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 | 18 | |
99ccbd22 MT |
19 | #ifdef CONFIG_CEPH_FSCACHE |
20 | #include <linux/fscache.h> | |
21 | #endif | |
22 | ||
de57606c SW |
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. */ | |
92a49fb0 | 28 | #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ |
de57606c SW |
29 | #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) |
30 | ||
3d14c5d2 YS |
31 | #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ |
32 | #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ | |
33 | #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ | |
ad1fee96 | 34 | #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ |
a40dc6cc | 35 | #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ |
99ccbd22 | 36 | #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ |
6a259382 | 37 | |
3d14c5d2 | 38 | #define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES) |
de57606c | 39 | |
3d14c5d2 YS |
40 | #define ceph_set_mount_opt(fsc, opt) \ |
41 | (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt; | |
42 | #define ceph_test_mount_opt(fsc, opt) \ | |
43 | (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) | |
de57606c | 44 | |
83817e35 SW |
45 | #define CEPH_RSIZE_DEFAULT 0 /* max read size */ |
46 | #define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */ | |
3d14c5d2 YS |
47 | #define CEPH_MAX_READDIR_DEFAULT 1024 |
48 | #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) | |
49 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" | |
de57606c | 50 | |
3d14c5d2 | 51 | struct ceph_mount_options { |
6e19a16e | 52 | int flags; |
3d14c5d2 YS |
53 | int sb_flags; |
54 | ||
83817e35 SW |
55 | int wsize; /* max write size */ |
56 | int rsize; /* max read size */ | |
57 | int rasize; /* max readahead */ | |
6e19a16e SW |
58 | int congestion_kb; /* max writeback in flight */ |
59 | int caps_wanted_delay_min, caps_wanted_delay_max; | |
60 | int cap_release_safety; | |
23804d91 SW |
61 | int max_readdir; /* max readdir result (entires) */ |
62 | int max_readdir_bytes; /* max readdir result (bytes) */ | |
de57606c | 63 | |
3d14c5d2 YS |
64 | /* |
65 | * everything above this point can be memcmp'd; everything below | |
66 | * is handled in compare_mount_options() | |
67 | */ | |
de57606c | 68 | |
3d14c5d2 YS |
69 | char *snapdir_name; /* default ".snap" */ |
70 | }; | |
de57606c | 71 | |
3d14c5d2 | 72 | struct ceph_fs_client { |
de57606c SW |
73 | struct super_block *sb; |
74 | ||
3d14c5d2 YS |
75 | struct ceph_mount_options *mount_options; |
76 | struct ceph_client *client; | |
de57606c | 77 | |
3d14c5d2 | 78 | unsigned long mount_state; |
85ccce43 SW |
79 | int min_caps; /* min caps i added */ |
80 | ||
3d14c5d2 | 81 | struct ceph_mds_client *mdsc; |
de57606c SW |
82 | |
83 | /* writeback */ | |
84 | mempool_t *wb_pagevec_pool; | |
85 | struct workqueue_struct *wb_wq; | |
86 | struct workqueue_struct *pg_inv_wq; | |
87 | struct workqueue_struct *trunc_wq; | |
2baba250 | 88 | atomic_long_t writeback_count; |
de57606c SW |
89 | |
90 | struct backing_dev_info backing_dev_info; | |
0743304d SW |
91 | |
92 | #ifdef CONFIG_DEBUG_FS | |
3d14c5d2 | 93 | struct dentry *debugfs_dentry_lru, *debugfs_caps; |
2baba250 | 94 | struct dentry *debugfs_congestion_kb; |
06edf046 | 95 | struct dentry *debugfs_bdi; |
3d14c5d2 | 96 | struct dentry *debugfs_mdsc, *debugfs_mdsmap; |
0743304d | 97 | #endif |
99ccbd22 MT |
98 | |
99 | #ifdef CONFIG_CEPH_FSCACHE | |
100 | struct fscache_cookie *fscache; | |
101 | struct workqueue_struct *revalidate_wq; | |
102 | #endif | |
de57606c SW |
103 | }; |
104 | ||
3d14c5d2 | 105 | |
de57606c SW |
106 | /* |
107 | * File i/o capability. This tracks shared state with the metadata | |
108 | * server that allows us to cache or writeback attributes or to read | |
109 | * and write data. For any given inode, we should have one or more | |
110 | * capabilities, one issued by each metadata server, and our | |
111 | * cumulative access is the OR of all issued capabilities. | |
112 | * | |
113 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds | |
114 | * session capability lists. | |
115 | */ | |
116 | struct ceph_cap { | |
117 | struct ceph_inode_info *ci; | |
118 | struct rb_node ci_node; /* per-ci cap tree */ | |
119 | struct ceph_mds_session *session; | |
120 | struct list_head session_caps; /* per-session caplist */ | |
121 | int mds; | |
122 | u64 cap_id; /* unique cap id (mds provided) */ | |
123 | int issued; /* latest, from the mds */ | |
124 | int implemented; /* implemented superset of issued (for revocation) */ | |
125 | int mds_wanted; | |
685f9a5d SW |
126 | u32 seq, issue_seq, mseq; |
127 | u32 cap_gen; /* active/stale cycle */ | |
de57606c SW |
128 | unsigned long last_used; |
129 | struct list_head caps_item; | |
130 | }; | |
131 | ||
132 | #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ | |
133 | #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ | |
134 | #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ | |
135 | ||
136 | /* | |
137 | * Snapped cap state that is pending flush to mds. When a snapshot occurs, | |
138 | * we first complete any in-process sync writes and writeback any dirty | |
139 | * data before flushing the snapped state (tracked here) back to the MDS. | |
140 | */ | |
141 | struct ceph_cap_snap { | |
142 | atomic_t nref; | |
143 | struct ceph_inode_info *ci; | |
144 | struct list_head ci_item, flushing_item; | |
145 | ||
146 | u64 follows, flush_tid; | |
147 | int issued, dirty; | |
148 | struct ceph_snap_context *context; | |
149 | ||
5706b27d | 150 | umode_t mode; |
05cb11c1 EB |
151 | kuid_t uid; |
152 | kgid_t gid; | |
de57606c | 153 | |
4a625be4 | 154 | struct ceph_buffer *xattr_blob; |
de57606c SW |
155 | u64 xattr_version; |
156 | ||
157 | u64 size; | |
158 | struct timespec mtime, atime, ctime; | |
159 | u64 time_warp_seq; | |
160 | int writing; /* a sync write is still in progress */ | |
161 | int dirty_pages; /* dirty pages awaiting writeback */ | |
162 | }; | |
163 | ||
164 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) | |
165 | { | |
4a625be4 SW |
166 | if (atomic_dec_and_test(&capsnap->nref)) { |
167 | if (capsnap->xattr_blob) | |
168 | ceph_buffer_put(capsnap->xattr_blob); | |
de57606c | 169 | kfree(capsnap); |
4a625be4 | 170 | } |
de57606c SW |
171 | } |
172 | ||
173 | /* | |
174 | * The frag tree describes how a directory is fragmented, potentially across | |
175 | * multiple metadata servers. It is also used to indicate points where | |
176 | * metadata authority is delegated, and whether/where metadata is replicated. | |
177 | * | |
178 | * A _leaf_ frag will be present in the i_fragtree IFF there is | |
179 | * delegation info. That is, if mds >= 0 || ndist > 0. | |
180 | */ | |
181 | #define CEPH_MAX_DIRFRAG_REP 4 | |
182 | ||
183 | struct ceph_inode_frag { | |
184 | struct rb_node node; | |
185 | ||
186 | /* fragtree state */ | |
187 | u32 frag; | |
188 | int split_by; /* i.e. 2^(split_by) children */ | |
189 | ||
190 | /* delegation and replication info */ | |
191 | int mds; /* -1 if same authority as parent */ | |
192 | int ndist; /* >0 if replicated */ | |
193 | int dist[CEPH_MAX_DIRFRAG_REP]; | |
194 | }; | |
195 | ||
196 | /* | |
197 | * We cache inode xattrs as an encoded blob until they are first used, | |
198 | * at which point we parse them into an rbtree. | |
199 | */ | |
200 | struct ceph_inode_xattr { | |
201 | struct rb_node node; | |
202 | ||
203 | const char *name; | |
204 | int name_len; | |
205 | const char *val; | |
206 | int val_len; | |
207 | int dirty; | |
208 | ||
209 | int should_free_name; | |
210 | int should_free_val; | |
211 | }; | |
212 | ||
3d14c5d2 YS |
213 | /* |
214 | * Ceph dentry state | |
215 | */ | |
216 | struct ceph_dentry_info { | |
217 | struct ceph_mds_session *lease_session; | |
218 | u32 lease_gen, lease_shared_gen; | |
219 | u32 lease_seq; | |
220 | unsigned long lease_renew_after, lease_renew_from; | |
221 | struct list_head lru; | |
222 | struct dentry *dentry; | |
223 | u64 time; | |
224 | u64 offset; | |
225 | }; | |
226 | ||
de57606c SW |
227 | struct ceph_inode_xattrs_info { |
228 | /* | |
229 | * (still encoded) xattr blob. we avoid the overhead of parsing | |
230 | * this until someone actually calls getxattr, etc. | |
231 | * | |
232 | * blob->vec.iov_len == 4 implies there are no xattrs; blob == | |
233 | * NULL means we don't know. | |
234 | */ | |
235 | struct ceph_buffer *blob, *prealloc_blob; | |
236 | ||
237 | struct rb_root index; | |
238 | bool dirty; | |
239 | int count; | |
240 | int names_size; | |
241 | int vals_size; | |
242 | u64 version, index_version; | |
243 | }; | |
244 | ||
245 | /* | |
246 | * Ceph inode. | |
247 | */ | |
de57606c SW |
248 | struct ceph_inode_info { |
249 | struct ceph_vino i_vino; /* ceph ino + snap */ | |
250 | ||
be655596 SW |
251 | spinlock_t i_ceph_lock; |
252 | ||
de57606c SW |
253 | u64 i_version; |
254 | u32 i_time_warp_seq; | |
255 | ||
256 | unsigned i_ceph_flags; | |
2f276c51 YZ |
257 | atomic_t i_release_count; |
258 | atomic_t i_complete_count; | |
de57606c | 259 | |
6c0f3af7 | 260 | struct ceph_dir_layout i_dir_layout; |
de57606c SW |
261 | struct ceph_file_layout i_layout; |
262 | char *i_symlink; | |
263 | ||
264 | /* for dirs */ | |
265 | struct timespec i_rctime; | |
266 | u64 i_rbytes, i_rfiles, i_rsubdirs; | |
267 | u64 i_files, i_subdirs; | |
2f276c51 | 268 | u64 i_max_offset; /* largest readdir offset, set with complete dir */ |
de57606c SW |
269 | |
270 | struct rb_root i_fragtree; | |
271 | struct mutex i_fragtree_mutex; | |
272 | ||
273 | struct ceph_inode_xattrs_info i_xattrs; | |
274 | ||
be655596 | 275 | /* capabilities. protected _both_ by i_ceph_lock and cap->session's |
de57606c SW |
276 | * s_mutex. */ |
277 | struct rb_root i_caps; /* cap list */ | |
278 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ | |
279 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ | |
280 | struct list_head i_dirty_item, i_flushing_item; | |
281 | u64 i_cap_flush_seq; | |
282 | /* we need to track cap writeback on a per-cap-bit basis, to allow | |
283 | * overlapping, pipelined cap flushes to the mds. we can probably | |
284 | * reduce the tid to 8 bits if we're concerned about inode size. */ | |
285 | u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; | |
286 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ | |
287 | unsigned long i_hold_caps_min; /* jiffies */ | |
288 | unsigned long i_hold_caps_max; /* jiffies */ | |
289 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ | |
de57606c SW |
290 | struct ceph_cap_reservation i_cap_migration_resv; |
291 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ | |
7d8cb26d SW |
292 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or |
293 | dirty|flushing caps */ | |
de57606c | 294 | unsigned i_snap_caps; /* cap bits for snapped files */ |
11df2dfb | 295 | unsigned i_cap_exporting_issued; |
de57606c SW |
296 | |
297 | int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ | |
298 | ||
b0d7c223 | 299 | struct mutex i_truncate_mutex; |
de57606c SW |
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 | ||
99ccbd22 MT |
331 | #ifdef CONFIG_CEPH_FSCACHE |
332 | struct fscache_cookie *fscache; | |
333 | u32 i_fscache_gen; /* sequence, for delayed fscache validate */ | |
334 | struct work_struct i_revalidate_work; | |
335 | #endif | |
de57606c SW |
336 | struct inode vfs_inode; /* at end */ |
337 | }; | |
338 | ||
339 | static inline struct ceph_inode_info *ceph_inode(struct inode *inode) | |
340 | { | |
fbbccec9 | 341 | return container_of(inode, struct ceph_inode_info, vfs_inode); |
de57606c SW |
342 | } |
343 | ||
ad1fee96 YS |
344 | static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) |
345 | { | |
346 | return (struct ceph_fs_client *)inode->i_sb->s_fs_info; | |
347 | } | |
348 | ||
349 | static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) | |
350 | { | |
351 | return (struct ceph_fs_client *)sb->s_fs_info; | |
352 | } | |
353 | ||
3d14c5d2 YS |
354 | static inline struct ceph_vino ceph_vino(struct inode *inode) |
355 | { | |
356 | return ceph_inode(inode)->i_vino; | |
357 | } | |
358 | ||
359 | /* | |
360 | * ino_t is <64 bits on many architectures, blech. | |
361 | * | |
ad1fee96 YS |
362 | * i_ino (kernel inode) st_ino (userspace) |
363 | * i386 32 32 | |
364 | * x86_64+ino32 64 32 | |
365 | * x86_64 64 64 | |
366 | */ | |
3310f754 | 367 | static inline u32 ceph_ino_to_ino32(__u64 vino) |
ad1fee96 | 368 | { |
3310f754 AO |
369 | u32 ino = vino & 0xffffffff; |
370 | ino ^= vino >> 32; | |
ad1fee96 | 371 | if (!ino) |
a661fc56 | 372 | ino = 2; |
ad1fee96 YS |
373 | return ino; |
374 | } | |
375 | ||
376 | /* | |
377 | * kernel i_ino value | |
3d14c5d2 YS |
378 | */ |
379 | static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) | |
380 | { | |
3d14c5d2 | 381 | #if BITS_PER_LONG == 32 |
3310f754 AO |
382 | return ceph_ino_to_ino32(vino.ino); |
383 | #else | |
384 | return (ino_t)vino.ino; | |
3d14c5d2 | 385 | #endif |
3d14c5d2 YS |
386 | } |
387 | ||
ad1fee96 YS |
388 | /* |
389 | * user-visible ino (stat, filldir) | |
390 | */ | |
391 | #if BITS_PER_LONG == 32 | |
392 | static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) | |
393 | { | |
394 | return ino; | |
395 | } | |
396 | #else | |
397 | static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) | |
398 | { | |
399 | if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) | |
400 | ino = ceph_ino_to_ino32(ino); | |
401 | return ino; | |
402 | } | |
403 | #endif | |
404 | ||
405 | ||
3d14c5d2 YS |
406 | /* for printf-style formatting */ |
407 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap | |
408 | ||
409 | static inline u64 ceph_ino(struct inode *inode) | |
410 | { | |
411 | return ceph_inode(inode)->i_vino.ino; | |
412 | } | |
413 | static inline u64 ceph_snap(struct inode *inode) | |
414 | { | |
415 | return ceph_inode(inode)->i_vino.snap; | |
416 | } | |
417 | ||
418 | static inline int ceph_ino_compare(struct inode *inode, void *data) | |
419 | { | |
420 | struct ceph_vino *pvino = (struct ceph_vino *)data; | |
421 | struct ceph_inode_info *ci = ceph_inode(inode); | |
422 | return ci->i_vino.ino == pvino->ino && | |
423 | ci->i_vino.snap == pvino->snap; | |
424 | } | |
425 | ||
426 | static inline struct inode *ceph_find_inode(struct super_block *sb, | |
427 | struct ceph_vino vino) | |
428 | { | |
429 | ino_t t = ceph_vino_to_ino(vino); | |
430 | return ilookup5(sb, t, ceph_ino_compare, &vino); | |
431 | } | |
432 | ||
433 | ||
434 | /* | |
435 | * Ceph inode. | |
436 | */ | |
3d14c5d2 YS |
437 | #define CEPH_I_NODELAY 4 /* do not delay cap release */ |
438 | #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ | |
439 | #define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ | |
440 | ||
2f276c51 YZ |
441 | static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, |
442 | int release_count) | |
de57606c | 443 | { |
2f276c51 | 444 | atomic_set(&ci->i_complete_count, release_count); |
de57606c SW |
445 | } |
446 | ||
2f276c51 | 447 | static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) |
de57606c | 448 | { |
2f276c51 YZ |
449 | atomic_inc(&ci->i_release_count); |
450 | } | |
de57606c | 451 | |
2f276c51 YZ |
452 | static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) |
453 | { | |
454 | return atomic_read(&ci->i_complete_count) == | |
455 | atomic_read(&ci->i_release_count); | |
de57606c SW |
456 | } |
457 | ||
2f276c51 | 458 | static inline void ceph_dir_clear_complete(struct inode *inode) |
de57606c | 459 | { |
2f276c51 YZ |
460 | __ceph_dir_clear_complete(ceph_inode(inode)); |
461 | } | |
de57606c | 462 | |
2f276c51 YZ |
463 | static inline bool ceph_dir_is_complete(struct inode *inode) |
464 | { | |
465 | return __ceph_dir_is_complete(ceph_inode(inode)); | |
de57606c SW |
466 | } |
467 | ||
468 | ||
469 | /* find a specific frag @f */ | |
470 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, | |
471 | u32 f); | |
472 | ||
473 | /* | |
474 | * choose fragment for value @v. copy frag content to pfrag, if leaf | |
475 | * exists | |
476 | */ | |
477 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, | |
478 | struct ceph_inode_frag *pfrag, | |
479 | int *found); | |
480 | ||
de57606c SW |
481 | static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) |
482 | { | |
483 | return (struct ceph_dentry_info *)dentry->d_fsdata; | |
484 | } | |
485 | ||
486 | static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) | |
487 | { | |
488 | return ((loff_t)frag << 32) | (loff_t)off; | |
489 | } | |
490 | ||
de57606c SW |
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; | |
be655596 | 507 | spin_lock(&ci->i_ceph_lock); |
de57606c | 508 | issued = __ceph_caps_issued(ci, NULL); |
be655596 | 509 | spin_unlock(&ci->i_ceph_lock); |
de57606c SW |
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; | |
be655596 | 517 | spin_lock(&ci->i_ceph_lock); |
de57606c | 518 | r = __ceph_caps_issued_mask(ci, mask, touch); |
be655596 | 519 | spin_unlock(&ci->i_ceph_lock); |
de57606c SW |
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 | } | |
fca65b4a | 527 | extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); |
de57606c | 528 | |
9563f88c YZ |
529 | extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, |
530 | struct ceph_cap *ocap, int mask); | |
de57606c SW |
531 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); |
532 | extern int __ceph_caps_used(struct ceph_inode_info *ci); | |
533 | ||
534 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); | |
535 | ||
536 | /* | |
537 | * wanted, by virtue of open file modes AND cap refs (buffered/cached data) | |
538 | */ | |
539 | static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) | |
540 | { | |
541 | int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); | |
542 | if (w & CEPH_CAP_FILE_BUFFER) | |
543 | w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ | |
544 | return w; | |
545 | } | |
546 | ||
547 | /* what the mds thinks we want */ | |
548 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); | |
549 | ||
37151668 YS |
550 | extern void ceph_caps_init(struct ceph_mds_client *mdsc); |
551 | extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); | |
552 | extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); | |
93faca6e | 553 | extern void ceph_reserve_caps(struct ceph_mds_client *mdsc, |
37151668 YS |
554 | struct ceph_cap_reservation *ctx, int need); |
555 | extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, | |
556 | struct ceph_cap_reservation *ctx); | |
3d14c5d2 | 557 | extern void ceph_reservation_status(struct ceph_fs_client *client, |
de57606c | 558 | int *total, int *avail, int *used, |
85ccce43 | 559 | int *reserved, int *min); |
de57606c | 560 | |
de57606c | 561 | |
de57606c SW |
562 | |
563 | /* | |
564 | * we keep buffered readdir results attached to file->private_data | |
565 | */ | |
4918b6d1 | 566 | #define CEPH_F_SYNC 1 |
9cfa1098 | 567 | #define CEPH_F_ATEND 2 |
4918b6d1 | 568 | |
de57606c | 569 | struct ceph_file_info { |
252c6728 SW |
570 | short fmode; /* initialized on open */ |
571 | short flags; /* CEPH_F_* */ | |
de57606c SW |
572 | |
573 | /* readdir: position within the dir */ | |
574 | u32 frag; | |
575 | struct ceph_mds_request *last_readdir; | |
de57606c SW |
576 | |
577 | /* readdir: position within a frag */ | |
578 | unsigned offset; /* offset of last chunk, adjusted for . and .. */ | |
579 | u64 next_offset; /* offset of next chunk (last_name's + 1) */ | |
580 | char *last_name; /* last entry in previous chunk */ | |
581 | struct dentry *dentry; /* next dentry (for dcache readdir) */ | |
2f276c51 | 582 | int dir_release_count; |
de57606c SW |
583 | |
584 | /* used for -o dirstat read() on directory thing */ | |
585 | char *dir_info; | |
586 | int dir_info_len; | |
587 | }; | |
588 | ||
589 | ||
590 | ||
de57606c SW |
591 | /* |
592 | * A "snap realm" describes a subset of the file hierarchy sharing | |
593 | * the same set of snapshots that apply to it. The realms themselves | |
594 | * are organized into a hierarchy, such that children inherit (some of) | |
595 | * the snapshots of their parents. | |
596 | * | |
597 | * All inodes within the realm that have capabilities are linked into a | |
598 | * per-realm list. | |
599 | */ | |
600 | struct ceph_snap_realm { | |
601 | u64 ino; | |
602 | atomic_t nref; | |
a105f00c SW |
603 | struct rb_node node; |
604 | ||
de57606c SW |
605 | u64 created, seq; |
606 | u64 parent_ino; | |
607 | u64 parent_since; /* snapid when our current parent became so */ | |
608 | ||
609 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ | |
aa711ee3 | 610 | u32 num_prior_parent_snaps; /* had prior to parent_since */ |
de57606c | 611 | u64 *snaps; /* snaps specific to this realm */ |
aa711ee3 | 612 | u32 num_snaps; |
de57606c SW |
613 | |
614 | struct ceph_snap_realm *parent; | |
615 | struct list_head children; /* list of child realms */ | |
616 | struct list_head child_item; | |
617 | ||
618 | struct list_head empty_item; /* if i have ref==0 */ | |
619 | ||
ae00d4f3 SW |
620 | struct list_head dirty_item; /* if realm needs new context */ |
621 | ||
de57606c SW |
622 | /* the current set of snaps for this realm */ |
623 | struct ceph_snap_context *cached_context; | |
624 | ||
625 | struct list_head inodes_with_caps; | |
626 | spinlock_t inodes_with_caps_lock; | |
627 | }; | |
628 | ||
3d14c5d2 | 629 | static inline int default_congestion_kb(void) |
de57606c | 630 | { |
3d14c5d2 YS |
631 | int congestion_kb; |
632 | ||
633 | /* | |
634 | * Copied from NFS | |
635 | * | |
636 | * congestion size, scale with available memory. | |
637 | * | |
638 | * 64MB: 8192k | |
639 | * 128MB: 11585k | |
640 | * 256MB: 16384k | |
641 | * 512MB: 23170k | |
642 | * 1GB: 32768k | |
643 | * 2GB: 46340k | |
644 | * 4GB: 65536k | |
645 | * 8GB: 92681k | |
646 | * 16GB: 131072k | |
647 | * | |
648 | * This allows larger machines to have larger/more transfers. | |
649 | * Limit the default to 256M | |
650 | */ | |
651 | congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); | |
652 | if (congestion_kb > 256*1024) | |
653 | congestion_kb = 256*1024; | |
654 | ||
655 | return congestion_kb; | |
de57606c SW |
656 | } |
657 | ||
658 | ||
659 | ||
660 | /* snap.c */ | |
661 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, | |
662 | u64 ino); | |
663 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, | |
664 | struct ceph_snap_realm *realm); | |
665 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, | |
666 | struct ceph_snap_realm *realm); | |
667 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, | |
668 | void *p, void *e, bool deletion); | |
669 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, | |
2600d2dd | 670 | struct ceph_mds_session *session, |
de57606c | 671 | struct ceph_msg *msg); |
fc837c8f | 672 | extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); |
de57606c SW |
673 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, |
674 | struct ceph_cap_snap *capsnap); | |
675 | extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); | |
676 | ||
677 | /* | |
678 | * a cap_snap is "pending" if it is still awaiting an in-progress | |
679 | * sync write (that may/may not still update size, mtime, etc.). | |
680 | */ | |
681 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) | |
682 | { | |
683 | return !list_empty(&ci->i_cap_snaps) && | |
684 | list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, | |
685 | ci_item)->writing; | |
686 | } | |
687 | ||
de57606c SW |
688 | /* inode.c */ |
689 | extern const struct inode_operations ceph_file_iops; | |
690 | ||
691 | extern struct inode *ceph_alloc_inode(struct super_block *sb); | |
692 | extern void ceph_destroy_inode(struct inode *inode); | |
9f12bd11 | 693 | extern int ceph_drop_inode(struct inode *inode); |
de57606c SW |
694 | |
695 | extern struct inode *ceph_get_inode(struct super_block *sb, | |
696 | struct ceph_vino vino); | |
697 | extern struct inode *ceph_get_snapdir(struct inode *parent); | |
698 | extern int ceph_fill_file_size(struct inode *inode, int issued, | |
699 | u32 truncate_seq, u64 truncate_size, u64 size); | |
700 | extern void ceph_fill_file_time(struct inode *inode, int issued, | |
701 | u64 time_warp_seq, struct timespec *ctime, | |
702 | struct timespec *mtime, struct timespec *atime); | |
703 | extern int ceph_fill_trace(struct super_block *sb, | |
704 | struct ceph_mds_request *req, | |
705 | struct ceph_mds_session *session); | |
706 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, | |
707 | struct ceph_mds_session *session); | |
708 | ||
709 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); | |
710 | ||
711 | extern int ceph_inode_set_size(struct inode *inode, loff_t size); | |
b415bf4f | 712 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); |
3c6f6b79 SW |
713 | extern void ceph_queue_vmtruncate(struct inode *inode); |
714 | ||
715 | extern void ceph_queue_invalidate(struct inode *inode); | |
716 | extern void ceph_queue_writeback(struct inode *inode); | |
de57606c SW |
717 | |
718 | extern int ceph_do_getattr(struct inode *inode, int mask); | |
10556cb2 | 719 | extern int ceph_permission(struct inode *inode, int mask); |
de57606c SW |
720 | extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); |
721 | extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, | |
722 | struct kstat *stat); | |
723 | ||
724 | /* xattr.c */ | |
725 | extern int ceph_setxattr(struct dentry *, const char *, const void *, | |
726 | size_t, int); | |
7221fe4c GZ |
727 | int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int); |
728 | ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); | |
729 | int __ceph_removexattr(struct dentry *, const char *); | |
de57606c SW |
730 | extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); |
731 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); | |
732 | extern int ceph_removexattr(struct dentry *, const char *); | |
733 | extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); | |
734 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); | |
3ce6cd12 AE |
735 | extern void __init ceph_xattr_init(void); |
736 | extern void ceph_xattr_exit(void); | |
de57606c | 737 | |
7221fe4c | 738 | /* acl.c */ |
7221fe4c GZ |
739 | extern const struct xattr_handler *ceph_xattr_handlers[]; |
740 | ||
741 | #ifdef CONFIG_CEPH_FS_POSIX_ACL | |
742 | ||
743 | struct posix_acl *ceph_get_acl(struct inode *, int); | |
72466d0b | 744 | int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); |
7221fe4c | 745 | int ceph_init_acl(struct dentry *, struct inode *, struct inode *); |
7221fe4c GZ |
746 | void ceph_forget_all_cached_acls(struct inode *inode); |
747 | ||
748 | #else | |
749 | ||
750 | #define ceph_get_acl NULL | |
72466d0b | 751 | #define ceph_set_acl NULL |
7221fe4c GZ |
752 | |
753 | static inline int ceph_init_acl(struct dentry *dentry, struct inode *inode, | |
754 | struct inode *dir) | |
755 | { | |
756 | return 0; | |
757 | } | |
758 | ||
759 | static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) | |
760 | { | |
761 | return 0; | |
762 | } | |
763 | ||
764 | static inline void ceph_forget_all_cached_acls(struct inode *inode) | |
765 | { | |
766 | } | |
767 | ||
768 | #endif | |
769 | ||
de57606c SW |
770 | /* caps.c */ |
771 | extern const char *ceph_cap_string(int c); | |
772 | extern void ceph_handle_caps(struct ceph_mds_session *session, | |
773 | struct ceph_msg *msg); | |
774 | extern int ceph_add_cap(struct inode *inode, | |
775 | struct ceph_mds_session *session, u64 cap_id, | |
776 | int fmode, unsigned issued, unsigned wanted, | |
777 | unsigned cap, unsigned seq, u64 realmino, int flags, | |
778 | struct ceph_cap_reservation *caps_reservation); | |
a096b09a | 779 | extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); |
37151668 YS |
780 | extern void ceph_put_cap(struct ceph_mds_client *mdsc, |
781 | struct ceph_cap *cap); | |
9215aeea | 782 | extern int ceph_is_any_caps(struct inode *inode); |
de57606c | 783 | |
d40ee0dc YZ |
784 | extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino, |
785 | u64 cap_id, u32 migrate_seq, u32 issue_seq); | |
de57606c | 786 | extern void ceph_queue_caps_release(struct inode *inode); |
f1a3d572 | 787 | extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); |
02c24a82 JB |
788 | extern int ceph_fsync(struct file *file, loff_t start, loff_t end, |
789 | int datasync); | |
de57606c SW |
790 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, |
791 | struct ceph_mds_session *session); | |
2bc50259 GF |
792 | extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, |
793 | int mds); | |
de57606c SW |
794 | extern int ceph_get_cap_mds(struct inode *inode); |
795 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); | |
796 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); | |
797 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, | |
798 | struct ceph_snap_context *snapc); | |
799 | extern void __ceph_flush_snaps(struct ceph_inode_info *ci, | |
e835124c SW |
800 | struct ceph_mds_session **psession, |
801 | int again); | |
de57606c SW |
802 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, |
803 | struct ceph_mds_session *session); | |
afcdaea3 SW |
804 | extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); |
805 | extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); | |
de57606c SW |
806 | |
807 | extern int ceph_encode_inode_release(void **p, struct inode *inode, | |
808 | int mds, int drop, int unless, int force); | |
809 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, | |
810 | int mds, int drop, int unless); | |
811 | ||
812 | extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, | |
813 | int *got, loff_t endoff); | |
814 | ||
815 | /* for counting open files by mode */ | |
816 | static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) | |
817 | { | |
818 | ci->i_nr_by_mode[mode]++; | |
819 | } | |
820 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); | |
821 | ||
822 | /* addr.c */ | |
823 | extern const struct address_space_operations ceph_aops; | |
824 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); | |
825 | ||
826 | /* file.c */ | |
827 | extern const struct file_operations ceph_file_fops; | |
828 | extern const struct address_space_operations ceph_aops; | |
9e0eb85d | 829 | |
de57606c | 830 | extern int ceph_open(struct inode *inode, struct file *file); |
5ef50c3b SW |
831 | extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, |
832 | struct file *file, unsigned flags, umode_t mode, | |
833 | int *opened); | |
de57606c | 834 | extern int ceph_release(struct inode *inode, struct file *filp); |
de57606c SW |
835 | |
836 | /* dir.c */ | |
837 | extern const struct file_operations ceph_dir_fops; | |
838 | extern const struct inode_operations ceph_dir_iops; | |
52dfb8ac | 839 | extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, |
de57606c SW |
840 | ceph_snapdir_dentry_ops; |
841 | ||
842 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); | |
468640e3 SW |
843 | extern int ceph_handle_snapdir(struct ceph_mds_request *req, |
844 | struct dentry *dentry, int err); | |
de57606c SW |
845 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, |
846 | struct dentry *dentry, int err); | |
847 | ||
848 | extern void ceph_dentry_lru_add(struct dentry *dn); | |
849 | extern void ceph_dentry_lru_touch(struct dentry *dn); | |
850 | extern void ceph_dentry_lru_del(struct dentry *dn); | |
81a6cf2d | 851 | extern void ceph_invalidate_dentry_lease(struct dentry *dentry); |
e5f86dc3 | 852 | extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); |
5f21c96d | 853 | extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry); |
de57606c SW |
854 | |
855 | /* | |
856 | * our d_ops vary depending on whether the inode is live, | |
857 | * snapshotted (read-only), or a virtual ".snap" directory. | |
858 | */ | |
859 | int ceph_init_dentry(struct dentry *dentry); | |
860 | ||
861 | ||
862 | /* ioctl.c */ | |
863 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | |
864 | ||
865 | /* export.c */ | |
866 | extern const struct export_operations ceph_export_ops; | |
867 | ||
40819f6f GF |
868 | /* locks.c */ |
869 | extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); | |
870 | extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); | |
871 | extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); | |
39be95e9 JS |
872 | extern int ceph_encode_locks_to_buffer(struct inode *inode, |
873 | struct ceph_filelock *flocks, | |
874 | int num_fcntl_locks, | |
875 | int num_flock_locks); | |
876 | extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, | |
877 | struct ceph_pagelist *pagelist, | |
878 | int num_fcntl_locks, int num_flock_locks); | |
40819f6f GF |
879 | extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); |
880 | ||
3d14c5d2 YS |
881 | /* debugfs.c */ |
882 | extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); | |
883 | extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); | |
884 | ||
de57606c | 885 | #endif /* _FS_CEPH_SUPER_H */ |