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> | |
d45d0d97 | 13 | #include <linux/slab.h> |
de57606c | 14 | #include <linux/wait.h> |
f1a3d572 | 15 | #include <linux/writeback.h> |
5a0e3ad6 | 16 | #include <linux/slab.h> |
de57606c SW |
17 | |
18 | #include "types.h" | |
19 | #include "messenger.h" | |
20 | #include "msgpool.h" | |
21 | #include "mon_client.h" | |
22 | #include "mds_client.h" | |
23 | #include "osd_client.h" | |
24 | #include "ceph_fs.h" | |
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 20 /* 1 MB */ | |
32 | #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) | |
33 | ||
34 | /* | |
35 | * mount options | |
36 | */ | |
37 | #define CEPH_OPT_FSID (1<<0) | |
38 | #define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */ | |
39 | #define CEPH_OPT_MYIP (1<<2) /* specified my ip */ | |
40 | #define CEPH_OPT_DIRSTAT (1<<4) /* funky `cat dirname` for stats */ | |
41 | #define CEPH_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ | |
42 | #define CEPH_OPT_NOCRC (1<<6) /* no data crc on writes */ | |
43 | #define CEPH_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ | |
44 | ||
45 | #define CEPH_OPT_DEFAULT (CEPH_OPT_RBYTES) | |
46 | ||
47 | #define ceph_set_opt(client, opt) \ | |
6b805185 | 48 | (client)->mount_args->flags |= CEPH_OPT_##opt; |
de57606c | 49 | #define ceph_test_opt(client, opt) \ |
6b805185 | 50 | (!!((client)->mount_args->flags & CEPH_OPT_##opt)) |
de57606c SW |
51 | |
52 | ||
de57606c SW |
53 | struct ceph_mount_args { |
54 | int sb_flags; | |
6b805185 SW |
55 | int num_mon; |
56 | struct ceph_entity_addr *mon_addr; | |
de57606c SW |
57 | int flags; |
58 | int mount_timeout; | |
f5a2041b | 59 | int osd_idle_ttl; |
de57606c SW |
60 | int caps_wanted_delay_min, caps_wanted_delay_max; |
61 | struct ceph_fsid fsid; | |
62 | struct ceph_entity_addr my_addr; | |
63 | int wsize; | |
64 | int rsize; /* max readahead */ | |
65 | int max_readdir; /* max readdir size */ | |
2baba250 | 66 | int congestion_kb; /* max readdir size */ |
de57606c | 67 | int osd_timeout; |
422d2cb8 | 68 | int osd_keepalive_timeout; |
de57606c | 69 | char *snapdir_name; /* default ".snap" */ |
4e7a5dcd | 70 | char *name; |
de57606c SW |
71 | char *secret; |
72 | int cap_release_safety; | |
73 | }; | |
74 | ||
75 | /* | |
76 | * defaults | |
77 | */ | |
78 | #define CEPH_MOUNT_TIMEOUT_DEFAULT 60 | |
422d2cb8 YS |
79 | #define CEPH_OSD_TIMEOUT_DEFAULT 60 /* seconds */ |
80 | #define CEPH_OSD_KEEPALIVE_DEFAULT 5 | |
f5a2041b | 81 | #define CEPH_OSD_IDLE_TTL_DEFAULT 60 |
8fa97655 | 82 | #define CEPH_MOUNT_RSIZE_DEFAULT (512*1024) /* readahead */ |
de57606c SW |
83 | |
84 | #define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024) | |
85 | #define CEPH_MSG_MAX_DATA_LEN (16*1024*1024) | |
86 | ||
87 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" | |
4e7a5dcd | 88 | #define CEPH_AUTH_NAME_DEFAULT "guest" |
de57606c SW |
89 | |
90 | /* | |
91 | * Delay telling the MDS we no longer want caps, in case we reopen | |
92 | * the file. Delay a minimum amount of time, even if we send a cap | |
93 | * message for some other reason. Otherwise, take the oppotunity to | |
94 | * update the mds to avoid sending another message later. | |
95 | */ | |
96 | #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ | |
97 | #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ | |
98 | ||
99 | ||
100 | /* mount state */ | |
101 | enum { | |
102 | CEPH_MOUNT_MOUNTING, | |
103 | CEPH_MOUNT_MOUNTED, | |
104 | CEPH_MOUNT_UNMOUNTING, | |
105 | CEPH_MOUNT_UNMOUNTED, | |
106 | CEPH_MOUNT_SHUTDOWN, | |
107 | }; | |
108 | ||
109 | /* | |
110 | * subtract jiffies | |
111 | */ | |
112 | static inline unsigned long time_sub(unsigned long a, unsigned long b) | |
113 | { | |
114 | BUG_ON(time_after(b, a)); | |
115 | return (long)a - (long)b; | |
116 | } | |
117 | ||
118 | /* | |
119 | * per-filesystem client state | |
120 | * | |
121 | * possibly shared by multiple mount points, if they are | |
122 | * mounting the same ceph filesystem/cluster. | |
123 | */ | |
124 | struct ceph_client { | |
0743304d SW |
125 | struct ceph_fsid fsid; |
126 | bool have_fsid; | |
de57606c SW |
127 | |
128 | struct mutex mount_mutex; /* serialize mount attempts */ | |
6b805185 | 129 | struct ceph_mount_args *mount_args; |
de57606c SW |
130 | |
131 | struct super_block *sb; | |
132 | ||
133 | unsigned long mount_state; | |
9bd2e6f8 | 134 | wait_queue_head_t auth_wq; |
de57606c | 135 | |
9bd2e6f8 | 136 | int auth_err; |
de57606c | 137 | |
85ccce43 SW |
138 | int min_caps; /* min caps i added */ |
139 | ||
de57606c SW |
140 | struct ceph_messenger *msgr; /* messenger instance */ |
141 | struct ceph_mon_client monc; | |
142 | struct ceph_mds_client mdsc; | |
143 | struct ceph_osd_client osdc; | |
144 | ||
145 | /* writeback */ | |
146 | mempool_t *wb_pagevec_pool; | |
147 | struct workqueue_struct *wb_wq; | |
148 | struct workqueue_struct *pg_inv_wq; | |
149 | struct workqueue_struct *trunc_wq; | |
2baba250 | 150 | atomic_long_t writeback_count; |
de57606c SW |
151 | |
152 | struct backing_dev_info backing_dev_info; | |
0743304d SW |
153 | |
154 | #ifdef CONFIG_DEBUG_FS | |
155 | struct dentry *debugfs_monmap; | |
156 | struct dentry *debugfs_mdsmap, *debugfs_osdmap; | |
157 | struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps; | |
2baba250 | 158 | struct dentry *debugfs_congestion_kb; |
06edf046 | 159 | struct dentry *debugfs_bdi; |
0743304d | 160 | #endif |
de57606c SW |
161 | }; |
162 | ||
163 | static inline struct ceph_client *ceph_client(struct super_block *sb) | |
164 | { | |
165 | return sb->s_fs_info; | |
166 | } | |
167 | ||
168 | ||
169 | /* | |
170 | * File i/o capability. This tracks shared state with the metadata | |
171 | * server that allows us to cache or writeback attributes or to read | |
172 | * and write data. For any given inode, we should have one or more | |
173 | * capabilities, one issued by each metadata server, and our | |
174 | * cumulative access is the OR of all issued capabilities. | |
175 | * | |
176 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds | |
177 | * session capability lists. | |
178 | */ | |
179 | struct ceph_cap { | |
180 | struct ceph_inode_info *ci; | |
181 | struct rb_node ci_node; /* per-ci cap tree */ | |
182 | struct ceph_mds_session *session; | |
183 | struct list_head session_caps; /* per-session caplist */ | |
184 | int mds; | |
185 | u64 cap_id; /* unique cap id (mds provided) */ | |
186 | int issued; /* latest, from the mds */ | |
187 | int implemented; /* implemented superset of issued (for revocation) */ | |
188 | int mds_wanted; | |
685f9a5d SW |
189 | u32 seq, issue_seq, mseq; |
190 | u32 cap_gen; /* active/stale cycle */ | |
de57606c SW |
191 | unsigned long last_used; |
192 | struct list_head caps_item; | |
193 | }; | |
194 | ||
195 | #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ | |
196 | #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ | |
197 | #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ | |
198 | ||
199 | /* | |
200 | * Snapped cap state that is pending flush to mds. When a snapshot occurs, | |
201 | * we first complete any in-process sync writes and writeback any dirty | |
202 | * data before flushing the snapped state (tracked here) back to the MDS. | |
203 | */ | |
204 | struct ceph_cap_snap { | |
205 | atomic_t nref; | |
206 | struct ceph_inode_info *ci; | |
207 | struct list_head ci_item, flushing_item; | |
208 | ||
209 | u64 follows, flush_tid; | |
210 | int issued, dirty; | |
211 | struct ceph_snap_context *context; | |
212 | ||
213 | mode_t mode; | |
214 | uid_t uid; | |
215 | gid_t gid; | |
216 | ||
217 | void *xattr_blob; | |
218 | int xattr_len; | |
219 | u64 xattr_version; | |
220 | ||
221 | u64 size; | |
222 | struct timespec mtime, atime, ctime; | |
223 | u64 time_warp_seq; | |
224 | int writing; /* a sync write is still in progress */ | |
225 | int dirty_pages; /* dirty pages awaiting writeback */ | |
226 | }; | |
227 | ||
228 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) | |
229 | { | |
230 | if (atomic_dec_and_test(&capsnap->nref)) | |
231 | kfree(capsnap); | |
232 | } | |
233 | ||
234 | /* | |
235 | * The frag tree describes how a directory is fragmented, potentially across | |
236 | * multiple metadata servers. It is also used to indicate points where | |
237 | * metadata authority is delegated, and whether/where metadata is replicated. | |
238 | * | |
239 | * A _leaf_ frag will be present in the i_fragtree IFF there is | |
240 | * delegation info. That is, if mds >= 0 || ndist > 0. | |
241 | */ | |
242 | #define CEPH_MAX_DIRFRAG_REP 4 | |
243 | ||
244 | struct ceph_inode_frag { | |
245 | struct rb_node node; | |
246 | ||
247 | /* fragtree state */ | |
248 | u32 frag; | |
249 | int split_by; /* i.e. 2^(split_by) children */ | |
250 | ||
251 | /* delegation and replication info */ | |
252 | int mds; /* -1 if same authority as parent */ | |
253 | int ndist; /* >0 if replicated */ | |
254 | int dist[CEPH_MAX_DIRFRAG_REP]; | |
255 | }; | |
256 | ||
257 | /* | |
258 | * We cache inode xattrs as an encoded blob until they are first used, | |
259 | * at which point we parse them into an rbtree. | |
260 | */ | |
261 | struct ceph_inode_xattr { | |
262 | struct rb_node node; | |
263 | ||
264 | const char *name; | |
265 | int name_len; | |
266 | const char *val; | |
267 | int val_len; | |
268 | int dirty; | |
269 | ||
270 | int should_free_name; | |
271 | int should_free_val; | |
272 | }; | |
273 | ||
274 | struct ceph_inode_xattrs_info { | |
275 | /* | |
276 | * (still encoded) xattr blob. we avoid the overhead of parsing | |
277 | * this until someone actually calls getxattr, etc. | |
278 | * | |
279 | * blob->vec.iov_len == 4 implies there are no xattrs; blob == | |
280 | * NULL means we don't know. | |
281 | */ | |
282 | struct ceph_buffer *blob, *prealloc_blob; | |
283 | ||
284 | struct rb_root index; | |
285 | bool dirty; | |
286 | int count; | |
287 | int names_size; | |
288 | int vals_size; | |
289 | u64 version, index_version; | |
290 | }; | |
291 | ||
292 | /* | |
293 | * Ceph inode. | |
294 | */ | |
295 | #define CEPH_I_COMPLETE 1 /* we have complete directory cached */ | |
296 | #define CEPH_I_NODELAY 4 /* do not delay cap release */ | |
297 | #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ | |
e9964c10 | 298 | #define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ |
de57606c SW |
299 | |
300 | struct ceph_inode_info { | |
301 | struct ceph_vino i_vino; /* ceph ino + snap */ | |
302 | ||
303 | u64 i_version; | |
304 | u32 i_time_warp_seq; | |
305 | ||
306 | unsigned i_ceph_flags; | |
307 | unsigned long i_release_count; | |
308 | ||
309 | struct ceph_file_layout i_layout; | |
310 | char *i_symlink; | |
311 | ||
312 | /* for dirs */ | |
313 | struct timespec i_rctime; | |
314 | u64 i_rbytes, i_rfiles, i_rsubdirs; | |
315 | u64 i_files, i_subdirs; | |
316 | u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */ | |
317 | ||
318 | struct rb_root i_fragtree; | |
319 | struct mutex i_fragtree_mutex; | |
320 | ||
321 | struct ceph_inode_xattrs_info i_xattrs; | |
322 | ||
323 | /* capabilities. protected _both_ by i_lock and cap->session's | |
324 | * s_mutex. */ | |
325 | struct rb_root i_caps; /* cap list */ | |
326 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ | |
327 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ | |
328 | struct list_head i_dirty_item, i_flushing_item; | |
329 | u64 i_cap_flush_seq; | |
330 | /* we need to track cap writeback on a per-cap-bit basis, to allow | |
331 | * overlapping, pipelined cap flushes to the mds. we can probably | |
332 | * reduce the tid to 8 bits if we're concerned about inode size. */ | |
333 | u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; | |
334 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ | |
335 | unsigned long i_hold_caps_min; /* jiffies */ | |
336 | unsigned long i_hold_caps_max; /* jiffies */ | |
337 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ | |
338 | int i_cap_exporting_mds; /* to handle cap migration between */ | |
339 | unsigned i_cap_exporting_mseq; /* mds's. */ | |
340 | unsigned i_cap_exporting_issued; | |
341 | struct ceph_cap_reservation i_cap_migration_resv; | |
342 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ | |
343 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */ | |
344 | unsigned i_snap_caps; /* cap bits for snapped files */ | |
345 | ||
346 | int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ | |
347 | ||
348 | u32 i_truncate_seq; /* last truncate to smaller size */ | |
349 | u64 i_truncate_size; /* and the size we last truncated down to */ | |
350 | int i_truncate_pending; /* still need to call vmtruncate */ | |
351 | ||
352 | u64 i_max_size; /* max file size authorized by mds */ | |
353 | u64 i_reported_size; /* (max_)size reported to or requested of mds */ | |
354 | u64 i_wanted_max_size; /* offset we'd like to write too */ | |
355 | u64 i_requested_max_size; /* max_size we've requested */ | |
356 | ||
357 | /* held references to caps */ | |
358 | int i_pin_ref; | |
359 | int i_rd_ref, i_rdcache_ref, i_wr_ref; | |
360 | int i_wrbuffer_ref, i_wrbuffer_ref_head; | |
361 | u32 i_shared_gen; /* increment each time we get FILE_SHARED */ | |
362 | u32 i_rdcache_gen; /* we increment this each time we get | |
363 | FILE_CACHE. If it's non-zero, we | |
364 | _may_ have cached pages. */ | |
365 | u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ | |
366 | ||
367 | struct list_head i_unsafe_writes; /* uncommitted sync writes */ | |
368 | struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ | |
369 | spinlock_t i_unsafe_lock; | |
370 | ||
371 | struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ | |
372 | int i_snap_realm_counter; /* snap realm (if caps) */ | |
373 | struct list_head i_snap_realm_item; | |
374 | struct list_head i_snap_flush_item; | |
375 | ||
376 | struct work_struct i_wb_work; /* writeback work */ | |
377 | struct work_struct i_pg_inv_work; /* page invalidation work */ | |
378 | ||
379 | struct work_struct i_vmtruncate_work; | |
380 | ||
381 | struct inode vfs_inode; /* at end */ | |
382 | }; | |
383 | ||
384 | static inline struct ceph_inode_info *ceph_inode(struct inode *inode) | |
385 | { | |
fbbccec9 | 386 | return container_of(inode, struct ceph_inode_info, vfs_inode); |
de57606c SW |
387 | } |
388 | ||
389 | static inline void ceph_i_clear(struct inode *inode, unsigned mask) | |
390 | { | |
391 | struct ceph_inode_info *ci = ceph_inode(inode); | |
392 | ||
393 | spin_lock(&inode->i_lock); | |
394 | ci->i_ceph_flags &= ~mask; | |
395 | spin_unlock(&inode->i_lock); | |
396 | } | |
397 | ||
398 | static inline void ceph_i_set(struct inode *inode, unsigned mask) | |
399 | { | |
400 | struct ceph_inode_info *ci = ceph_inode(inode); | |
401 | ||
402 | spin_lock(&inode->i_lock); | |
403 | ci->i_ceph_flags |= mask; | |
404 | spin_unlock(&inode->i_lock); | |
405 | } | |
406 | ||
407 | static inline bool ceph_i_test(struct inode *inode, unsigned mask) | |
408 | { | |
409 | struct ceph_inode_info *ci = ceph_inode(inode); | |
410 | bool r; | |
411 | ||
412 | smp_mb(); | |
413 | r = (ci->i_ceph_flags & mask) == mask; | |
414 | return r; | |
415 | } | |
416 | ||
417 | ||
418 | /* find a specific frag @f */ | |
419 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, | |
420 | u32 f); | |
421 | ||
422 | /* | |
423 | * choose fragment for value @v. copy frag content to pfrag, if leaf | |
424 | * exists | |
425 | */ | |
426 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, | |
427 | struct ceph_inode_frag *pfrag, | |
428 | int *found); | |
429 | ||
430 | /* | |
431 | * Ceph dentry state | |
432 | */ | |
433 | struct ceph_dentry_info { | |
434 | struct ceph_mds_session *lease_session; | |
435 | u32 lease_gen, lease_shared_gen; | |
436 | u32 lease_seq; | |
437 | unsigned long lease_renew_after, lease_renew_from; | |
438 | struct list_head lru; | |
439 | struct dentry *dentry; | |
440 | u64 time; | |
441 | u64 offset; | |
442 | }; | |
443 | ||
444 | static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) | |
445 | { | |
446 | return (struct ceph_dentry_info *)dentry->d_fsdata; | |
447 | } | |
448 | ||
449 | static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) | |
450 | { | |
451 | return ((loff_t)frag << 32) | (loff_t)off; | |
452 | } | |
453 | ||
454 | /* | |
455 | * ino_t is <64 bits on many architectures, blech. | |
456 | * | |
457 | * don't include snap in ino hash, at least for now. | |
458 | */ | |
459 | static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) | |
460 | { | |
461 | ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */ | |
462 | #if BITS_PER_LONG == 32 | |
463 | ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8; | |
464 | if (!ino) | |
465 | ino = 1; | |
466 | #endif | |
467 | return ino; | |
468 | } | |
469 | ||
470 | static inline int ceph_set_ino_cb(struct inode *inode, void *data) | |
471 | { | |
472 | ceph_inode(inode)->i_vino = *(struct ceph_vino *)data; | |
473 | inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data); | |
474 | return 0; | |
475 | } | |
476 | ||
477 | static inline struct ceph_vino ceph_vino(struct inode *inode) | |
478 | { | |
479 | return ceph_inode(inode)->i_vino; | |
480 | } | |
481 | ||
482 | /* for printf-style formatting */ | |
483 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap | |
484 | ||
485 | static inline u64 ceph_ino(struct inode *inode) | |
486 | { | |
487 | return ceph_inode(inode)->i_vino.ino; | |
488 | } | |
489 | static inline u64 ceph_snap(struct inode *inode) | |
490 | { | |
491 | return ceph_inode(inode)->i_vino.snap; | |
492 | } | |
493 | ||
494 | static inline int ceph_ino_compare(struct inode *inode, void *data) | |
495 | { | |
496 | struct ceph_vino *pvino = (struct ceph_vino *)data; | |
497 | struct ceph_inode_info *ci = ceph_inode(inode); | |
498 | return ci->i_vino.ino == pvino->ino && | |
499 | ci->i_vino.snap == pvino->snap; | |
500 | } | |
501 | ||
502 | static inline struct inode *ceph_find_inode(struct super_block *sb, | |
503 | struct ceph_vino vino) | |
504 | { | |
505 | ino_t t = ceph_vino_to_ino(vino); | |
506 | return ilookup5(sb, t, ceph_ino_compare, &vino); | |
507 | } | |
508 | ||
509 | ||
510 | /* | |
511 | * caps helpers | |
512 | */ | |
513 | static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) | |
514 | { | |
515 | return !RB_EMPTY_ROOT(&ci->i_caps); | |
516 | } | |
517 | ||
518 | extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); | |
519 | extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); | |
520 | extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, | |
521 | struct ceph_cap *cap); | |
522 | ||
523 | static inline int ceph_caps_issued(struct ceph_inode_info *ci) | |
524 | { | |
525 | int issued; | |
526 | spin_lock(&ci->vfs_inode.i_lock); | |
527 | issued = __ceph_caps_issued(ci, NULL); | |
528 | spin_unlock(&ci->vfs_inode.i_lock); | |
529 | return issued; | |
530 | } | |
531 | ||
532 | static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, | |
533 | int touch) | |
534 | { | |
535 | int r; | |
536 | spin_lock(&ci->vfs_inode.i_lock); | |
537 | r = __ceph_caps_issued_mask(ci, mask, touch); | |
538 | spin_unlock(&ci->vfs_inode.i_lock); | |
539 | return r; | |
540 | } | |
541 | ||
542 | static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) | |
543 | { | |
544 | return ci->i_dirty_caps | ci->i_flushing_caps; | |
545 | } | |
afcdaea3 | 546 | extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); |
de57606c SW |
547 | |
548 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); | |
549 | extern int __ceph_caps_used(struct ceph_inode_info *ci); | |
550 | ||
551 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); | |
552 | ||
553 | /* | |
554 | * wanted, by virtue of open file modes AND cap refs (buffered/cached data) | |
555 | */ | |
556 | static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) | |
557 | { | |
558 | int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); | |
559 | if (w & CEPH_CAP_FILE_BUFFER) | |
560 | w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ | |
561 | return w; | |
562 | } | |
563 | ||
564 | /* what the mds thinks we want */ | |
565 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); | |
566 | ||
567 | extern void ceph_caps_init(void); | |
568 | extern void ceph_caps_finalize(void); | |
85ccce43 | 569 | extern void ceph_adjust_min_caps(int delta); |
de57606c SW |
570 | extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need); |
571 | extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx); | |
572 | extern void ceph_reservation_status(struct ceph_client *client, | |
573 | int *total, int *avail, int *used, | |
85ccce43 | 574 | int *reserved, int *min); |
de57606c SW |
575 | |
576 | static inline struct ceph_client *ceph_inode_to_client(struct inode *inode) | |
577 | { | |
578 | return (struct ceph_client *)inode->i_sb->s_fs_info; | |
579 | } | |
580 | ||
581 | static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb) | |
582 | { | |
583 | return (struct ceph_client *)sb->s_fs_info; | |
584 | } | |
585 | ||
de57606c SW |
586 | |
587 | /* | |
588 | * we keep buffered readdir results attached to file->private_data | |
589 | */ | |
590 | struct ceph_file_info { | |
591 | int fmode; /* initialized on open */ | |
592 | ||
593 | /* readdir: position within the dir */ | |
594 | u32 frag; | |
595 | struct ceph_mds_request *last_readdir; | |
596 | int at_end; | |
597 | ||
598 | /* readdir: position within a frag */ | |
599 | unsigned offset; /* offset of last chunk, adjusted for . and .. */ | |
600 | u64 next_offset; /* offset of next chunk (last_name's + 1) */ | |
601 | char *last_name; /* last entry in previous chunk */ | |
602 | struct dentry *dentry; /* next dentry (for dcache readdir) */ | |
603 | unsigned long dir_release_count; | |
604 | ||
605 | /* used for -o dirstat read() on directory thing */ | |
606 | char *dir_info; | |
607 | int dir_info_len; | |
608 | }; | |
609 | ||
610 | ||
611 | ||
612 | /* | |
613 | * snapshots | |
614 | */ | |
615 | ||
616 | /* | |
617 | * A "snap context" is the set of existing snapshots when we | |
618 | * write data. It is used by the OSD to guide its COW behavior. | |
619 | * | |
620 | * The ceph_snap_context is refcounted, and attached to each dirty | |
621 | * page, indicating which context the dirty data belonged when it was | |
622 | * dirtied. | |
623 | */ | |
624 | struct ceph_snap_context { | |
625 | atomic_t nref; | |
626 | u64 seq; | |
627 | int num_snaps; | |
628 | u64 snaps[]; | |
629 | }; | |
630 | ||
631 | static inline struct ceph_snap_context * | |
632 | ceph_get_snap_context(struct ceph_snap_context *sc) | |
633 | { | |
634 | /* | |
635 | printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), | |
636 | atomic_read(&sc->nref)+1); | |
637 | */ | |
638 | if (sc) | |
639 | atomic_inc(&sc->nref); | |
640 | return sc; | |
641 | } | |
642 | ||
643 | static inline void ceph_put_snap_context(struct ceph_snap_context *sc) | |
644 | { | |
645 | if (!sc) | |
646 | return; | |
647 | /* | |
648 | printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), | |
649 | atomic_read(&sc->nref)-1); | |
650 | */ | |
651 | if (atomic_dec_and_test(&sc->nref)) { | |
652 | /*printk(" deleting snap_context %p\n", sc);*/ | |
653 | kfree(sc); | |
654 | } | |
655 | } | |
656 | ||
657 | /* | |
658 | * A "snap realm" describes a subset of the file hierarchy sharing | |
659 | * the same set of snapshots that apply to it. The realms themselves | |
660 | * are organized into a hierarchy, such that children inherit (some of) | |
661 | * the snapshots of their parents. | |
662 | * | |
663 | * All inodes within the realm that have capabilities are linked into a | |
664 | * per-realm list. | |
665 | */ | |
666 | struct ceph_snap_realm { | |
667 | u64 ino; | |
668 | atomic_t nref; | |
a105f00c SW |
669 | struct rb_node node; |
670 | ||
de57606c SW |
671 | u64 created, seq; |
672 | u64 parent_ino; | |
673 | u64 parent_since; /* snapid when our current parent became so */ | |
674 | ||
675 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ | |
676 | int num_prior_parent_snaps; /* had prior to parent_since */ | |
677 | u64 *snaps; /* snaps specific to this realm */ | |
678 | int num_snaps; | |
679 | ||
680 | struct ceph_snap_realm *parent; | |
681 | struct list_head children; /* list of child realms */ | |
682 | struct list_head child_item; | |
683 | ||
684 | struct list_head empty_item; /* if i have ref==0 */ | |
685 | ||
686 | /* the current set of snaps for this realm */ | |
687 | struct ceph_snap_context *cached_context; | |
688 | ||
689 | struct list_head inodes_with_caps; | |
690 | spinlock_t inodes_with_caps_lock; | |
691 | }; | |
692 | ||
693 | ||
694 | ||
695 | /* | |
696 | * calculate the number of pages a given length and offset map onto, | |
697 | * if we align the data. | |
698 | */ | |
699 | static inline int calc_pages_for(u64 off, u64 len) | |
700 | { | |
701 | return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) - | |
702 | (off >> PAGE_CACHE_SHIFT); | |
703 | } | |
704 | ||
705 | ||
706 | ||
707 | /* snap.c */ | |
708 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, | |
709 | u64 ino); | |
710 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, | |
711 | struct ceph_snap_realm *realm); | |
712 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, | |
713 | struct ceph_snap_realm *realm); | |
714 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, | |
715 | void *p, void *e, bool deletion); | |
716 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, | |
2600d2dd | 717 | struct ceph_mds_session *session, |
de57606c | 718 | struct ceph_msg *msg); |
fc837c8f | 719 | extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); |
de57606c SW |
720 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, |
721 | struct ceph_cap_snap *capsnap); | |
722 | extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); | |
723 | ||
724 | /* | |
725 | * a cap_snap is "pending" if it is still awaiting an in-progress | |
726 | * sync write (that may/may not still update size, mtime, etc.). | |
727 | */ | |
728 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) | |
729 | { | |
730 | return !list_empty(&ci->i_cap_snaps) && | |
731 | list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, | |
732 | ci_item)->writing; | |
733 | } | |
734 | ||
735 | ||
736 | /* super.c */ | |
737 | extern struct kmem_cache *ceph_inode_cachep; | |
738 | extern struct kmem_cache *ceph_cap_cachep; | |
739 | extern struct kmem_cache *ceph_dentry_cachep; | |
740 | extern struct kmem_cache *ceph_file_cachep; | |
741 | ||
742 | extern const char *ceph_msg_type_name(int type); | |
0743304d | 743 | extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid); |
de57606c SW |
744 | |
745 | #define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \ | |
746 | "%02x%02x%02x%02x%02x%02x" | |
747 | #define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \ | |
748 | (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7], \ | |
749 | (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11], \ | |
750 | (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15] | |
751 | ||
752 | /* inode.c */ | |
753 | extern const struct inode_operations ceph_file_iops; | |
754 | ||
755 | extern struct inode *ceph_alloc_inode(struct super_block *sb); | |
756 | extern void ceph_destroy_inode(struct inode *inode); | |
757 | ||
758 | extern struct inode *ceph_get_inode(struct super_block *sb, | |
759 | struct ceph_vino vino); | |
760 | extern struct inode *ceph_get_snapdir(struct inode *parent); | |
761 | extern int ceph_fill_file_size(struct inode *inode, int issued, | |
762 | u32 truncate_seq, u64 truncate_size, u64 size); | |
763 | extern void ceph_fill_file_time(struct inode *inode, int issued, | |
764 | u64 time_warp_seq, struct timespec *ctime, | |
765 | struct timespec *mtime, struct timespec *atime); | |
766 | extern int ceph_fill_trace(struct super_block *sb, | |
767 | struct ceph_mds_request *req, | |
768 | struct ceph_mds_session *session); | |
769 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, | |
770 | struct ceph_mds_session *session); | |
771 | ||
772 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); | |
773 | ||
774 | extern int ceph_inode_set_size(struct inode *inode, loff_t size); | |
de57606c | 775 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); |
3c6f6b79 SW |
776 | extern void ceph_queue_vmtruncate(struct inode *inode); |
777 | ||
778 | extern void ceph_queue_invalidate(struct inode *inode); | |
779 | extern void ceph_queue_writeback(struct inode *inode); | |
de57606c SW |
780 | |
781 | extern int ceph_do_getattr(struct inode *inode, int mask); | |
782 | extern int ceph_permission(struct inode *inode, int mask); | |
783 | extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); | |
784 | extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, | |
785 | struct kstat *stat); | |
786 | ||
787 | /* xattr.c */ | |
788 | extern int ceph_setxattr(struct dentry *, const char *, const void *, | |
789 | size_t, int); | |
790 | extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); | |
791 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); | |
792 | extern int ceph_removexattr(struct dentry *, const char *); | |
793 | extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); | |
794 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); | |
795 | ||
796 | /* caps.c */ | |
797 | extern const char *ceph_cap_string(int c); | |
798 | extern void ceph_handle_caps(struct ceph_mds_session *session, | |
799 | struct ceph_msg *msg); | |
800 | extern int ceph_add_cap(struct inode *inode, | |
801 | struct ceph_mds_session *session, u64 cap_id, | |
802 | int fmode, unsigned issued, unsigned wanted, | |
803 | unsigned cap, unsigned seq, u64 realmino, int flags, | |
804 | struct ceph_cap_reservation *caps_reservation); | |
7c1332b8 | 805 | extern void __ceph_remove_cap(struct ceph_cap *cap); |
de57606c SW |
806 | static inline void ceph_remove_cap(struct ceph_cap *cap) |
807 | { | |
808 | struct inode *inode = &cap->ci->vfs_inode; | |
809 | spin_lock(&inode->i_lock); | |
7c1332b8 | 810 | __ceph_remove_cap(cap); |
de57606c SW |
811 | spin_unlock(&inode->i_lock); |
812 | } | |
7c1332b8 | 813 | extern void ceph_put_cap(struct ceph_cap *cap); |
de57606c SW |
814 | |
815 | extern void ceph_queue_caps_release(struct inode *inode); | |
f1a3d572 | 816 | extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); |
de57606c SW |
817 | extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync); |
818 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, | |
819 | struct ceph_mds_session *session); | |
820 | extern int ceph_get_cap_mds(struct inode *inode); | |
821 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); | |
822 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); | |
823 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, | |
824 | struct ceph_snap_context *snapc); | |
825 | extern void __ceph_flush_snaps(struct ceph_inode_info *ci, | |
826 | struct ceph_mds_session **psession); | |
827 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, | |
828 | struct ceph_mds_session *session); | |
afcdaea3 SW |
829 | extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); |
830 | extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); | |
de57606c SW |
831 | |
832 | extern int ceph_encode_inode_release(void **p, struct inode *inode, | |
833 | int mds, int drop, int unless, int force); | |
834 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, | |
835 | int mds, int drop, int unless); | |
836 | ||
837 | extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, | |
838 | int *got, loff_t endoff); | |
839 | ||
840 | /* for counting open files by mode */ | |
841 | static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) | |
842 | { | |
843 | ci->i_nr_by_mode[mode]++; | |
844 | } | |
845 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); | |
846 | ||
847 | /* addr.c */ | |
848 | extern const struct address_space_operations ceph_aops; | |
849 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); | |
850 | ||
851 | /* file.c */ | |
852 | extern const struct file_operations ceph_file_fops; | |
853 | extern const struct address_space_operations ceph_aops; | |
854 | extern int ceph_open(struct inode *inode, struct file *file); | |
855 | extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry, | |
856 | struct nameidata *nd, int mode, | |
857 | int locked_dir); | |
858 | extern int ceph_release(struct inode *inode, struct file *filp); | |
859 | extern void ceph_release_page_vector(struct page **pages, int num_pages); | |
860 | ||
861 | /* dir.c */ | |
862 | extern const struct file_operations ceph_dir_fops; | |
863 | extern const struct inode_operations ceph_dir_iops; | |
864 | extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, | |
865 | ceph_snapdir_dentry_ops; | |
866 | ||
867 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); | |
868 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, | |
869 | struct dentry *dentry, int err); | |
870 | ||
871 | extern void ceph_dentry_lru_add(struct dentry *dn); | |
872 | extern void ceph_dentry_lru_touch(struct dentry *dn); | |
873 | extern void ceph_dentry_lru_del(struct dentry *dn); | |
874 | ||
875 | /* | |
876 | * our d_ops vary depending on whether the inode is live, | |
877 | * snapshotted (read-only), or a virtual ".snap" directory. | |
878 | */ | |
879 | int ceph_init_dentry(struct dentry *dentry); | |
880 | ||
881 | ||
882 | /* ioctl.c */ | |
883 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | |
884 | ||
885 | /* export.c */ | |
886 | extern const struct export_operations ceph_export_ops; | |
887 | ||
888 | /* debugfs.c */ | |
889 | extern int ceph_debugfs_init(void); | |
890 | extern void ceph_debugfs_cleanup(void); | |
891 | extern int ceph_debugfs_client_init(struct ceph_client *client); | |
892 | extern void ceph_debugfs_client_cleanup(struct ceph_client *client); | |
893 | ||
894 | static inline struct inode *get_dentry_parent_inode(struct dentry *dentry) | |
895 | { | |
896 | if (dentry && dentry->d_parent) | |
897 | return dentry->d_parent->d_inode; | |
898 | ||
899 | return NULL; | |
900 | } | |
901 | ||
902 | #endif /* _FS_CEPH_SUPER_H */ |