Commit | Line | Data |
---|---|---|
de57606c SW |
1 | #ifndef _FS_CEPH_SUPER_H |
2 | #define _FS_CEPH_SUPER_H | |
3 | ||
3d14c5d2 | 4 | #include <linux/ceph/ceph_debug.h> |
de57606c SW |
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> | |
f1a3d572 | 14 | #include <linux/writeback.h> |
5a0e3ad6 | 15 | #include <linux/slab.h> |
de57606c | 16 | |
3d14c5d2 | 17 | #include <linux/ceph/libceph.h> |
de57606c SW |
18 | |
19 | /* f_type in struct statfs */ | |
20 | #define CEPH_SUPER_MAGIC 0x00c36400 | |
21 | ||
22 | /* large granularity for statfs utilization stats to facilitate | |
23 | * large volume sizes on 32-bit machines. */ | |
24 | #define CEPH_BLOCK_SHIFT 20 /* 1 MB */ | |
25 | #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) | |
26 | ||
3d14c5d2 YS |
27 | #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ |
28 | #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ | |
29 | #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ | |
ad1fee96 | 30 | #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ |
6a259382 | 31 | |
3d14c5d2 | 32 | #define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES) |
de57606c | 33 | |
3d14c5d2 YS |
34 | #define ceph_set_mount_opt(fsc, opt) \ |
35 | (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt; | |
36 | #define ceph_test_mount_opt(fsc, opt) \ | |
37 | (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) | |
de57606c | 38 | |
83817e35 SW |
39 | #define CEPH_RSIZE_DEFAULT 0 /* max read size */ |
40 | #define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */ | |
3d14c5d2 YS |
41 | #define CEPH_MAX_READDIR_DEFAULT 1024 |
42 | #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) | |
43 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" | |
de57606c | 44 | |
3d14c5d2 | 45 | struct ceph_mount_options { |
6e19a16e | 46 | int flags; |
3d14c5d2 YS |
47 | int sb_flags; |
48 | ||
83817e35 SW |
49 | int wsize; /* max write size */ |
50 | int rsize; /* max read size */ | |
51 | int rasize; /* max readahead */ | |
6e19a16e SW |
52 | int congestion_kb; /* max writeback in flight */ |
53 | int caps_wanted_delay_min, caps_wanted_delay_max; | |
54 | int cap_release_safety; | |
23804d91 SW |
55 | int max_readdir; /* max readdir result (entires) */ |
56 | int max_readdir_bytes; /* max readdir result (bytes) */ | |
de57606c | 57 | |
3d14c5d2 YS |
58 | /* |
59 | * everything above this point can be memcmp'd; everything below | |
60 | * is handled in compare_mount_options() | |
61 | */ | |
de57606c | 62 | |
3d14c5d2 YS |
63 | char *snapdir_name; /* default ".snap" */ |
64 | }; | |
de57606c | 65 | |
3d14c5d2 | 66 | struct ceph_fs_client { |
de57606c SW |
67 | struct super_block *sb; |
68 | ||
3d14c5d2 YS |
69 | struct ceph_mount_options *mount_options; |
70 | struct ceph_client *client; | |
de57606c | 71 | |
3d14c5d2 | 72 | unsigned long mount_state; |
85ccce43 SW |
73 | int min_caps; /* min caps i added */ |
74 | ||
3d14c5d2 | 75 | struct ceph_mds_client *mdsc; |
de57606c SW |
76 | |
77 | /* writeback */ | |
78 | mempool_t *wb_pagevec_pool; | |
79 | struct workqueue_struct *wb_wq; | |
80 | struct workqueue_struct *pg_inv_wq; | |
81 | struct workqueue_struct *trunc_wq; | |
2baba250 | 82 | atomic_long_t writeback_count; |
de57606c SW |
83 | |
84 | struct backing_dev_info backing_dev_info; | |
0743304d SW |
85 | |
86 | #ifdef CONFIG_DEBUG_FS | |
3d14c5d2 | 87 | struct dentry *debugfs_dentry_lru, *debugfs_caps; |
2baba250 | 88 | struct dentry *debugfs_congestion_kb; |
06edf046 | 89 | struct dentry *debugfs_bdi; |
3d14c5d2 | 90 | struct dentry *debugfs_mdsc, *debugfs_mdsmap; |
0743304d | 91 | #endif |
de57606c SW |
92 | }; |
93 | ||
3d14c5d2 | 94 | |
de57606c SW |
95 | /* |
96 | * File i/o capability. This tracks shared state with the metadata | |
97 | * server that allows us to cache or writeback attributes or to read | |
98 | * and write data. For any given inode, we should have one or more | |
99 | * capabilities, one issued by each metadata server, and our | |
100 | * cumulative access is the OR of all issued capabilities. | |
101 | * | |
102 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds | |
103 | * session capability lists. | |
104 | */ | |
105 | struct ceph_cap { | |
106 | struct ceph_inode_info *ci; | |
107 | struct rb_node ci_node; /* per-ci cap tree */ | |
108 | struct ceph_mds_session *session; | |
109 | struct list_head session_caps; /* per-session caplist */ | |
110 | int mds; | |
111 | u64 cap_id; /* unique cap id (mds provided) */ | |
112 | int issued; /* latest, from the mds */ | |
113 | int implemented; /* implemented superset of issued (for revocation) */ | |
114 | int mds_wanted; | |
685f9a5d SW |
115 | u32 seq, issue_seq, mseq; |
116 | u32 cap_gen; /* active/stale cycle */ | |
de57606c SW |
117 | unsigned long last_used; |
118 | struct list_head caps_item; | |
119 | }; | |
120 | ||
121 | #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ | |
122 | #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ | |
123 | #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ | |
124 | ||
125 | /* | |
126 | * Snapped cap state that is pending flush to mds. When a snapshot occurs, | |
127 | * we first complete any in-process sync writes and writeback any dirty | |
128 | * data before flushing the snapped state (tracked here) back to the MDS. | |
129 | */ | |
130 | struct ceph_cap_snap { | |
131 | atomic_t nref; | |
132 | struct ceph_inode_info *ci; | |
133 | struct list_head ci_item, flushing_item; | |
134 | ||
135 | u64 follows, flush_tid; | |
136 | int issued, dirty; | |
137 | struct ceph_snap_context *context; | |
138 | ||
139 | mode_t mode; | |
140 | uid_t uid; | |
141 | gid_t gid; | |
142 | ||
4a625be4 | 143 | struct ceph_buffer *xattr_blob; |
de57606c SW |
144 | u64 xattr_version; |
145 | ||
146 | u64 size; | |
147 | struct timespec mtime, atime, ctime; | |
148 | u64 time_warp_seq; | |
149 | int writing; /* a sync write is still in progress */ | |
150 | int dirty_pages; /* dirty pages awaiting writeback */ | |
151 | }; | |
152 | ||
153 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) | |
154 | { | |
4a625be4 SW |
155 | if (atomic_dec_and_test(&capsnap->nref)) { |
156 | if (capsnap->xattr_blob) | |
157 | ceph_buffer_put(capsnap->xattr_blob); | |
de57606c | 158 | kfree(capsnap); |
4a625be4 | 159 | } |
de57606c SW |
160 | } |
161 | ||
162 | /* | |
163 | * The frag tree describes how a directory is fragmented, potentially across | |
164 | * multiple metadata servers. It is also used to indicate points where | |
165 | * metadata authority is delegated, and whether/where metadata is replicated. | |
166 | * | |
167 | * A _leaf_ frag will be present in the i_fragtree IFF there is | |
168 | * delegation info. That is, if mds >= 0 || ndist > 0. | |
169 | */ | |
170 | #define CEPH_MAX_DIRFRAG_REP 4 | |
171 | ||
172 | struct ceph_inode_frag { | |
173 | struct rb_node node; | |
174 | ||
175 | /* fragtree state */ | |
176 | u32 frag; | |
177 | int split_by; /* i.e. 2^(split_by) children */ | |
178 | ||
179 | /* delegation and replication info */ | |
180 | int mds; /* -1 if same authority as parent */ | |
181 | int ndist; /* >0 if replicated */ | |
182 | int dist[CEPH_MAX_DIRFRAG_REP]; | |
183 | }; | |
184 | ||
185 | /* | |
186 | * We cache inode xattrs as an encoded blob until they are first used, | |
187 | * at which point we parse them into an rbtree. | |
188 | */ | |
189 | struct ceph_inode_xattr { | |
190 | struct rb_node node; | |
191 | ||
192 | const char *name; | |
193 | int name_len; | |
194 | const char *val; | |
195 | int val_len; | |
196 | int dirty; | |
197 | ||
198 | int should_free_name; | |
199 | int should_free_val; | |
200 | }; | |
201 | ||
3d14c5d2 YS |
202 | /* |
203 | * Ceph dentry state | |
204 | */ | |
205 | struct ceph_dentry_info { | |
b58dc410 | 206 | unsigned long flags; |
3d14c5d2 YS |
207 | struct ceph_mds_session *lease_session; |
208 | u32 lease_gen, lease_shared_gen; | |
209 | u32 lease_seq; | |
210 | unsigned long lease_renew_after, lease_renew_from; | |
211 | struct list_head lru; | |
212 | struct dentry *dentry; | |
213 | u64 time; | |
214 | u64 offset; | |
215 | }; | |
216 | ||
b58dc410 SW |
217 | /* |
218 | * dentry flags | |
219 | * | |
220 | * The locking for D_COMPLETE is a bit odd: | |
221 | * - we can clear it at almost any time (see ceph_d_prune) | |
222 | * - it is only meaningful if: | |
223 | * - we hold dir inode i_lock | |
224 | * - we hold dir FILE_SHARED caps | |
225 | * - the dentry D_COMPLETE is set | |
226 | */ | |
227 | #define CEPH_D_COMPLETE 1 /* if set, d_u.d_subdirs is complete directory */ | |
228 | ||
de57606c SW |
229 | struct ceph_inode_xattrs_info { |
230 | /* | |
231 | * (still encoded) xattr blob. we avoid the overhead of parsing | |
232 | * this until someone actually calls getxattr, etc. | |
233 | * | |
234 | * blob->vec.iov_len == 4 implies there are no xattrs; blob == | |
235 | * NULL means we don't know. | |
236 | */ | |
237 | struct ceph_buffer *blob, *prealloc_blob; | |
238 | ||
239 | struct rb_root index; | |
240 | bool dirty; | |
241 | int count; | |
242 | int names_size; | |
243 | int vals_size; | |
244 | u64 version, index_version; | |
245 | }; | |
246 | ||
247 | /* | |
248 | * Ceph inode. | |
249 | */ | |
de57606c SW |
250 | struct ceph_inode_info { |
251 | struct ceph_vino i_vino; /* ceph ino + snap */ | |
252 | ||
253 | u64 i_version; | |
254 | u32 i_time_warp_seq; | |
255 | ||
256 | unsigned i_ceph_flags; | |
257 | unsigned long i_release_count; | |
258 | ||
6c0f3af7 | 259 | struct ceph_dir_layout i_dir_layout; |
de57606c SW |
260 | struct ceph_file_layout i_layout; |
261 | char *i_symlink; | |
262 | ||
263 | /* for dirs */ | |
264 | struct timespec i_rctime; | |
265 | u64 i_rbytes, i_rfiles, i_rsubdirs; | |
266 | u64 i_files, i_subdirs; | |
267 | u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */ | |
268 | ||
269 | struct rb_root i_fragtree; | |
270 | struct mutex i_fragtree_mutex; | |
271 | ||
272 | struct ceph_inode_xattrs_info i_xattrs; | |
273 | ||
274 | /* capabilities. protected _both_ by i_lock and cap->session's | |
275 | * s_mutex. */ | |
276 | struct rb_root i_caps; /* cap list */ | |
277 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ | |
278 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ | |
279 | struct list_head i_dirty_item, i_flushing_item; | |
280 | u64 i_cap_flush_seq; | |
281 | /* we need to track cap writeback on a per-cap-bit basis, to allow | |
282 | * overlapping, pipelined cap flushes to the mds. we can probably | |
283 | * reduce the tid to 8 bits if we're concerned about inode size. */ | |
284 | u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; | |
285 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ | |
286 | unsigned long i_hold_caps_min; /* jiffies */ | |
287 | unsigned long i_hold_caps_max; /* jiffies */ | |
288 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ | |
289 | int i_cap_exporting_mds; /* to handle cap migration between */ | |
290 | unsigned i_cap_exporting_mseq; /* mds's. */ | |
291 | unsigned i_cap_exporting_issued; | |
292 | struct ceph_cap_reservation i_cap_migration_resv; | |
293 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ | |
7d8cb26d SW |
294 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or |
295 | dirty|flushing caps */ | |
de57606c SW |
296 | unsigned i_snap_caps; /* cap bits for snapped files */ |
297 | ||
298 | int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ | |
299 | ||
300 | u32 i_truncate_seq; /* last truncate to smaller size */ | |
301 | u64 i_truncate_size; /* and the size we last truncated down to */ | |
302 | int i_truncate_pending; /* still need to call vmtruncate */ | |
303 | ||
304 | u64 i_max_size; /* max file size authorized by mds */ | |
305 | u64 i_reported_size; /* (max_)size reported to or requested of mds */ | |
306 | u64 i_wanted_max_size; /* offset we'd like to write too */ | |
307 | u64 i_requested_max_size; /* max_size we've requested */ | |
308 | ||
309 | /* held references to caps */ | |
310 | int i_pin_ref; | |
d3d0720d | 311 | int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref; |
de57606c SW |
312 | int i_wrbuffer_ref, i_wrbuffer_ref_head; |
313 | u32 i_shared_gen; /* increment each time we get FILE_SHARED */ | |
cd045cb4 | 314 | u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ |
de57606c SW |
315 | u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ |
316 | ||
317 | struct list_head i_unsafe_writes; /* uncommitted sync writes */ | |
318 | struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ | |
319 | spinlock_t i_unsafe_lock; | |
320 | ||
321 | struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ | |
322 | int i_snap_realm_counter; /* snap realm (if caps) */ | |
323 | struct list_head i_snap_realm_item; | |
324 | struct list_head i_snap_flush_item; | |
325 | ||
326 | struct work_struct i_wb_work; /* writeback work */ | |
327 | struct work_struct i_pg_inv_work; /* page invalidation work */ | |
328 | ||
329 | struct work_struct i_vmtruncate_work; | |
330 | ||
331 | struct inode vfs_inode; /* at end */ | |
332 | }; | |
333 | ||
334 | static inline struct ceph_inode_info *ceph_inode(struct inode *inode) | |
335 | { | |
fbbccec9 | 336 | return container_of(inode, struct ceph_inode_info, vfs_inode); |
de57606c SW |
337 | } |
338 | ||
ad1fee96 YS |
339 | static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) |
340 | { | |
341 | return (struct ceph_fs_client *)inode->i_sb->s_fs_info; | |
342 | } | |
343 | ||
344 | static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) | |
345 | { | |
346 | return (struct ceph_fs_client *)sb->s_fs_info; | |
347 | } | |
348 | ||
3d14c5d2 YS |
349 | static inline struct ceph_vino ceph_vino(struct inode *inode) |
350 | { | |
351 | return ceph_inode(inode)->i_vino; | |
352 | } | |
353 | ||
354 | /* | |
355 | * ino_t is <64 bits on many architectures, blech. | |
356 | * | |
ad1fee96 YS |
357 | * i_ino (kernel inode) st_ino (userspace) |
358 | * i386 32 32 | |
359 | * x86_64+ino32 64 32 | |
360 | * x86_64 64 64 | |
361 | */ | |
3310f754 | 362 | static inline u32 ceph_ino_to_ino32(__u64 vino) |
ad1fee96 | 363 | { |
3310f754 AO |
364 | u32 ino = vino & 0xffffffff; |
365 | ino ^= vino >> 32; | |
ad1fee96 YS |
366 | if (!ino) |
367 | ino = 1; | |
368 | return ino; | |
369 | } | |
370 | ||
371 | /* | |
372 | * kernel i_ino value | |
3d14c5d2 YS |
373 | */ |
374 | static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) | |
375 | { | |
3d14c5d2 | 376 | #if BITS_PER_LONG == 32 |
3310f754 AO |
377 | return ceph_ino_to_ino32(vino.ino); |
378 | #else | |
379 | return (ino_t)vino.ino; | |
3d14c5d2 | 380 | #endif |
3d14c5d2 YS |
381 | } |
382 | ||
ad1fee96 YS |
383 | /* |
384 | * user-visible ino (stat, filldir) | |
385 | */ | |
386 | #if BITS_PER_LONG == 32 | |
387 | static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) | |
388 | { | |
389 | return ino; | |
390 | } | |
391 | #else | |
392 | static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) | |
393 | { | |
394 | if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) | |
395 | ino = ceph_ino_to_ino32(ino); | |
396 | return ino; | |
397 | } | |
398 | #endif | |
399 | ||
400 | ||
3d14c5d2 YS |
401 | /* for printf-style formatting */ |
402 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap | |
403 | ||
404 | static inline u64 ceph_ino(struct inode *inode) | |
405 | { | |
406 | return ceph_inode(inode)->i_vino.ino; | |
407 | } | |
408 | static inline u64 ceph_snap(struct inode *inode) | |
409 | { | |
410 | return ceph_inode(inode)->i_vino.snap; | |
411 | } | |
412 | ||
413 | static inline int ceph_ino_compare(struct inode *inode, void *data) | |
414 | { | |
415 | struct ceph_vino *pvino = (struct ceph_vino *)data; | |
416 | struct ceph_inode_info *ci = ceph_inode(inode); | |
417 | return ci->i_vino.ino == pvino->ino && | |
418 | ci->i_vino.snap == pvino->snap; | |
419 | } | |
420 | ||
421 | static inline struct inode *ceph_find_inode(struct super_block *sb, | |
422 | struct ceph_vino vino) | |
423 | { | |
424 | ino_t t = ceph_vino_to_ino(vino); | |
425 | return ilookup5(sb, t, ceph_ino_compare, &vino); | |
426 | } | |
427 | ||
428 | ||
429 | /* | |
430 | * Ceph inode. | |
431 | */ | |
432 | #define CEPH_I_COMPLETE 1 /* we have complete directory cached */ | |
433 | #define CEPH_I_NODELAY 4 /* do not delay cap release */ | |
434 | #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ | |
435 | #define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ | |
436 | ||
de57606c SW |
437 | static inline void ceph_i_clear(struct inode *inode, unsigned mask) |
438 | { | |
439 | struct ceph_inode_info *ci = ceph_inode(inode); | |
440 | ||
441 | spin_lock(&inode->i_lock); | |
442 | ci->i_ceph_flags &= ~mask; | |
443 | spin_unlock(&inode->i_lock); | |
444 | } | |
445 | ||
446 | static inline void ceph_i_set(struct inode *inode, unsigned mask) | |
447 | { | |
448 | struct ceph_inode_info *ci = ceph_inode(inode); | |
449 | ||
450 | spin_lock(&inode->i_lock); | |
451 | ci->i_ceph_flags |= mask; | |
452 | spin_unlock(&inode->i_lock); | |
453 | } | |
454 | ||
455 | static inline bool ceph_i_test(struct inode *inode, unsigned mask) | |
456 | { | |
457 | struct ceph_inode_info *ci = ceph_inode(inode); | |
458 | bool r; | |
459 | ||
efa4c120 | 460 | spin_lock(&inode->i_lock); |
de57606c | 461 | r = (ci->i_ceph_flags & mask) == mask; |
efa4c120 | 462 | spin_unlock(&inode->i_lock); |
de57606c SW |
463 | return r; |
464 | } | |
465 | ||
466 | ||
467 | /* find a specific frag @f */ | |
468 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, | |
469 | u32 f); | |
470 | ||
471 | /* | |
472 | * choose fragment for value @v. copy frag content to pfrag, if leaf | |
473 | * exists | |
474 | */ | |
475 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, | |
476 | struct ceph_inode_frag *pfrag, | |
477 | int *found); | |
478 | ||
de57606c SW |
479 | static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) |
480 | { | |
481 | return (struct ceph_dentry_info *)dentry->d_fsdata; | |
482 | } | |
483 | ||
484 | static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) | |
485 | { | |
486 | return ((loff_t)frag << 32) | (loff_t)off; | |
487 | } | |
488 | ||
de57606c SW |
489 | /* |
490 | * caps helpers | |
491 | */ | |
492 | static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) | |
493 | { | |
494 | return !RB_EMPTY_ROOT(&ci->i_caps); | |
495 | } | |
496 | ||
497 | extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); | |
498 | extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); | |
499 | extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, | |
500 | struct ceph_cap *cap); | |
501 | ||
502 | static inline int ceph_caps_issued(struct ceph_inode_info *ci) | |
503 | { | |
504 | int issued; | |
505 | spin_lock(&ci->vfs_inode.i_lock); | |
506 | issued = __ceph_caps_issued(ci, NULL); | |
507 | spin_unlock(&ci->vfs_inode.i_lock); | |
508 | return issued; | |
509 | } | |
510 | ||
511 | static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, | |
512 | int touch) | |
513 | { | |
514 | int r; | |
515 | spin_lock(&ci->vfs_inode.i_lock); | |
516 | r = __ceph_caps_issued_mask(ci, mask, touch); | |
517 | spin_unlock(&ci->vfs_inode.i_lock); | |
518 | return r; | |
519 | } | |
520 | ||
521 | static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) | |
522 | { | |
523 | return ci->i_dirty_caps | ci->i_flushing_caps; | |
524 | } | |
fca65b4a | 525 | extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); |
de57606c SW |
526 | |
527 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); | |
528 | extern int __ceph_caps_used(struct ceph_inode_info *ci); | |
529 | ||
530 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); | |
531 | ||
532 | /* | |
533 | * wanted, by virtue of open file modes AND cap refs (buffered/cached data) | |
534 | */ | |
535 | static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) | |
536 | { | |
537 | int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); | |
538 | if (w & CEPH_CAP_FILE_BUFFER) | |
539 | w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ | |
540 | return w; | |
541 | } | |
542 | ||
543 | /* what the mds thinks we want */ | |
544 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); | |
545 | ||
37151668 YS |
546 | extern void ceph_caps_init(struct ceph_mds_client *mdsc); |
547 | extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); | |
548 | extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); | |
549 | extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, | |
550 | struct ceph_cap_reservation *ctx, int need); | |
551 | extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, | |
552 | struct ceph_cap_reservation *ctx); | |
3d14c5d2 | 553 | extern void ceph_reservation_status(struct ceph_fs_client *client, |
de57606c | 554 | int *total, int *avail, int *used, |
85ccce43 | 555 | int *reserved, int *min); |
de57606c | 556 | |
de57606c | 557 | |
de57606c SW |
558 | |
559 | /* | |
560 | * we keep buffered readdir results attached to file->private_data | |
561 | */ | |
4918b6d1 | 562 | #define CEPH_F_SYNC 1 |
9cfa1098 | 563 | #define CEPH_F_ATEND 2 |
4918b6d1 | 564 | |
de57606c | 565 | struct ceph_file_info { |
252c6728 SW |
566 | short fmode; /* initialized on open */ |
567 | short flags; /* CEPH_F_* */ | |
de57606c SW |
568 | |
569 | /* readdir: position within the dir */ | |
570 | u32 frag; | |
571 | struct ceph_mds_request *last_readdir; | |
de57606c SW |
572 | |
573 | /* readdir: position within a frag */ | |
574 | unsigned offset; /* offset of last chunk, adjusted for . and .. */ | |
575 | u64 next_offset; /* offset of next chunk (last_name's + 1) */ | |
576 | char *last_name; /* last entry in previous chunk */ | |
577 | struct dentry *dentry; /* next dentry (for dcache readdir) */ | |
578 | unsigned long dir_release_count; | |
579 | ||
580 | /* used for -o dirstat read() on directory thing */ | |
581 | char *dir_info; | |
582 | int dir_info_len; | |
583 | }; | |
584 | ||
585 | ||
586 | ||
de57606c SW |
587 | /* |
588 | * A "snap realm" describes a subset of the file hierarchy sharing | |
589 | * the same set of snapshots that apply to it. The realms themselves | |
590 | * are organized into a hierarchy, such that children inherit (some of) | |
591 | * the snapshots of their parents. | |
592 | * | |
593 | * All inodes within the realm that have capabilities are linked into a | |
594 | * per-realm list. | |
595 | */ | |
596 | struct ceph_snap_realm { | |
597 | u64 ino; | |
598 | atomic_t nref; | |
a105f00c SW |
599 | struct rb_node node; |
600 | ||
de57606c SW |
601 | u64 created, seq; |
602 | u64 parent_ino; | |
603 | u64 parent_since; /* snapid when our current parent became so */ | |
604 | ||
605 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ | |
606 | int num_prior_parent_snaps; /* had prior to parent_since */ | |
607 | u64 *snaps; /* snaps specific to this realm */ | |
608 | int num_snaps; | |
609 | ||
610 | struct ceph_snap_realm *parent; | |
611 | struct list_head children; /* list of child realms */ | |
612 | struct list_head child_item; | |
613 | ||
614 | struct list_head empty_item; /* if i have ref==0 */ | |
615 | ||
ae00d4f3 SW |
616 | struct list_head dirty_item; /* if realm needs new context */ |
617 | ||
de57606c SW |
618 | /* the current set of snaps for this realm */ |
619 | struct ceph_snap_context *cached_context; | |
620 | ||
621 | struct list_head inodes_with_caps; | |
622 | spinlock_t inodes_with_caps_lock; | |
623 | }; | |
624 | ||
3d14c5d2 | 625 | static inline int default_congestion_kb(void) |
de57606c | 626 | { |
3d14c5d2 YS |
627 | int congestion_kb; |
628 | ||
629 | /* | |
630 | * Copied from NFS | |
631 | * | |
632 | * congestion size, scale with available memory. | |
633 | * | |
634 | * 64MB: 8192k | |
635 | * 128MB: 11585k | |
636 | * 256MB: 16384k | |
637 | * 512MB: 23170k | |
638 | * 1GB: 32768k | |
639 | * 2GB: 46340k | |
640 | * 4GB: 65536k | |
641 | * 8GB: 92681k | |
642 | * 16GB: 131072k | |
643 | * | |
644 | * This allows larger machines to have larger/more transfers. | |
645 | * Limit the default to 256M | |
646 | */ | |
647 | congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); | |
648 | if (congestion_kb > 256*1024) | |
649 | congestion_kb = 256*1024; | |
650 | ||
651 | return congestion_kb; | |
de57606c SW |
652 | } |
653 | ||
654 | ||
655 | ||
656 | /* snap.c */ | |
657 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, | |
658 | u64 ino); | |
659 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, | |
660 | struct ceph_snap_realm *realm); | |
661 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, | |
662 | struct ceph_snap_realm *realm); | |
663 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, | |
664 | void *p, void *e, bool deletion); | |
665 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, | |
2600d2dd | 666 | struct ceph_mds_session *session, |
de57606c | 667 | struct ceph_msg *msg); |
fc837c8f | 668 | extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); |
de57606c SW |
669 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, |
670 | struct ceph_cap_snap *capsnap); | |
671 | extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); | |
672 | ||
673 | /* | |
674 | * a cap_snap is "pending" if it is still awaiting an in-progress | |
675 | * sync write (that may/may not still update size, mtime, etc.). | |
676 | */ | |
677 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) | |
678 | { | |
679 | return !list_empty(&ci->i_cap_snaps) && | |
680 | list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, | |
681 | ci_item)->writing; | |
682 | } | |
683 | ||
de57606c SW |
684 | /* inode.c */ |
685 | extern const struct inode_operations ceph_file_iops; | |
686 | ||
687 | extern struct inode *ceph_alloc_inode(struct super_block *sb); | |
688 | extern void ceph_destroy_inode(struct inode *inode); | |
689 | ||
690 | extern struct inode *ceph_get_inode(struct super_block *sb, | |
691 | struct ceph_vino vino); | |
692 | extern struct inode *ceph_get_snapdir(struct inode *parent); | |
693 | extern int ceph_fill_file_size(struct inode *inode, int issued, | |
694 | u32 truncate_seq, u64 truncate_size, u64 size); | |
695 | extern void ceph_fill_file_time(struct inode *inode, int issued, | |
696 | u64 time_warp_seq, struct timespec *ctime, | |
697 | struct timespec *mtime, struct timespec *atime); | |
698 | extern int ceph_fill_trace(struct super_block *sb, | |
699 | struct ceph_mds_request *req, | |
700 | struct ceph_mds_session *session); | |
701 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, | |
702 | struct ceph_mds_session *session); | |
703 | ||
704 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); | |
705 | ||
706 | extern int ceph_inode_set_size(struct inode *inode, loff_t size); | |
de57606c | 707 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); |
3c6f6b79 SW |
708 | extern void ceph_queue_vmtruncate(struct inode *inode); |
709 | ||
710 | extern void ceph_queue_invalidate(struct inode *inode); | |
711 | extern void ceph_queue_writeback(struct inode *inode); | |
de57606c SW |
712 | |
713 | extern int ceph_do_getattr(struct inode *inode, int mask); | |
10556cb2 | 714 | extern int ceph_permission(struct inode *inode, int mask); |
de57606c SW |
715 | extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); |
716 | extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, | |
717 | struct kstat *stat); | |
718 | ||
719 | /* xattr.c */ | |
720 | extern int ceph_setxattr(struct dentry *, const char *, const void *, | |
721 | size_t, int); | |
722 | extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); | |
723 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); | |
724 | extern int ceph_removexattr(struct dentry *, const char *); | |
725 | extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); | |
726 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); | |
727 | ||
728 | /* caps.c */ | |
729 | extern const char *ceph_cap_string(int c); | |
730 | extern void ceph_handle_caps(struct ceph_mds_session *session, | |
731 | struct ceph_msg *msg); | |
732 | extern int ceph_add_cap(struct inode *inode, | |
733 | struct ceph_mds_session *session, u64 cap_id, | |
734 | int fmode, unsigned issued, unsigned wanted, | |
735 | unsigned cap, unsigned seq, u64 realmino, int flags, | |
736 | struct ceph_cap_reservation *caps_reservation); | |
7c1332b8 | 737 | extern void __ceph_remove_cap(struct ceph_cap *cap); |
de57606c SW |
738 | static inline void ceph_remove_cap(struct ceph_cap *cap) |
739 | { | |
740 | struct inode *inode = &cap->ci->vfs_inode; | |
741 | spin_lock(&inode->i_lock); | |
7c1332b8 | 742 | __ceph_remove_cap(cap); |
de57606c SW |
743 | spin_unlock(&inode->i_lock); |
744 | } | |
37151668 YS |
745 | extern void ceph_put_cap(struct ceph_mds_client *mdsc, |
746 | struct ceph_cap *cap); | |
de57606c SW |
747 | |
748 | extern void ceph_queue_caps_release(struct inode *inode); | |
f1a3d572 | 749 | extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); |
02c24a82 JB |
750 | extern int ceph_fsync(struct file *file, loff_t start, loff_t end, |
751 | int datasync); | |
de57606c SW |
752 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, |
753 | struct ceph_mds_session *session); | |
2bc50259 GF |
754 | extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, |
755 | int mds); | |
de57606c SW |
756 | extern int ceph_get_cap_mds(struct inode *inode); |
757 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); | |
758 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); | |
759 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, | |
760 | struct ceph_snap_context *snapc); | |
761 | extern void __ceph_flush_snaps(struct ceph_inode_info *ci, | |
e835124c SW |
762 | struct ceph_mds_session **psession, |
763 | int again); | |
de57606c SW |
764 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, |
765 | struct ceph_mds_session *session); | |
afcdaea3 SW |
766 | extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); |
767 | extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); | |
de57606c SW |
768 | |
769 | extern int ceph_encode_inode_release(void **p, struct inode *inode, | |
770 | int mds, int drop, int unless, int force); | |
771 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, | |
772 | int mds, int drop, int unless); | |
773 | ||
774 | extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, | |
775 | int *got, loff_t endoff); | |
776 | ||
777 | /* for counting open files by mode */ | |
778 | static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) | |
779 | { | |
780 | ci->i_nr_by_mode[mode]++; | |
781 | } | |
782 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); | |
783 | ||
784 | /* addr.c */ | |
785 | extern const struct address_space_operations ceph_aops; | |
786 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); | |
787 | ||
788 | /* file.c */ | |
789 | extern const struct file_operations ceph_file_fops; | |
790 | extern const struct address_space_operations ceph_aops; | |
3d14c5d2 YS |
791 | extern int ceph_copy_to_page_vector(struct page **pages, |
792 | const char *data, | |
793 | loff_t off, size_t len); | |
794 | extern int ceph_copy_from_page_vector(struct page **pages, | |
795 | char *data, | |
796 | loff_t off, size_t len); | |
797 | extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags); | |
de57606c SW |
798 | extern int ceph_open(struct inode *inode, struct file *file); |
799 | extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry, | |
800 | struct nameidata *nd, int mode, | |
801 | int locked_dir); | |
802 | extern int ceph_release(struct inode *inode, struct file *filp); | |
de57606c SW |
803 | |
804 | /* dir.c */ | |
805 | extern const struct file_operations ceph_dir_fops; | |
806 | extern const struct inode_operations ceph_dir_iops; | |
52dfb8ac | 807 | extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, |
de57606c SW |
808 | ceph_snapdir_dentry_ops; |
809 | ||
810 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); | |
468640e3 SW |
811 | extern int ceph_handle_snapdir(struct ceph_mds_request *req, |
812 | struct dentry *dentry, int err); | |
de57606c SW |
813 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, |
814 | struct dentry *dentry, int err); | |
815 | ||
816 | extern void ceph_dentry_lru_add(struct dentry *dn); | |
817 | extern void ceph_dentry_lru_touch(struct dentry *dn); | |
818 | extern void ceph_dentry_lru_del(struct dentry *dn); | |
81a6cf2d | 819 | extern void ceph_invalidate_dentry_lease(struct dentry *dentry); |
e5f86dc3 | 820 | extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); |
5f21c96d | 821 | extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry); |
de57606c SW |
822 | |
823 | /* | |
824 | * our d_ops vary depending on whether the inode is live, | |
825 | * snapshotted (read-only), or a virtual ".snap" directory. | |
826 | */ | |
827 | int ceph_init_dentry(struct dentry *dentry); | |
828 | ||
829 | ||
830 | /* ioctl.c */ | |
831 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | |
832 | ||
833 | /* export.c */ | |
834 | extern const struct export_operations ceph_export_ops; | |
835 | ||
40819f6f GF |
836 | /* locks.c */ |
837 | extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); | |
838 | extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); | |
839 | extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); | |
840 | extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p, | |
841 | int p_locks, int f_locks); | |
842 | extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); | |
843 | ||
3d14c5d2 YS |
844 | /* debugfs.c */ |
845 | extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); | |
846 | extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); | |
847 | ||
de57606c | 848 | #endif /* _FS_CEPH_SUPER_H */ |