diff options
author | Sage Weil <sage@newdream.net> | 2009-10-06 14:31:07 -0400 |
---|---|---|
committer | Sage Weil <sage@newdream.net> | 2009-10-06 14:31:07 -0400 |
commit | de57606c23afded22202825b3db8a5d61859f198 (patch) | |
tree | 87ca160cf9655161dbd3c42b1bdc23c4b5584e1d /fs/ceph/super.h | |
parent | 0dee3c28af2fbe22ca62739a7f57da5435d35793 (diff) |
ceph: client types
We first define constants, types, and prototypes for the kernel client
proper.
A few subsystems are defined separately later: the MDS, OSD, and
monitor clients, and the messaging layer.
Signed-off-by: Sage Weil <sage@newdream.net>
Diffstat (limited to 'fs/ceph/super.h')
-rw-r--r-- | fs/ceph/super.h | 890 |
1 files changed, 890 insertions, 0 deletions
diff --git a/fs/ceph/super.h b/fs/ceph/super.h new file mode 100644 index 00000000000..cfd39ef4023 --- /dev/null +++ b/fs/ceph/super.h | |||
@@ -0,0 +1,890 @@ | |||
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) \ | ||
45 | (client)->mount_args.flags |= CEPH_OPT_##opt; | ||
46 | #define ceph_test_opt(client, opt) \ | ||
47 | (!!((client)->mount_args.flags & CEPH_OPT_##opt)) | ||
48 | |||
49 | |||
50 | #define CEPH_MAX_MON_MOUNT_ADDR 5 | ||
51 | |||
52 | struct ceph_mount_args { | ||
53 | int sb_flags; | ||
54 | int flags; | ||
55 | int mount_timeout; | ||
56 | int caps_wanted_delay_min, caps_wanted_delay_max; | ||
57 | struct ceph_fsid fsid; | ||
58 | struct ceph_entity_addr my_addr; | ||
59 | int wsize; | ||
60 | int rsize; /* max readahead */ | ||
61 | int max_readdir; /* max readdir size */ | ||
62 | int osd_timeout; | ||
63 | char *snapdir_name; /* default ".snap" */ | ||
64 | char *secret; | ||
65 | int cap_release_safety; | ||
66 | }; | ||
67 | |||
68 | /* | ||
69 | * defaults | ||
70 | */ | ||
71 | #define CEPH_MOUNT_TIMEOUT_DEFAULT 60 | ||
72 | #define CEPH_MOUNT_RSIZE_DEFAULT (128*1024) /* readahead */ | ||
73 | |||
74 | #define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024) | ||
75 | #define CEPH_MSG_MAX_DATA_LEN (16*1024*1024) | ||
76 | |||
77 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" | ||
78 | |||
79 | /* | ||
80 | * Delay telling the MDS we no longer want caps, in case we reopen | ||
81 | * the file. Delay a minimum amount of time, even if we send a cap | ||
82 | * message for some other reason. Otherwise, take the oppotunity to | ||
83 | * update the mds to avoid sending another message later. | ||
84 | */ | ||
85 | #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ | ||
86 | #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ | ||
87 | |||
88 | |||
89 | /* mount state */ | ||
90 | enum { | ||
91 | CEPH_MOUNT_MOUNTING, | ||
92 | CEPH_MOUNT_MOUNTED, | ||
93 | CEPH_MOUNT_UNMOUNTING, | ||
94 | CEPH_MOUNT_UNMOUNTED, | ||
95 | CEPH_MOUNT_SHUTDOWN, | ||
96 | }; | ||
97 | |||
98 | /* | ||
99 | * subtract jiffies | ||
100 | */ | ||
101 | static inline unsigned long time_sub(unsigned long a, unsigned long b) | ||
102 | { | ||
103 | BUG_ON(time_after(b, a)); | ||
104 | return (long)a - (long)b; | ||
105 | } | ||
106 | |||
107 | /* | ||
108 | * per-filesystem client state | ||
109 | * | ||
110 | * possibly shared by multiple mount points, if they are | ||
111 | * mounting the same ceph filesystem/cluster. | ||
112 | */ | ||
113 | struct ceph_client { | ||
114 | __s64 whoami; /* my client number */ | ||
115 | struct dentry *debugfs_monmap; | ||
116 | struct dentry *debugfs_mdsmap, *debugfs_osdmap; | ||
117 | struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps; | ||
118 | |||
119 | struct mutex mount_mutex; /* serialize mount attempts */ | ||
120 | struct ceph_mount_args mount_args; | ||
121 | struct ceph_fsid fsid; | ||
122 | |||
123 | struct super_block *sb; | ||
124 | |||
125 | unsigned long mount_state; | ||
126 | wait_queue_head_t mount_wq; | ||
127 | |||
128 | int mount_err; | ||
129 | void *signed_ticket; /* our keys to the kingdom */ | ||
130 | int signed_ticket_len; | ||
131 | |||
132 | struct ceph_messenger *msgr; /* messenger instance */ | ||
133 | struct ceph_mon_client monc; | ||
134 | struct ceph_mds_client mdsc; | ||
135 | struct ceph_osd_client osdc; | ||
136 | |||
137 | /* writeback */ | ||
138 | mempool_t *wb_pagevec_pool; | ||
139 | struct workqueue_struct *wb_wq; | ||
140 | struct workqueue_struct *pg_inv_wq; | ||
141 | struct workqueue_struct *trunc_wq; | ||
142 | |||
143 | struct backing_dev_info backing_dev_info; | ||
144 | }; | ||
145 | |||
146 | static inline struct ceph_client *ceph_client(struct super_block *sb) | ||
147 | { | ||
148 | return sb->s_fs_info; | ||
149 | } | ||
150 | |||
151 | |||
152 | /* | ||
153 | * File i/o capability. This tracks shared state with the metadata | ||
154 | * server that allows us to cache or writeback attributes or to read | ||
155 | * and write data. For any given inode, we should have one or more | ||
156 | * capabilities, one issued by each metadata server, and our | ||
157 | * cumulative access is the OR of all issued capabilities. | ||
158 | * | ||
159 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds | ||
160 | * session capability lists. | ||
161 | */ | ||
162 | struct ceph_cap { | ||
163 | struct ceph_inode_info *ci; | ||
164 | struct rb_node ci_node; /* per-ci cap tree */ | ||
165 | struct ceph_mds_session *session; | ||
166 | struct list_head session_caps; /* per-session caplist */ | ||
167 | int mds; | ||
168 | u64 cap_id; /* unique cap id (mds provided) */ | ||
169 | int issued; /* latest, from the mds */ | ||
170 | int implemented; /* implemented superset of issued (for revocation) */ | ||
171 | int mds_wanted; | ||
172 | u32 seq, issue_seq, mseq, gen; | ||
173 | unsigned long last_used; | ||
174 | struct list_head caps_item; | ||
175 | }; | ||
176 | |||
177 | #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ | ||
178 | #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ | ||
179 | #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ | ||
180 | |||
181 | /* | ||
182 | * Snapped cap state that is pending flush to mds. When a snapshot occurs, | ||
183 | * we first complete any in-process sync writes and writeback any dirty | ||
184 | * data before flushing the snapped state (tracked here) back to the MDS. | ||
185 | */ | ||
186 | struct ceph_cap_snap { | ||
187 | atomic_t nref; | ||
188 | struct ceph_inode_info *ci; | ||
189 | struct list_head ci_item, flushing_item; | ||
190 | |||
191 | u64 follows, flush_tid; | ||
192 | int issued, dirty; | ||
193 | struct ceph_snap_context *context; | ||
194 | |||
195 | mode_t mode; | ||
196 | uid_t uid; | ||
197 | gid_t gid; | ||
198 | |||
199 | void *xattr_blob; | ||
200 | int xattr_len; | ||
201 | u64 xattr_version; | ||
202 | |||
203 | u64 size; | ||
204 | struct timespec mtime, atime, ctime; | ||
205 | u64 time_warp_seq; | ||
206 | int writing; /* a sync write is still in progress */ | ||
207 | int dirty_pages; /* dirty pages awaiting writeback */ | ||
208 | }; | ||
209 | |||
210 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) | ||
211 | { | ||
212 | if (atomic_dec_and_test(&capsnap->nref)) | ||
213 | kfree(capsnap); | ||
214 | } | ||
215 | |||
216 | /* | ||
217 | * The frag tree describes how a directory is fragmented, potentially across | ||
218 | * multiple metadata servers. It is also used to indicate points where | ||
219 | * metadata authority is delegated, and whether/where metadata is replicated. | ||
220 | * | ||
221 | * A _leaf_ frag will be present in the i_fragtree IFF there is | ||
222 | * delegation info. That is, if mds >= 0 || ndist > 0. | ||
223 | */ | ||
224 | #define CEPH_MAX_DIRFRAG_REP 4 | ||
225 | |||
226 | struct ceph_inode_frag { | ||
227 | struct rb_node node; | ||
228 | |||
229 | /* fragtree state */ | ||
230 | u32 frag; | ||
231 | int split_by; /* i.e. 2^(split_by) children */ | ||
232 | |||
233 | /* delegation and replication info */ | ||
234 | int mds; /* -1 if same authority as parent */ | ||
235 | int ndist; /* >0 if replicated */ | ||
236 | int dist[CEPH_MAX_DIRFRAG_REP]; | ||
237 | }; | ||
238 | |||
239 | /* | ||
240 | * We cache inode xattrs as an encoded blob until they are first used, | ||
241 | * at which point we parse them into an rbtree. | ||
242 | */ | ||
243 | struct ceph_inode_xattr { | ||
244 | struct rb_node node; | ||
245 | |||
246 | const char *name; | ||
247 | int name_len; | ||
248 | const char *val; | ||
249 | int val_len; | ||
250 | int dirty; | ||
251 | |||
252 | int should_free_name; | ||
253 | int should_free_val; | ||
254 | }; | ||
255 | |||
256 | struct ceph_inode_xattrs_info { | ||
257 | /* | ||
258 | * (still encoded) xattr blob. we avoid the overhead of parsing | ||
259 | * this until someone actually calls getxattr, etc. | ||
260 | * | ||
261 | * blob->vec.iov_len == 4 implies there are no xattrs; blob == | ||
262 | * NULL means we don't know. | ||
263 | */ | ||
264 | struct ceph_buffer *blob, *prealloc_blob; | ||
265 | |||
266 | struct rb_root index; | ||
267 | bool dirty; | ||
268 | int count; | ||
269 | int names_size; | ||
270 | int vals_size; | ||
271 | u64 version, index_version; | ||
272 | }; | ||
273 | |||
274 | /* | ||
275 | * Ceph inode. | ||
276 | */ | ||
277 | #define CEPH_I_COMPLETE 1 /* we have complete directory cached */ | ||
278 | #define CEPH_I_NODELAY 4 /* do not delay cap release */ | ||
279 | #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ | ||
280 | |||
281 | struct ceph_inode_info { | ||
282 | struct ceph_vino i_vino; /* ceph ino + snap */ | ||
283 | |||
284 | u64 i_version; | ||
285 | u32 i_time_warp_seq; | ||
286 | |||
287 | unsigned i_ceph_flags; | ||
288 | unsigned long i_release_count; | ||
289 | |||
290 | struct ceph_file_layout i_layout; | ||
291 | char *i_symlink; | ||
292 | |||
293 | /* for dirs */ | ||
294 | struct timespec i_rctime; | ||
295 | u64 i_rbytes, i_rfiles, i_rsubdirs; | ||
296 | u64 i_files, i_subdirs; | ||
297 | u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */ | ||
298 | |||
299 | struct rb_root i_fragtree; | ||
300 | struct mutex i_fragtree_mutex; | ||
301 | |||
302 | struct ceph_inode_xattrs_info i_xattrs; | ||
303 | |||
304 | /* capabilities. protected _both_ by i_lock and cap->session's | ||
305 | * s_mutex. */ | ||
306 | struct rb_root i_caps; /* cap list */ | ||
307 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ | ||
308 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ | ||
309 | struct list_head i_dirty_item, i_flushing_item; | ||
310 | u64 i_cap_flush_seq; | ||
311 | /* we need to track cap writeback on a per-cap-bit basis, to allow | ||
312 | * overlapping, pipelined cap flushes to the mds. we can probably | ||
313 | * reduce the tid to 8 bits if we're concerned about inode size. */ | ||
314 | u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; | ||
315 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ | ||
316 | unsigned long i_hold_caps_min; /* jiffies */ | ||
317 | unsigned long i_hold_caps_max; /* jiffies */ | ||
318 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ | ||
319 | int i_cap_exporting_mds; /* to handle cap migration between */ | ||
320 | unsigned i_cap_exporting_mseq; /* mds's. */ | ||
321 | unsigned i_cap_exporting_issued; | ||
322 | struct ceph_cap_reservation i_cap_migration_resv; | ||
323 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ | ||
324 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */ | ||
325 | unsigned i_snap_caps; /* cap bits for snapped files */ | ||
326 | |||
327 | int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ | ||
328 | |||
329 | u32 i_truncate_seq; /* last truncate to smaller size */ | ||
330 | u64 i_truncate_size; /* and the size we last truncated down to */ | ||
331 | int i_truncate_pending; /* still need to call vmtruncate */ | ||
332 | |||
333 | u64 i_max_size; /* max file size authorized by mds */ | ||
334 | u64 i_reported_size; /* (max_)size reported to or requested of mds */ | ||
335 | u64 i_wanted_max_size; /* offset we'd like to write too */ | ||
336 | u64 i_requested_max_size; /* max_size we've requested */ | ||
337 | |||
338 | /* held references to caps */ | ||
339 | int i_pin_ref; | ||
340 | int i_rd_ref, i_rdcache_ref, i_wr_ref; | ||
341 | int i_wrbuffer_ref, i_wrbuffer_ref_head; | ||
342 | u32 i_shared_gen; /* increment each time we get FILE_SHARED */ | ||
343 | u32 i_rdcache_gen; /* we increment this each time we get | ||
344 | FILE_CACHE. If it's non-zero, we | ||
345 | _may_ have cached pages. */ | ||
346 | u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ | ||
347 | |||
348 | struct list_head i_unsafe_writes; /* uncommitted sync writes */ | ||
349 | struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ | ||
350 | spinlock_t i_unsafe_lock; | ||
351 | |||
352 | struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ | ||
353 | int i_snap_realm_counter; /* snap realm (if caps) */ | ||
354 | struct list_head i_snap_realm_item; | ||
355 | struct list_head i_snap_flush_item; | ||
356 | |||
357 | struct work_struct i_wb_work; /* writeback work */ | ||
358 | struct work_struct i_pg_inv_work; /* page invalidation work */ | ||
359 | |||
360 | struct work_struct i_vmtruncate_work; | ||
361 | |||
362 | struct inode vfs_inode; /* at end */ | ||
363 | }; | ||
364 | |||
365 | static inline struct ceph_inode_info *ceph_inode(struct inode *inode) | ||
366 | { | ||
367 | return list_entry(inode, struct ceph_inode_info, vfs_inode); | ||
368 | } | ||
369 | |||
370 | static inline void ceph_i_clear(struct inode *inode, unsigned mask) | ||
371 | { | ||
372 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
373 | |||
374 | spin_lock(&inode->i_lock); | ||
375 | ci->i_ceph_flags &= ~mask; | ||
376 | spin_unlock(&inode->i_lock); | ||
377 | } | ||
378 | |||
379 | static inline void ceph_i_set(struct inode *inode, unsigned mask) | ||
380 | { | ||
381 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
382 | |||
383 | spin_lock(&inode->i_lock); | ||
384 | ci->i_ceph_flags |= mask; | ||
385 | spin_unlock(&inode->i_lock); | ||
386 | } | ||
387 | |||
388 | static inline bool ceph_i_test(struct inode *inode, unsigned mask) | ||
389 | { | ||
390 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
391 | bool r; | ||
392 | |||
393 | smp_mb(); | ||
394 | r = (ci->i_ceph_flags & mask) == mask; | ||
395 | return r; | ||
396 | } | ||
397 | |||
398 | |||
399 | /* find a specific frag @f */ | ||
400 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, | ||
401 | u32 f); | ||
402 | |||
403 | /* | ||
404 | * choose fragment for value @v. copy frag content to pfrag, if leaf | ||
405 | * exists | ||
406 | */ | ||
407 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, | ||
408 | struct ceph_inode_frag *pfrag, | ||
409 | int *found); | ||
410 | |||
411 | /* | ||
412 | * Ceph dentry state | ||
413 | */ | ||
414 | struct ceph_dentry_info { | ||
415 | struct ceph_mds_session *lease_session; | ||
416 | u32 lease_gen, lease_shared_gen; | ||
417 | u32 lease_seq; | ||
418 | unsigned long lease_renew_after, lease_renew_from; | ||
419 | struct list_head lru; | ||
420 | struct dentry *dentry; | ||
421 | u64 time; | ||
422 | u64 offset; | ||
423 | }; | ||
424 | |||
425 | static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) | ||
426 | { | ||
427 | return (struct ceph_dentry_info *)dentry->d_fsdata; | ||
428 | } | ||
429 | |||
430 | static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) | ||
431 | { | ||
432 | return ((loff_t)frag << 32) | (loff_t)off; | ||
433 | } | ||
434 | |||
435 | /* | ||
436 | * ino_t is <64 bits on many architectures, blech. | ||
437 | * | ||
438 | * don't include snap in ino hash, at least for now. | ||
439 | */ | ||
440 | static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) | ||
441 | { | ||
442 | ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */ | ||
443 | #if BITS_PER_LONG == 32 | ||
444 | ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8; | ||
445 | if (!ino) | ||
446 | ino = 1; | ||
447 | #endif | ||
448 | return ino; | ||
449 | } | ||
450 | |||
451 | static inline int ceph_set_ino_cb(struct inode *inode, void *data) | ||
452 | { | ||
453 | ceph_inode(inode)->i_vino = *(struct ceph_vino *)data; | ||
454 | inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data); | ||
455 | return 0; | ||
456 | } | ||
457 | |||
458 | static inline struct ceph_vino ceph_vino(struct inode *inode) | ||
459 | { | ||
460 | return ceph_inode(inode)->i_vino; | ||
461 | } | ||
462 | |||
463 | /* for printf-style formatting */ | ||
464 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap | ||
465 | |||
466 | static inline u64 ceph_ino(struct inode *inode) | ||
467 | { | ||
468 | return ceph_inode(inode)->i_vino.ino; | ||
469 | } | ||
470 | static inline u64 ceph_snap(struct inode *inode) | ||
471 | { | ||
472 | return ceph_inode(inode)->i_vino.snap; | ||
473 | } | ||
474 | |||
475 | static inline int ceph_ino_compare(struct inode *inode, void *data) | ||
476 | { | ||
477 | struct ceph_vino *pvino = (struct ceph_vino *)data; | ||
478 | struct ceph_inode_info *ci = ceph_inode(inode); | ||
479 | return ci->i_vino.ino == pvino->ino && | ||
480 | ci->i_vino.snap == pvino->snap; | ||
481 | } | ||
482 | |||
483 | static inline struct inode *ceph_find_inode(struct super_block *sb, | ||
484 | struct ceph_vino vino) | ||
485 | { | ||
486 | ino_t t = ceph_vino_to_ino(vino); | ||
487 | return ilookup5(sb, t, ceph_ino_compare, &vino); | ||
488 | } | ||
489 | |||
490 | |||
491 | /* | ||
492 | * caps helpers | ||
493 | */ | ||
494 | static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) | ||
495 | { | ||
496 | return !RB_EMPTY_ROOT(&ci->i_caps); | ||
497 | } | ||
498 | |||
499 | extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); | ||
500 | extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); | ||
501 | extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, | ||
502 | struct ceph_cap *cap); | ||
503 | |||
504 | static inline int ceph_caps_issued(struct ceph_inode_info *ci) | ||
505 | { | ||
506 | int issued; | ||
507 | spin_lock(&ci->vfs_inode.i_lock); | ||
508 | issued = __ceph_caps_issued(ci, NULL); | ||
509 | spin_unlock(&ci->vfs_inode.i_lock); | ||
510 | return issued; | ||
511 | } | ||
512 | |||
513 | static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, | ||
514 | int touch) | ||
515 | { | ||
516 | int r; | ||
517 | spin_lock(&ci->vfs_inode.i_lock); | ||
518 | r = __ceph_caps_issued_mask(ci, mask, touch); | ||
519 | spin_unlock(&ci->vfs_inode.i_lock); | ||
520 | return r; | ||
521 | } | ||
522 | |||
523 | static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) | ||
524 | { | ||
525 | return ci->i_dirty_caps | ci->i_flushing_caps; | ||
526 | } | ||
527 | extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); | ||
528 | |||
529 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); | ||
530 | extern int __ceph_caps_used(struct ceph_inode_info *ci); | ||
531 | |||
532 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); | ||
533 | |||
534 | /* | ||
535 | * wanted, by virtue of open file modes AND cap refs (buffered/cached data) | ||
536 | */ | ||
537 | static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) | ||
538 | { | ||
539 | int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); | ||
540 | if (w & CEPH_CAP_FILE_BUFFER) | ||
541 | w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ | ||
542 | return w; | ||
543 | } | ||
544 | |||
545 | /* what the mds thinks we want */ | ||
546 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); | ||
547 | |||
548 | extern void ceph_caps_init(void); | ||
549 | extern void ceph_caps_finalize(void); | ||
550 | extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need); | ||
551 | extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx); | ||
552 | extern void ceph_reservation_status(struct ceph_client *client, | ||
553 | int *total, int *avail, int *used, | ||
554 | int *reserved); | ||
555 | |||
556 | static inline struct ceph_client *ceph_inode_to_client(struct inode *inode) | ||
557 | { | ||
558 | return (struct ceph_client *)inode->i_sb->s_fs_info; | ||
559 | } | ||
560 | |||
561 | static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb) | ||
562 | { | ||
563 | return (struct ceph_client *)sb->s_fs_info; | ||
564 | } | ||
565 | |||
566 | static inline int ceph_queue_writeback(struct inode *inode) | ||
567 | { | ||
568 | return queue_work(ceph_inode_to_client(inode)->wb_wq, | ||
569 | &ceph_inode(inode)->i_wb_work); | ||
570 | } | ||
571 | |||
572 | static inline int ceph_queue_page_invalidation(struct inode *inode) | ||
573 | { | ||
574 | return queue_work(ceph_inode_to_client(inode)->pg_inv_wq, | ||
575 | &ceph_inode(inode)->i_pg_inv_work); | ||
576 | } | ||
577 | |||
578 | |||
579 | /* | ||
580 | * we keep buffered readdir results attached to file->private_data | ||
581 | */ | ||
582 | struct ceph_file_info { | ||
583 | int fmode; /* initialized on open */ | ||
584 | |||
585 | /* readdir: position within the dir */ | ||
586 | u32 frag; | ||
587 | struct ceph_mds_request *last_readdir; | ||
588 | int at_end; | ||
589 | |||
590 | /* readdir: position within a frag */ | ||
591 | unsigned offset; /* offset of last chunk, adjusted for . and .. */ | ||
592 | u64 next_offset; /* offset of next chunk (last_name's + 1) */ | ||
593 | char *last_name; /* last entry in previous chunk */ | ||
594 | struct dentry *dentry; /* next dentry (for dcache readdir) */ | ||
595 | unsigned long dir_release_count; | ||
596 | |||
597 | /* used for -o dirstat read() on directory thing */ | ||
598 | char *dir_info; | ||
599 | int dir_info_len; | ||
600 | }; | ||
601 | |||
602 | |||
603 | |||
604 | /* | ||
605 | * snapshots | ||
606 | */ | ||
607 | |||
608 | /* | ||
609 | * A "snap context" is the set of existing snapshots when we | ||
610 | * write data. It is used by the OSD to guide its COW behavior. | ||
611 | * | ||
612 | * The ceph_snap_context is refcounted, and attached to each dirty | ||
613 | * page, indicating which context the dirty data belonged when it was | ||
614 | * dirtied. | ||
615 | */ | ||
616 | struct ceph_snap_context { | ||
617 | atomic_t nref; | ||
618 | u64 seq; | ||
619 | int num_snaps; | ||
620 | u64 snaps[]; | ||
621 | }; | ||
622 | |||
623 | static inline struct ceph_snap_context * | ||
624 | ceph_get_snap_context(struct ceph_snap_context *sc) | ||
625 | { | ||
626 | /* | ||
627 | printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), | ||
628 | atomic_read(&sc->nref)+1); | ||
629 | */ | ||
630 | if (sc) | ||
631 | atomic_inc(&sc->nref); | ||
632 | return sc; | ||
633 | } | ||
634 | |||
635 | static inline void ceph_put_snap_context(struct ceph_snap_context *sc) | ||
636 | { | ||
637 | if (!sc) | ||
638 | return; | ||
639 | /* | ||
640 | printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), | ||
641 | atomic_read(&sc->nref)-1); | ||
642 | */ | ||
643 | if (atomic_dec_and_test(&sc->nref)) { | ||
644 | /*printk(" deleting snap_context %p\n", sc);*/ | ||
645 | kfree(sc); | ||
646 | } | ||
647 | } | ||
648 | |||
649 | /* | ||
650 | * A "snap realm" describes a subset of the file hierarchy sharing | ||
651 | * the same set of snapshots that apply to it. The realms themselves | ||
652 | * are organized into a hierarchy, such that children inherit (some of) | ||
653 | * the snapshots of their parents. | ||
654 | * | ||
655 | * All inodes within the realm that have capabilities are linked into a | ||
656 | * per-realm list. | ||
657 | */ | ||
658 | struct ceph_snap_realm { | ||
659 | u64 ino; | ||
660 | atomic_t nref; | ||
661 | u64 created, seq; | ||
662 | u64 parent_ino; | ||
663 | u64 parent_since; /* snapid when our current parent became so */ | ||
664 | |||
665 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ | ||
666 | int num_prior_parent_snaps; /* had prior to parent_since */ | ||
667 | u64 *snaps; /* snaps specific to this realm */ | ||
668 | int num_snaps; | ||
669 | |||
670 | struct ceph_snap_realm *parent; | ||
671 | struct list_head children; /* list of child realms */ | ||
672 | struct list_head child_item; | ||
673 | |||
674 | struct list_head empty_item; /* if i have ref==0 */ | ||
675 | |||
676 | /* the current set of snaps for this realm */ | ||
677 | struct ceph_snap_context *cached_context; | ||
678 | |||
679 | struct list_head inodes_with_caps; | ||
680 | spinlock_t inodes_with_caps_lock; | ||
681 | }; | ||
682 | |||
683 | |||
684 | |||
685 | /* | ||
686 | * calculate the number of pages a given length and offset map onto, | ||
687 | * if we align the data. | ||
688 | */ | ||
689 | static inline int calc_pages_for(u64 off, u64 len) | ||
690 | { | ||
691 | return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) - | ||
692 | (off >> PAGE_CACHE_SHIFT); | ||
693 | } | ||
694 | |||
695 | |||
696 | |||
697 | /* snap.c */ | ||
698 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, | ||
699 | u64 ino); | ||
700 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, | ||
701 | struct ceph_snap_realm *realm); | ||
702 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, | ||
703 | struct ceph_snap_realm *realm); | ||
704 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, | ||
705 | void *p, void *e, bool deletion); | ||
706 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, | ||
707 | struct ceph_msg *msg); | ||
708 | extern void ceph_queue_cap_snap(struct ceph_inode_info *ci, | ||
709 | struct ceph_snap_context *snapc); | ||
710 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, | ||
711 | struct ceph_cap_snap *capsnap); | ||
712 | extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); | ||
713 | |||
714 | /* | ||
715 | * a cap_snap is "pending" if it is still awaiting an in-progress | ||
716 | * sync write (that may/may not still update size, mtime, etc.). | ||
717 | */ | ||
718 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) | ||
719 | { | ||
720 | return !list_empty(&ci->i_cap_snaps) && | ||
721 | list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, | ||
722 | ci_item)->writing; | ||
723 | } | ||
724 | |||
725 | |||
726 | /* super.c */ | ||
727 | extern struct kmem_cache *ceph_inode_cachep; | ||
728 | extern struct kmem_cache *ceph_cap_cachep; | ||
729 | extern struct kmem_cache *ceph_dentry_cachep; | ||
730 | extern struct kmem_cache *ceph_file_cachep; | ||
731 | |||
732 | extern const char *ceph_msg_type_name(int type); | ||
733 | |||
734 | #define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \ | ||
735 | "%02x%02x%02x%02x%02x%02x" | ||
736 | #define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \ | ||
737 | (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7], \ | ||
738 | (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11], \ | ||
739 | (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15] | ||
740 | |||
741 | /* inode.c */ | ||
742 | extern const struct inode_operations ceph_file_iops; | ||
743 | |||
744 | extern struct inode *ceph_alloc_inode(struct super_block *sb); | ||
745 | extern void ceph_destroy_inode(struct inode *inode); | ||
746 | |||
747 | extern struct inode *ceph_get_inode(struct super_block *sb, | ||
748 | struct ceph_vino vino); | ||
749 | extern struct inode *ceph_get_snapdir(struct inode *parent); | ||
750 | extern int ceph_fill_file_size(struct inode *inode, int issued, | ||
751 | u32 truncate_seq, u64 truncate_size, u64 size); | ||
752 | extern void ceph_fill_file_time(struct inode *inode, int issued, | ||
753 | u64 time_warp_seq, struct timespec *ctime, | ||
754 | struct timespec *mtime, struct timespec *atime); | ||
755 | extern int ceph_fill_trace(struct super_block *sb, | ||
756 | struct ceph_mds_request *req, | ||
757 | struct ceph_mds_session *session); | ||
758 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, | ||
759 | struct ceph_mds_session *session); | ||
760 | |||
761 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); | ||
762 | |||
763 | extern int ceph_inode_set_size(struct inode *inode, loff_t size); | ||
764 | extern void ceph_inode_writeback(struct work_struct *work); | ||
765 | extern void ceph_vmtruncate_work(struct work_struct *work); | ||
766 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); | ||
767 | extern void __ceph_queue_vmtruncate(struct inode *inode); | ||
768 | |||
769 | extern int ceph_do_getattr(struct inode *inode, int mask); | ||
770 | extern int ceph_permission(struct inode *inode, int mask); | ||
771 | extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); | ||
772 | extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, | ||
773 | struct kstat *stat); | ||
774 | |||
775 | /* xattr.c */ | ||
776 | extern int ceph_setxattr(struct dentry *, const char *, const void *, | ||
777 | size_t, int); | ||
778 | extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); | ||
779 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); | ||
780 | extern int ceph_removexattr(struct dentry *, const char *); | ||
781 | extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); | ||
782 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); | ||
783 | |||
784 | /* caps.c */ | ||
785 | extern const char *ceph_cap_string(int c); | ||
786 | extern void ceph_handle_caps(struct ceph_mds_session *session, | ||
787 | struct ceph_msg *msg); | ||
788 | extern int ceph_add_cap(struct inode *inode, | ||
789 | struct ceph_mds_session *session, u64 cap_id, | ||
790 | int fmode, unsigned issued, unsigned wanted, | ||
791 | unsigned cap, unsigned seq, u64 realmino, int flags, | ||
792 | struct ceph_cap_reservation *caps_reservation); | ||
793 | extern void __ceph_remove_cap(struct ceph_cap *cap, | ||
794 | struct ceph_cap_reservation *ctx); | ||
795 | static inline void ceph_remove_cap(struct ceph_cap *cap) | ||
796 | { | ||
797 | struct inode *inode = &cap->ci->vfs_inode; | ||
798 | spin_lock(&inode->i_lock); | ||
799 | __ceph_remove_cap(cap, NULL); | ||
800 | spin_unlock(&inode->i_lock); | ||
801 | } | ||
802 | |||
803 | extern void ceph_queue_caps_release(struct inode *inode); | ||
804 | extern int ceph_write_inode(struct inode *inode, int unused); | ||
805 | extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync); | ||
806 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, | ||
807 | struct ceph_mds_session *session); | ||
808 | extern int ceph_get_cap_mds(struct inode *inode); | ||
809 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); | ||
810 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); | ||
811 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, | ||
812 | struct ceph_snap_context *snapc); | ||
813 | extern void __ceph_flush_snaps(struct ceph_inode_info *ci, | ||
814 | struct ceph_mds_session **psession); | ||
815 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, | ||
816 | struct ceph_mds_session *session); | ||
817 | extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc, | ||
818 | int flushdirty); | ||
819 | |||
820 | extern int ceph_encode_inode_release(void **p, struct inode *inode, | ||
821 | int mds, int drop, int unless, int force); | ||
822 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, | ||
823 | int mds, int drop, int unless); | ||
824 | |||
825 | extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, | ||
826 | int *got, loff_t endoff); | ||
827 | |||
828 | /* for counting open files by mode */ | ||
829 | static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) | ||
830 | { | ||
831 | ci->i_nr_by_mode[mode]++; | ||
832 | } | ||
833 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); | ||
834 | |||
835 | /* addr.c */ | ||
836 | extern const struct address_space_operations ceph_aops; | ||
837 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); | ||
838 | |||
839 | /* file.c */ | ||
840 | extern const struct file_operations ceph_file_fops; | ||
841 | extern const struct address_space_operations ceph_aops; | ||
842 | extern int ceph_open(struct inode *inode, struct file *file); | ||
843 | extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry, | ||
844 | struct nameidata *nd, int mode, | ||
845 | int locked_dir); | ||
846 | extern int ceph_release(struct inode *inode, struct file *filp); | ||
847 | extern void ceph_release_page_vector(struct page **pages, int num_pages); | ||
848 | |||
849 | /* dir.c */ | ||
850 | extern const struct file_operations ceph_dir_fops; | ||
851 | extern const struct inode_operations ceph_dir_iops; | ||
852 | extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, | ||
853 | ceph_snapdir_dentry_ops; | ||
854 | |||
855 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); | ||
856 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, | ||
857 | struct dentry *dentry, int err); | ||
858 | |||
859 | extern void ceph_dentry_lru_add(struct dentry *dn); | ||
860 | extern void ceph_dentry_lru_touch(struct dentry *dn); | ||
861 | extern void ceph_dentry_lru_del(struct dentry *dn); | ||
862 | |||
863 | /* | ||
864 | * our d_ops vary depending on whether the inode is live, | ||
865 | * snapshotted (read-only), or a virtual ".snap" directory. | ||
866 | */ | ||
867 | int ceph_init_dentry(struct dentry *dentry); | ||
868 | |||
869 | |||
870 | /* ioctl.c */ | ||
871 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | ||
872 | |||
873 | /* export.c */ | ||
874 | extern const struct export_operations ceph_export_ops; | ||
875 | |||
876 | /* debugfs.c */ | ||
877 | extern int ceph_debugfs_init(void); | ||
878 | extern void ceph_debugfs_cleanup(void); | ||
879 | extern int ceph_debugfs_client_init(struct ceph_client *client); | ||
880 | extern void ceph_debugfs_client_cleanup(struct ceph_client *client); | ||
881 | |||
882 | static inline struct inode *get_dentry_parent_inode(struct dentry *dentry) | ||
883 | { | ||
884 | if (dentry && dentry->d_parent) | ||
885 | return dentry->d_parent->d_inode; | ||
886 | |||
887 | return NULL; | ||
888 | } | ||
889 | |||
890 | #endif /* _FS_CEPH_SUPER_H */ | ||