | |
508 | return !RB_EMPTY_ROOT(&ci->i_caps); | |
509 | } | |
510 | ||
511 | extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); | |
512 | extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); | |
513 | extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, | |
514 | struct ceph_cap *cap); | |
515 | ||
516 | static inline int ceph_caps_issued(struct ceph_inode_info *ci) | |
517 | { | |
518 | int issued; | |
519 | spin_lock(&ci->vfs_inode.i_lock); | |
520 | issued = __ceph_caps_issued(ci, NULL); | |
521 | spin_unlock(&ci->vfs_inode.i_lock); | |
522 | return issued; | |
523 | } | |
524 | ||
525 | static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, | |
526 | int touch) | |
527 | { | |
528 | int r; | |
529 | spin_lock(&ci->vfs_inode.i_lock); | |
530 | r = __ceph_caps_issued_mask(ci, mask, touch); | |
531 | spin_unlock(&ci->vfs_inode.i_lock); | |
532 | return r; | |
533 | } | |
534 | ||
535 | static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) | |
536 | { | |
537 | return ci->i_dirty_caps | ci->i_flushing_caps; | |
538 | } | |
afcdaea3 | 539 | extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); |
de57606c SW |
540 | |
541 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); | |
542 | extern int __ceph_caps_used(struct ceph_inode_info *ci); | |
543 | ||
544 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); | |
545 | ||
546 | /* | |
547 | * wanted, by virtue of open file modes AND cap refs (buffered/cached data) | |
548 | */ | |
549 | static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) | |
550 | { | |
551 | int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); | |
552 | if (w & CEPH_CAP_FILE_BUFFER) | |
553 | w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ | |
554 | return w; | |
555 | } | |
556 | ||
557 | /* what the mds thinks we want */ | |
558 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); | |
559 | ||
560 | extern void ceph_caps_init(void); | |
561 | extern void ceph_caps_finalize(void); | |
85ccce43 | 562 | extern void ceph_adjust_min_caps(int delta); |
de57606c SW |
563 | extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need); |
564 | extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx); | |
565 | extern void ceph_reservation_status(struct ceph_client *client, | |
566 | int *total, int *avail, int *used, | |
85ccce43 | 567 | int *reserved, int *min); |
de57606c SW |
568 | |
569 | static inline struct ceph_client *ceph_inode_to_client(struct inode *inode) | |
570 | { | |
571 | return (struct ceph_client *)inode->i_sb->s_fs_info; | |
572 | } | |
573 | ||
574 | static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb) | |
575 | { | |
576 | return (struct ceph_client *)sb->s_fs_info; | |
577 | } | |
578 | ||
de57606c SW |
579 | |
580 | /* | |
581 | * we keep buffered readdir results attached to file->private_data | |
582 | */ | |
583 | struct ceph_file_info { | |
584 | int fmode; /* initialized on open */ | |
585 | ||
586 | /* readdir: position within the dir */ | |
587 | u32 frag; | |
588 | struct ceph_mds_request *last_readdir; | |
589 | int at_end; | |
590 | ||
591 | /* readdir: position within a frag */ | |
592 | unsigned offset; /* offset of last chunk, adjusted for . and .. */ | |
593 | u64 next_offset; /* offset of next chunk (last_name's + 1) */ | |
594 | char *last_name; /* last entry in previous chunk */ | |
595 | struct dentry *dentry; /* next dentry (for dcache readdir) */ | |
596 | unsigned long dir_release_count; | |
597 | ||
598 | /* used for -o dirstat read() on directory thing */ | |
599 | char *dir_info; | |
600 | int dir_info_len; | |
601 | }; | |
602 | ||
603 | ||
604 | ||
605 | /* | |
606 | * snapshots | |
607 | */ | |
608 | ||
609 | /* | |
610 | * A "snap context" is the set of existing snapshots when we | |
611 | * write data. It is used by the OSD to guide its COW behavior. | |
612 | * | |
613 | * The ceph_snap_context is refcounted, and attached to each dirty | |
614 | * page, indicating which context the dirty data belonged when it was | |
615 | * dirtied. | |
616 | */ | |
617 | struct ceph_snap_context { | |
618 | atomic_t nref; | |
619 | u64 seq; | |
620 | int num_snaps; | |
621 | u64 snaps[]; | |
622 | }; | |
623 | ||
624 | static inline struct ceph_snap_context * | |
625 | ceph_get_snap_context(struct ceph_snap_context *sc) | |
626 | { | |
627 | /* | |
628 | printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), | |
629 | atomic_read(&sc->nref)+1); | |
630 | */ | |
631 | if (sc) | |
632 | atomic_inc(&sc->nref); | |
633 | return sc; | |
634 | } | |
635 | ||
636 | static inline void ceph_put_snap_context(struct ceph_snap_context *sc) | |
637 | { | |
638 | if (!sc) | |
639 | return; | |
640 | /* | |
641 | printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), | |
642 | atomic_read(&sc->nref)-1); | |
643 | */ | |
644 | if (atomic_dec_and_test(&sc->nref)) { | |
645 | /*printk(" deleting snap_context %p\n", sc);*/ | |
646 | kfree(sc); | |
647 | } | |
648 | } | |
649 | ||
650 | /* | |
651 | * A "snap realm" describes a subset of the file hierarchy sharing | |
652 | * the same set of snapshots that apply to it. The realms themselves | |
653 | * are organized into a hierarchy, such that children inherit (some of) | |
654 | * the snapshots of their parents. | |
655 | * | |
656 | * All inodes within the realm that have capabilities are linked into a | |
657 | * per-realm list. | |
658 | */ | |
659 | struct ceph_snap_realm { | |
660 | u64 ino; | |
661 | atomic_t nref; | |
a105f00c SW |
662 | struct rb_node node; |
663 | ||
de57606c SW |
664 | u64 created, seq; |
665 | u64 parent_ino; | |
666 | u64 parent_since; /* snapid when our current parent became so */ | |
667 | ||
668 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ | |
669 | int num_prior_parent_snaps; /* had prior to parent_since */ | |
670 | u64 *snaps; /* snaps specific to this realm */ | |
671 | int num_snaps; | |
672 | ||
673 | struct ceph_snap_realm *parent; | |
674 | struct list_head children; /* list of child realms */ | |
675 | struct list_head child_item; | |
676 | ||
677 | struct list_head empty_item; /* if i have ref==0 */ | |
678 | ||
679 | /* the current set of snaps for this realm */ | |
680 | struct ceph_snap_context *cached_context; | |
681 | ||
682 | struct list_head inodes_with_caps; | |
683 | spinlock_t inodes_with_caps_lock; | |
684 | }; | |
685 | ||
686 | ||
687 | ||
688 | /* | |
689 | * calculate the number of pages a given length and offset map onto, | |
690 | * if we align the data. | |
691 | */ | |
692 | static inline int calc_pages_for(u64 off, u64 len) | |
693 | { | |
694 | return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) - | |
695 | (off >> PAGE_CACHE_SHIFT); | |
696 | } | |
697 | ||
698 | ||
699 | ||
700 | /* snap.c */ | |
701 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, | |
702 | u64 ino); | |
703 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, | |
704 | struct ceph_snap_realm *realm); | |
705 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, | |
706 | struct ceph_snap_realm *realm); | |
707 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, | |
708 | void *p, void *e, bool deletion); | |
709 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, | |
710 | struct ceph_msg *msg); | |
711 | extern void ceph_queue_cap_snap(struct ceph_inode_info *ci, | |
712 | struct ceph_snap_context *snapc); | |
713 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, | |
714 | struct ceph_cap_snap *capsnap); | |
715 | extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); | |
716 | ||
717 | /* | |
718 | * a cap_snap is "pending" if it is still awaiting an in-progress | |
719 | * sync write (that may/may not still update size, mtime, etc.). | |
720 | */ | |
721 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) | |
722 | { | |
723 | return !list_empty(&ci->i_cap_snaps) && | |
724 | list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, | |
725 | ci_item)->writing; | |
726 | } | |
727 | ||
728 | ||
729 | /* super.c */ | |
730 | extern struct kmem_cache *ceph_inode_cachep; | |
731 | extern struct kmem_cache *ceph_cap_cachep; | |
732 | extern struct kmem_cache *ceph_dentry_cachep; | |
733 | extern struct kmem_cache *ceph_file_cachep; | |
734 | ||
735 | extern const char *ceph_msg_type_name(int type); | |
0743304d | 736 | extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid); |
de57606c SW |
737 | |
738 | #define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \ | |
739 | "%02x%02x%02x%02x%02x%02x" | |
740 | #define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \ | |
741 | (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7], \ | |
742 | (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11], \ | |
743 | (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15] | |
744 | ||
745 | /* inode.c */ | |
746 | extern const struct inode_operations ceph_file_iops; | |
747 | ||
748 | extern struct inode *ceph_alloc_inode(struct super_block *sb); | |
749 | extern void ceph_destroy_inode(struct inode *inode); | |
750 | ||
751 | extern struct inode *ceph_get_inode(struct super_block *sb, | |
752 | struct ceph_vino vino); | |
753 | extern struct inode *ceph_get_snapdir(struct inode *parent); | |
754 | extern int ceph_fill_file_size(struct inode *inode, int issued, | |
755 | u32 truncate_seq, u64 truncate_size, u64 size); | |
756 | extern void ceph_fill_file_time(struct inode *inode, int issued, | |
757 | u64 time_warp_seq, struct timespec *ctime, | |
758 | struct timespec *mtime, struct timespec *atime); | |
759 | extern int ceph_fill_trace(struct super_block *sb, | |
760 | struct ceph_mds_request *req, | |
761 | struct ceph_mds_session *session); | |
762 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, | |
763 | struct ceph_mds_session *session); | |
764 | ||
765 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); | |
766 | ||
767 | extern int ceph_inode_set_size(struct inode *inode, loff_t size); | |
de57606c | 768 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); |
3c6f6b79 SW |
769 | extern void ceph_queue_vmtruncate(struct inode *inode); |
770 | ||
771 | extern void ceph_queue_invalidate(struct inode *inode); | |
772 | extern void ceph_queue_writeback(struct inode *inode); | |
de57606c SW |
773 | |
774 | extern int ceph_do_getattr(struct inode *inode, int mask); | |
775 | extern int ceph_permission(struct inode *inode, int mask); | |
776 | extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); | |
777 | extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, | |
778 | struct kstat *stat); | |
779 | ||
780 | /* xattr.c */ | |
781 | extern int ceph_setxattr(struct dentry *, const char *, const void *, | |
782 | size_t, int); | |
783 | extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); | |
784 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); | |
785 | extern int ceph_removexattr(struct dentry *, const char *); | |
786 | extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); | |
787 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); | |
788 | ||
789 | /* caps.c */ | |
790 | extern const char *ceph_cap_string(int c); | |
791 | extern void ceph_handle_caps(struct ceph_mds_session *session, | |
792 | struct ceph_msg *msg); | |
793 | extern int ceph_add_cap(struct inode *inode, | |
794 | struct ceph_mds_session *session, u64 cap_id, | |
795 | int fmode, unsigned issued, unsigned wanted, | |
796 | unsigned cap, unsigned seq, u64 realmino, int flags, | |
797 | struct ceph_cap_reservation *caps_reservation); | |
7c1332b8 | 798 | extern void __ceph_remove_cap(struct ceph_cap *cap); |
de57606c SW |
799 | static inline void ceph_remove_cap(struct ceph_cap *cap) |
800 | { | |
801 | struct inode *inode = &cap->ci->vfs_inode; | |
802 | spin_lock(&inode->i_lock); | |
7c1332b8 | 803 | __ceph_remove_cap(cap); |
de57606c SW |
804 | spin_unlock(&inode->i_lock); |
805 | } | |
7c1332b8 | 806 | extern void ceph_put_cap(struct ceph_cap *cap); |
de57606c SW |
807 | |
808 | extern void ceph_queue_caps_release(struct inode *inode); | |
809 | extern int ceph_write_inode(struct inode *inode, int unused); | |
810 | extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync); | |
811 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, | |
812 | struct ceph_mds_session *session); | |
813 | extern int ceph_get_cap_mds(struct inode *inode); | |
814 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); | |
815 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); | |
816 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, | |
817 | struct ceph_snap_context *snapc); | |
818 | extern void __ceph_flush_snaps(struct ceph_inode_info *ci, | |
819 | struct ceph_mds_session **psession); | |
820 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, | |
821 | struct ceph_mds_session *session); | |
afcdaea3 SW |
822 | extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); |
823 | extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); | |
de57606c SW |
824 | |
825 | extern int ceph_encode_inode_release(void **p, struct inode *inode, | |
826 | int mds, int drop, int unless, int force); | |
827 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, | |
828 | int mds, int drop, int unless); | |
829 | ||
830 | extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, | |
831 | int *got, loff_t endoff); | |
832 | ||
833 | /* for counting open files by mode */ | |
834 | static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) | |
835 | { | |
836 | ci->i_nr_by_mode[mode]++; | |
837 | } | |
838 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); | |
839 | ||
840 | /* addr.c */ | |
841 | extern const struct address_space_operations ceph_aops; | |
842 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); | |
843 | ||
844 | /* file.c */ | |
845 | extern const struct file_operations ceph_file_fops; | |
846 | extern const struct address_space_operations ceph_aops; | |
847 | extern int ceph_open(struct inode *inode, struct file *file); | |
848 | extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry, | |
849 | struct nameidata *nd, int mode, | |
850 | int locked_dir); | |
851 | extern int ceph_release(struct inode *inode, struct file *filp); | |
852 | extern void ceph_release_page_vector(struct page **pages, int num_pages); | |
853 | ||
854 | /* dir.c */ | |
855 | extern const struct file_operations ceph_dir_fops; | |
856 | extern const struct inode_operations ceph_dir_iops; | |
857 | extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, | |
858 | ceph_snapdir_dentry_ops; | |
859 | ||
860 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); | |
861 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, | |
862 | struct dentry *dentry, int err); | |
863 | ||
864 | extern void ceph_dentry_lru_add(struct dentry *dn); | |
865 | extern void ceph_dentry_lru_touch(struct dentry *dn); | |
866 | extern void ceph_dentry_lru_del(struct dentry *dn); | |
867 | ||
868 | /* | |
869 | * our d_ops vary depending on whether the inode is live, | |
870 | * snapshotted (read-only), or a virtual ".snap" directory. | |
871 | */ | |
872 | int ceph_init_dentry(struct dentry *dentry); | |
873 | ||
874 | ||
875 | /* ioctl.c */ | |
876 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | |
877 | ||
878 | /* export.c */ | |
879 | extern const struct export_operations ceph_export_ops; | |
880 | ||
881 | /* debugfs.c */ | |
882 | extern int ceph_debugfs_init(void); | |
883 | extern void ceph_debugfs_cleanup(void); | |
884 | extern int ceph_debugfs_client_init(struct ceph_client *client); | |
885 | extern void ceph_debugfs_client_cleanup(struct ceph_client *client); | |
886 | ||
887 | static inline struct inode *get_dentry_parent_inode(struct dentry *dentry) | |
888 | { | |
889 | if (dentry && dentry->d_parent) | |
890 | return dentry->d_parent->d_inode; | |
891 | ||
892 | return NULL; | |
893 | } | |
894 | ||
895 | #endif /* _FS_CEPH_SUPER_H */ |