]>
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> |
c969d9bf | 16 | #include <linux/posix_acl.h> |
de57606c | 17 | |
3d14c5d2 | 18 | #include <linux/ceph/libceph.h> |
de57606c | 19 | |
99ccbd22 MT |
20 | #ifdef CONFIG_CEPH_FSCACHE |
21 | #include <linux/fscache.h> | |
22 | #endif | |
23 | ||
de57606c SW |
24 | /* f_type in struct statfs */ |
25 | #define CEPH_SUPER_MAGIC 0x00c36400 | |
26 | ||
27 | /* large granularity for statfs utilization stats to facilitate | |
28 | * large volume sizes on 32-bit machines. */ | |
92a49fb0 | 29 | #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ |
de57606c SW |
30 | #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) |
31 | ||
3d14c5d2 YS |
32 | #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ |
33 | #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ | |
34 | #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ | |
ad1fee96 | 35 | #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ |
a40dc6cc | 36 | #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ |
99ccbd22 | 37 | #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ |
6a259382 | 38 | |
3d14c5d2 | 39 | #define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES) |
de57606c | 40 | |
3d14c5d2 YS |
41 | #define ceph_set_mount_opt(fsc, opt) \ |
42 | (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt; | |
43 | #define ceph_test_mount_opt(fsc, opt) \ | |
44 | (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) | |
de57606c | 45 | |
83817e35 SW |
46 | #define CEPH_RSIZE_DEFAULT 0 /* max read size */ |
47 | #define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */ | |
3d14c5d2 YS |
48 | #define CEPH_MAX_READDIR_DEFAULT 1024 |
49 | #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) | |
50 | #define CEPH_SNAPDIRNAME_DEFAULT ".snap" | |
de57606c | 51 | |
3d14c5d2 | 52 | struct ceph_mount_options { |
6e19a16e | 53 | int flags; |
3d14c5d2 YS |
54 | int sb_flags; |
55 | ||
83817e35 SW |
56 | int wsize; /* max write size */ |
57 | int rsize; /* max read size */ | |
58 | int rasize; /* max readahead */ | |
6e19a16e SW |
59 | int congestion_kb; /* max writeback in flight */ |
60 | int caps_wanted_delay_min, caps_wanted_delay_max; | |
61 | int cap_release_safety; | |
23804d91 SW |
62 | int max_readdir; /* max readdir result (entires) */ |
63 | int max_readdir_bytes; /* max readdir result (bytes) */ | |
de57606c | 64 | |
3d14c5d2 YS |
65 | /* |
66 | * everything above this point can be memcmp'd; everything below | |
67 | * is handled in compare_mount_options() | |
68 | */ | |
de57606c | 69 | |
3d14c5d2 YS |
70 | char *snapdir_name; /* default ".snap" */ |
71 | }; | |
de57606c | 72 | |
3d14c5d2 | 73 | struct ceph_fs_client { |
de57606c SW |
74 | struct super_block *sb; |
75 | ||
3d14c5d2 YS |
76 | struct ceph_mount_options *mount_options; |
77 | struct ceph_client *client; | |
de57606c | 78 | |
3d14c5d2 | 79 | unsigned long mount_state; |
85ccce43 SW |
80 | int min_caps; /* min caps i added */ |
81 | ||
3d14c5d2 | 82 | struct ceph_mds_client *mdsc; |
de57606c SW |
83 | |
84 | /* writeback */ | |
85 | mempool_t *wb_pagevec_pool; | |
86 | struct workqueue_struct *wb_wq; | |
87 | struct workqueue_struct *pg_inv_wq; | |
88 | struct workqueue_struct *trunc_wq; | |
2baba250 | 89 | atomic_long_t writeback_count; |
de57606c SW |
90 | |
91 | struct backing_dev_info backing_dev_info; | |
0743304d SW |
92 | |
93 | #ifdef CONFIG_DEBUG_FS | |
3d14c5d2 | 94 | struct dentry *debugfs_dentry_lru, *debugfs_caps; |
2baba250 | 95 | struct dentry *debugfs_congestion_kb; |
06edf046 | 96 | struct dentry *debugfs_bdi; |
3d14c5d2 | 97 | struct dentry *debugfs_mdsc, *debugfs_mdsmap; |
0743304d | 98 | #endif |
99ccbd22 MT |
99 | |
100 | #ifdef CONFIG_CEPH_FSCACHE | |
101 | struct fscache_cookie *fscache; | |
102 | struct workqueue_struct *revalidate_wq; | |
103 | #endif | |
de57606c SW |
104 | }; |
105 | ||
3d14c5d2 | 106 | |
de57606c SW |
107 | /* |
108 | * File i/o capability. This tracks shared state with the metadata | |
109 | * server that allows us to cache or writeback attributes or to read | |
110 | * and write data. For any given inode, we should have one or more | |
111 | * capabilities, one issued by each metadata server, and our | |
112 | * cumulative access is the OR of all issued capabilities. | |
113 | * | |
114 | * Each cap is referenced by the inode's i_caps rbtree and by per-mds | |
115 | * session capability lists. | |
116 | */ | |
117 | struct ceph_cap { | |
118 | struct ceph_inode_info *ci; | |
119 | struct rb_node ci_node; /* per-ci cap tree */ | |
120 | struct ceph_mds_session *session; | |
121 | struct list_head session_caps; /* per-session caplist */ | |
122 | int mds; | |
123 | u64 cap_id; /* unique cap id (mds provided) */ | |
124 | int issued; /* latest, from the mds */ | |
125 | int implemented; /* implemented superset of issued (for revocation) */ | |
126 | int mds_wanted; | |
685f9a5d SW |
127 | u32 seq, issue_seq, mseq; |
128 | u32 cap_gen; /* active/stale cycle */ | |
de57606c SW |
129 | unsigned long last_used; |
130 | struct list_head caps_item; | |
131 | }; | |
132 | ||
133 | #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ | |
134 | #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ | |
135 | #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ | |
136 | ||
137 | /* | |
138 | * Snapped cap state that is pending flush to mds. When a snapshot occurs, | |
139 | * we first complete any in-process sync writes and writeback any dirty | |
140 | * data before flushing the snapped state (tracked here) back to the MDS. | |
141 | */ | |
142 | struct ceph_cap_snap { | |
143 | atomic_t nref; | |
144 | struct ceph_inode_info *ci; | |
145 | struct list_head ci_item, flushing_item; | |
146 | ||
147 | u64 follows, flush_tid; | |
148 | int issued, dirty; | |
149 | struct ceph_snap_context *context; | |
150 | ||
5706b27d | 151 | umode_t mode; |
05cb11c1 EB |
152 | kuid_t uid; |
153 | kgid_t gid; | |
de57606c | 154 | |
4a625be4 | 155 | struct ceph_buffer *xattr_blob; |
de57606c SW |
156 | u64 xattr_version; |
157 | ||
158 | u64 size; | |
159 | struct timespec mtime, atime, ctime; | |
160 | u64 time_warp_seq; | |
161 | int writing; /* a sync write is still in progress */ | |
162 | int dirty_pages; /* dirty pages awaiting writeback */ | |
163 | }; | |
164 | ||
165 | static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) | |
166 | { | |
4a625be4 SW |
167 | if (atomic_dec_and_test(&capsnap->nref)) { |
168 | if (capsnap->xattr_blob) | |
169 | ceph_buffer_put(capsnap->xattr_blob); | |
de57606c | 170 | kfree(capsnap); |
4a625be4 | 171 | } |
de57606c SW |
172 | } |
173 | ||
174 | /* | |
175 | * The frag tree describes how a directory is fragmented, potentially across | |
176 | * multiple metadata servers. It is also used to indicate points where | |
177 | * metadata authority is delegated, and whether/where metadata is replicated. | |
178 | * | |
179 | * A _leaf_ frag will be present in the i_fragtree IFF there is | |
180 | * delegation info. That is, if mds >= 0 || ndist > 0. | |
181 | */ | |
182 | #define CEPH_MAX_DIRFRAG_REP 4 | |
183 | ||
184 | struct ceph_inode_frag { | |
185 | struct rb_node node; | |
186 | ||
187 | /* fragtree state */ | |
188 | u32 frag; | |
189 | int split_by; /* i.e. 2^(split_by) children */ | |
190 | ||
191 | /* delegation and replication info */ | |
192 | int mds; /* -1 if same authority as parent */ | |
193 | int ndist; /* >0 if replicated */ | |
194 | int dist[CEPH_MAX_DIRFRAG_REP]; | |
195 | }; | |
196 | ||
197 | /* | |
198 | * We cache inode xattrs as an encoded blob until they are first used, | |
199 | * at which point we parse them into an rbtree. | |
200 | */ | |
201 | struct ceph_inode_xattr { | |
202 | struct rb_node node; | |
203 | ||
204 | const char *name; | |
205 | int name_len; | |
206 | const char *val; | |
207 | int val_len; | |
208 | int dirty; | |
209 | ||
210 | int should_free_name; | |
211 | int should_free_val; | |
212 | }; | |
213 | ||
3d14c5d2 YS |
214 | /* |
215 | * Ceph dentry state | |
216 | */ | |
217 | struct ceph_dentry_info { | |
218 | struct ceph_mds_session *lease_session; | |
219 | u32 lease_gen, lease_shared_gen; | |
220 | u32 lease_seq; | |
221 | unsigned long lease_renew_after, lease_renew_from; | |
222 | struct list_head lru; | |
223 | struct dentry *dentry; | |
224 | u64 time; | |
225 | u64 offset; | |
226 | }; | |
227 | ||
de57606c SW |
228 | struct ceph_inode_xattrs_info { |
229 | /* | |
230 | * (still encoded) xattr blob. we avoid the overhead of parsing | |
231 | * this until someone actually calls getxattr, etc. | |
232 | * | |
233 | * blob->vec.iov_len == 4 implies there are no xattrs; blob == | |
234 | * NULL means we don't know. | |
235 | */ | |
236 | struct ceph_buffer *blob, *prealloc_blob; | |
237 | ||
238 | struct rb_root index; | |
239 | bool dirty; | |
240 | int count; | |
241 | int names_size; | |
242 | int vals_size; | |
243 | u64 version, index_version; | |
244 | }; | |
245 | ||
246 | /* | |
247 | * Ceph inode. | |
248 | */ | |
de57606c SW |
249 | struct ceph_inode_info { |
250 | struct ceph_vino i_vino; /* ceph ino + snap */ | |
251 | ||
be655596 SW |
252 | spinlock_t i_ceph_lock; |
253 | ||
de57606c SW |
254 | u64 i_version; |
255 | u32 i_time_warp_seq; | |
256 | ||
257 | unsigned i_ceph_flags; | |
2f276c51 YZ |
258 | atomic_t i_release_count; |
259 | atomic_t i_complete_count; | |
de57606c | 260 | |
6c0f3af7 | 261 | struct ceph_dir_layout i_dir_layout; |
de57606c SW |
262 | struct ceph_file_layout i_layout; |
263 | char *i_symlink; | |
264 | ||
265 | /* for dirs */ | |
266 | struct timespec i_rctime; | |
267 | u64 i_rbytes, i_rfiles, i_rsubdirs; | |
268 | u64 i_files, i_subdirs; | |
2f276c51 | 269 | u64 i_max_offset; /* largest readdir offset, set with complete dir */ |
de57606c SW |
270 | |
271 | struct rb_root i_fragtree; | |
272 | struct mutex i_fragtree_mutex; | |
273 | ||
274 | struct ceph_inode_xattrs_info i_xattrs; | |
275 | ||
be655596 | 276 | /* capabilities. protected _both_ by i_ceph_lock and cap->session's |
de57606c SW |
277 | * s_mutex. */ |
278 | struct rb_root i_caps; /* cap list */ | |
279 | struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ | |
280 | unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ | |
281 | struct list_head i_dirty_item, i_flushing_item; | |
282 | u64 i_cap_flush_seq; | |
283 | /* we need to track cap writeback on a per-cap-bit basis, to allow | |
284 | * overlapping, pipelined cap flushes to the mds. we can probably | |
285 | * reduce the tid to 8 bits if we're concerned about inode size. */ | |
286 | u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; | |
287 | wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ | |
288 | unsigned long i_hold_caps_min; /* jiffies */ | |
289 | unsigned long i_hold_caps_max; /* jiffies */ | |
290 | struct list_head i_cap_delay_list; /* for delayed cap release to mds */ | |
de57606c SW |
291 | struct ceph_cap_reservation i_cap_migration_resv; |
292 | struct list_head i_cap_snaps; /* snapped state pending flush to mds */ | |
7d8cb26d SW |
293 | struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or |
294 | dirty|flushing caps */ | |
de57606c | 295 | unsigned i_snap_caps; /* cap bits for snapped files */ |
11df2dfb | 296 | unsigned i_cap_exporting_issued; |
de57606c SW |
297 | |
298 | int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ | |
299 | ||
b0d7c223 | 300 | struct mutex i_truncate_mutex; |
de57606c SW |
301 | u32 i_truncate_seq; /* last truncate to smaller size */ |
302 | u64 i_truncate_size; /* and the size we last truncated down to */ | |
303 | int i_truncate_pending; /* still need to call vmtruncate */ | |
304 | ||
305 | u64 i_max_size; /* max file size authorized by mds */ | |
306 | u64 i_reported_size; /* (max_)size reported to or requested of mds */ | |
307 | u64 i_wanted_max_size; /* offset we'd like to write too */ | |
308 | u64 i_requested_max_size; /* max_size we've requested */ | |
309 | ||
310 | /* held references to caps */ | |
311 | int i_pin_ref; | |
d3d0720d | 312 | int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref; |
de57606c SW |
313 | int i_wrbuffer_ref, i_wrbuffer_ref_head; |
314 | u32 i_shared_gen; /* increment each time we get FILE_SHARED */ | |
cd045cb4 | 315 | u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ |
de57606c SW |
316 | u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ |
317 | ||
318 | struct list_head i_unsafe_writes; /* uncommitted sync writes */ | |
319 | struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ | |
320 | spinlock_t i_unsafe_lock; | |
321 | ||
322 | struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ | |
323 | int i_snap_realm_counter; /* snap realm (if caps) */ | |
324 | struct list_head i_snap_realm_item; | |
325 | struct list_head i_snap_flush_item; | |
326 | ||
327 | struct work_struct i_wb_work; /* writeback work */ | |
328 | struct work_struct i_pg_inv_work; /* page invalidation work */ | |
329 | ||
330 | struct work_struct i_vmtruncate_work; | |
331 | ||
99ccbd22 MT |
332 | #ifdef CONFIG_CEPH_FSCACHE |
333 | struct fscache_cookie *fscache; | |
334 | u32 i_fscache_gen; /* sequence, for delayed fscache validate */ | |
335 | struct work_struct i_revalidate_work; | |
336 | #endif | |
de57606c SW |
337 | struct inode vfs_inode; /* at end */ |
338 | }; | |
339 | ||
340 | static inline struct ceph_inode_info *ceph_inode(struct inode *inode) | |
341 | { | |
fbbccec9 | 342 | return container_of(inode, struct ceph_inode_info, vfs_inode); |
de57606c SW |
343 | } |
344 | ||
ad1fee96 YS |
345 | static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) |
346 | { | |
347 | return (struct ceph_fs_client *)inode->i_sb->s_fs_info; | |
348 | } | |
349 | ||
350 | static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) | |
351 | { | |
352 | return (struct ceph_fs_client *)sb->s_fs_info; | |
353 | } | |
354 | ||
3d14c5d2 YS |
355 | static inline struct ceph_vino ceph_vino(struct inode *inode) |
356 | { | |
357 | return ceph_inode(inode)->i_vino; | |
358 | } | |
359 | ||
360 | /* | |
361 | * ino_t is <64 bits on many architectures, blech. | |
362 | * | |
ad1fee96 YS |
363 | * i_ino (kernel inode) st_ino (userspace) |
364 | * i386 32 32 | |
365 | * x86_64+ino32 64 32 | |
366 | * x86_64 64 64 | |
367 | */ | |
3310f754 | 368 | static inline u32 ceph_ino_to_ino32(__u64 vino) |
ad1fee96 | 369 | { |
3310f754 AO |
370 | u32 ino = vino & 0xffffffff; |
371 | ino ^= vino >> 32; | |
ad1fee96 | 372 | if (!ino) |
a661fc56 | 373 | ino = 2; |
ad1fee96 YS |
374 | return ino; |
375 | } | |
376 | ||
377 | /* | |
378 | * kernel i_ino value | |
3d14c5d2 YS |
379 | */ |
380 | static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) | |
381 | { | |
3d14c5d2 | 382 | #if BITS_PER_LONG == 32 |
3310f754 AO |
383 | return ceph_ino_to_ino32(vino.ino); |
384 | #else | |
385 | return (ino_t)vino.ino; | |
3d14c5d2 | 386 | #endif |
3d14c5d2 YS |
387 | } |
388 | ||
ad1fee96 YS |
389 | /* |
390 | * user-visible ino (stat, filldir) | |
391 | */ | |
392 | #if BITS_PER_LONG == 32 | |
393 | static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) | |
394 | { | |
395 | return ino; | |
396 | } | |
397 | #else | |
398 | static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) | |
399 | { | |
400 | if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) | |
401 | ino = ceph_ino_to_ino32(ino); | |
402 | return ino; | |
403 | } | |
404 | #endif | |
405 | ||
406 | ||
3d14c5d2 YS |
407 | /* for printf-style formatting */ |
408 | #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap | |
409 | ||
410 | static inline u64 ceph_ino(struct inode *inode) | |
411 | { | |
412 | return ceph_inode(inode)->i_vino.ino; | |
413 | } | |
414 | static inline u64 ceph_snap(struct inode *inode) | |
415 | { | |
416 | return ceph_inode(inode)->i_vino.snap; | |
417 | } | |
418 | ||
419 | static inline int ceph_ino_compare(struct inode *inode, void *data) | |
420 | { | |
421 | struct ceph_vino *pvino = (struct ceph_vino *)data; | |
422 | struct ceph_inode_info *ci = ceph_inode(inode); | |
423 | return ci->i_vino.ino == pvino->ino && | |
424 | ci->i_vino.snap == pvino->snap; | |
425 | } | |
426 | ||
427 | static inline struct inode *ceph_find_inode(struct super_block *sb, | |
428 | struct ceph_vino vino) | |
429 | { | |
430 | ino_t t = ceph_vino_to_ino(vino); | |
431 | return ilookup5(sb, t, ceph_ino_compare, &vino); | |
432 | } | |
433 | ||
434 | ||
435 | /* | |
436 | * Ceph inode. | |
437 | */ | |
3d14c5d2 YS |
438 | #define CEPH_I_NODELAY 4 /* do not delay cap release */ |
439 | #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ | |
440 | #define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ | |
441 | ||
2f276c51 YZ |
442 | static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, |
443 | int release_count) | |
de57606c | 444 | { |
2f276c51 | 445 | atomic_set(&ci->i_complete_count, release_count); |
de57606c SW |
446 | } |
447 | ||
2f276c51 | 448 | static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) |
de57606c | 449 | { |
2f276c51 YZ |
450 | atomic_inc(&ci->i_release_count); |
451 | } | |
de57606c | 452 | |
2f276c51 YZ |
453 | static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) |
454 | { | |
455 | return atomic_read(&ci->i_complete_count) == | |
456 | atomic_read(&ci->i_release_count); | |
de57606c SW |
457 | } |
458 | ||
2f276c51 | 459 | static inline void ceph_dir_clear_complete(struct inode *inode) |
de57606c | 460 | { |
2f276c51 YZ |
461 | __ceph_dir_clear_complete(ceph_inode(inode)); |
462 | } | |
de57606c | 463 | |
2f276c51 YZ |
464 | static inline bool ceph_dir_is_complete(struct inode *inode) |
465 | { | |
466 | return __ceph_dir_is_complete(ceph_inode(inode)); | |
de57606c SW |
467 | } |
468 | ||
469 | ||
470 | /* find a specific frag @f */ | |
471 | extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, | |
472 | u32 f); | |
473 | ||
474 | /* | |
475 | * choose fragment for value @v. copy frag content to pfrag, if leaf | |
476 | * exists | |
477 | */ | |
478 | extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, | |
479 | struct ceph_inode_frag *pfrag, | |
480 | int *found); | |
481 | ||
de57606c SW |
482 | static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) |
483 | { | |
484 | return (struct ceph_dentry_info *)dentry->d_fsdata; | |
485 | } | |
486 | ||
487 | static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) | |
488 | { | |
489 | return ((loff_t)frag << 32) | (loff_t)off; | |
490 | } | |
491 | ||
de57606c SW |
492 | /* |
493 | * caps helpers | |
494 | */ | |
495 | static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) | |
496 | { | |
497 | return !RB_EMPTY_ROOT(&ci->i_caps); | |
498 | } | |
499 | ||
500 | extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); | |
501 | extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); | |
502 | extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, | |
503 | struct ceph_cap *cap); | |
504 | ||
505 | static inline int ceph_caps_issued(struct ceph_inode_info *ci) | |
506 | { | |
507 | int issued; | |
be655596 | 508 | spin_lock(&ci->i_ceph_lock); |
de57606c | 509 | issued = __ceph_caps_issued(ci, NULL); |
be655596 | 510 | spin_unlock(&ci->i_ceph_lock); |
de57606c SW |
511 | return issued; |
512 | } | |
513 | ||
514 | static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, | |
515 | int touch) | |
516 | { | |
517 | int r; | |
be655596 | 518 | spin_lock(&ci->i_ceph_lock); |
de57606c | 519 | r = __ceph_caps_issued_mask(ci, mask, touch); |
be655596 | 520 | spin_unlock(&ci->i_ceph_lock); |
de57606c SW |
521 | return r; |
522 | } | |
523 | ||
524 | static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) | |
525 | { | |
526 | return ci->i_dirty_caps | ci->i_flushing_caps; | |
527 | } | |
fca65b4a | 528 | extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); |
de57606c | 529 | |
9563f88c YZ |
530 | extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, |
531 | struct ceph_cap *ocap, int mask); | |
de57606c SW |
532 | extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); |
533 | extern int __ceph_caps_used(struct ceph_inode_info *ci); | |
534 | ||
535 | extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); | |
536 | ||
537 | /* | |
538 | * wanted, by virtue of open file modes AND cap refs (buffered/cached data) | |
539 | */ | |
540 | static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) | |
541 | { | |
542 | int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); | |
543 | if (w & CEPH_CAP_FILE_BUFFER) | |
544 | w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ | |
545 | return w; | |
546 | } | |
547 | ||
548 | /* what the mds thinks we want */ | |
549 | extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); | |
550 | ||
37151668 YS |
551 | extern void ceph_caps_init(struct ceph_mds_client *mdsc); |
552 | extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); | |
553 | extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); | |
93faca6e | 554 | extern void ceph_reserve_caps(struct ceph_mds_client *mdsc, |
37151668 YS |
555 | struct ceph_cap_reservation *ctx, int need); |
556 | extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, | |
557 | struct ceph_cap_reservation *ctx); | |
3d14c5d2 | 558 | extern void ceph_reservation_status(struct ceph_fs_client *client, |
de57606c | 559 | int *total, int *avail, int *used, |
85ccce43 | 560 | int *reserved, int *min); |
de57606c | 561 | |
de57606c | 562 | |
de57606c SW |
563 | |
564 | /* | |
565 | * we keep buffered readdir results attached to file->private_data | |
566 | */ | |
4918b6d1 | 567 | #define CEPH_F_SYNC 1 |
9cfa1098 | 568 | #define CEPH_F_ATEND 2 |
4918b6d1 | 569 | |
de57606c | 570 | struct ceph_file_info { |
252c6728 SW |
571 | short fmode; /* initialized on open */ |
572 | short flags; /* CEPH_F_* */ | |
de57606c SW |
573 | |
574 | /* readdir: position within the dir */ | |
575 | u32 frag; | |
576 | struct ceph_mds_request *last_readdir; | |
de57606c SW |
577 | |
578 | /* readdir: position within a frag */ | |
579 | unsigned offset; /* offset of last chunk, adjusted for . and .. */ | |
f0494206 | 580 | unsigned next_offset; /* offset of next chunk (last_name's + 1) */ |
de57606c SW |
581 | char *last_name; /* last entry in previous chunk */ |
582 | struct dentry *dentry; /* next dentry (for dcache readdir) */ | |
2f276c51 | 583 | int dir_release_count; |
de57606c SW |
584 | |
585 | /* used for -o dirstat read() on directory thing */ | |
586 | char *dir_info; | |
587 | int dir_info_len; | |
588 | }; | |
589 | ||
590 | ||
591 | ||
de57606c SW |
592 | /* |
593 | * A "snap realm" describes a subset of the file hierarchy sharing | |
594 | * the same set of snapshots that apply to it. The realms themselves | |
595 | * are organized into a hierarchy, such that children inherit (some of) | |
596 | * the snapshots of their parents. | |
597 | * | |
598 | * All inodes within the realm that have capabilities are linked into a | |
599 | * per-realm list. | |
600 | */ | |
601 | struct ceph_snap_realm { | |
602 | u64 ino; | |
603 | atomic_t nref; | |
a105f00c SW |
604 | struct rb_node node; |
605 | ||
de57606c SW |
606 | u64 created, seq; |
607 | u64 parent_ino; | |
608 | u64 parent_since; /* snapid when our current parent became so */ | |
609 | ||
610 | u64 *prior_parent_snaps; /* snaps inherited from any parents we */ | |
aa711ee3 | 611 | u32 num_prior_parent_snaps; /* had prior to parent_since */ |
de57606c | 612 | u64 *snaps; /* snaps specific to this realm */ |
aa711ee3 | 613 | u32 num_snaps; |
de57606c SW |
614 | |
615 | struct ceph_snap_realm *parent; | |
616 | struct list_head children; /* list of child realms */ | |
617 | struct list_head child_item; | |
618 | ||
619 | struct list_head empty_item; /* if i have ref==0 */ | |
620 | ||
ae00d4f3 SW |
621 | struct list_head dirty_item; /* if realm needs new context */ |
622 | ||
de57606c SW |
623 | /* the current set of snaps for this realm */ |
624 | struct ceph_snap_context *cached_context; | |
625 | ||
626 | struct list_head inodes_with_caps; | |
627 | spinlock_t inodes_with_caps_lock; | |
628 | }; | |
629 | ||
3d14c5d2 | 630 | static inline int default_congestion_kb(void) |
de57606c | 631 | { |
3d14c5d2 YS |
632 | int congestion_kb; |
633 | ||
634 | /* | |
635 | * Copied from NFS | |
636 | * | |
637 | * congestion size, scale with available memory. | |
638 | * | |
639 | * 64MB: 8192k | |
640 | * 128MB: 11585k | |
641 | * 256MB: 16384k | |
642 | * 512MB: 23170k | |
643 | * 1GB: 32768k | |
644 | * 2GB: 46340k | |
645 | * 4GB: 65536k | |
646 | * 8GB: 92681k | |
647 | * 16GB: 131072k | |
648 | * | |
649 | * This allows larger machines to have larger/more transfers. | |
650 | * Limit the default to 256M | |
651 | */ | |
652 | congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); | |
653 | if (congestion_kb > 256*1024) | |
654 | congestion_kb = 256*1024; | |
655 | ||
656 | return congestion_kb; | |
de57606c SW |
657 | } |
658 | ||
659 | ||
660 | ||
661 | /* snap.c */ | |
662 | struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, | |
663 | u64 ino); | |
664 | extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, | |
665 | struct ceph_snap_realm *realm); | |
666 | extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, | |
667 | struct ceph_snap_realm *realm); | |
668 | extern int ceph_update_snap_trace(struct ceph_mds_client *m, | |
669 | void *p, void *e, bool deletion); | |
670 | extern void ceph_handle_snap(struct ceph_mds_client *mdsc, | |
2600d2dd | 671 | struct ceph_mds_session *session, |
de57606c | 672 | struct ceph_msg *msg); |
fc837c8f | 673 | extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); |
de57606c SW |
674 | extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, |
675 | struct ceph_cap_snap *capsnap); | |
676 | extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); | |
677 | ||
678 | /* | |
679 | * a cap_snap is "pending" if it is still awaiting an in-progress | |
680 | * sync write (that may/may not still update size, mtime, etc.). | |
681 | */ | |
682 | static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) | |
683 | { | |
684 | return !list_empty(&ci->i_cap_snaps) && | |
685 | list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, | |
686 | ci_item)->writing; | |
687 | } | |
688 | ||
de57606c SW |
689 | /* inode.c */ |
690 | extern const struct inode_operations ceph_file_iops; | |
691 | ||
692 | extern struct inode *ceph_alloc_inode(struct super_block *sb); | |
693 | extern void ceph_destroy_inode(struct inode *inode); | |
9f12bd11 | 694 | extern int ceph_drop_inode(struct inode *inode); |
de57606c SW |
695 | |
696 | extern struct inode *ceph_get_inode(struct super_block *sb, | |
697 | struct ceph_vino vino); | |
698 | extern struct inode *ceph_get_snapdir(struct inode *parent); | |
699 | extern int ceph_fill_file_size(struct inode *inode, int issued, | |
700 | u32 truncate_seq, u64 truncate_size, u64 size); | |
701 | extern void ceph_fill_file_time(struct inode *inode, int issued, | |
702 | u64 time_warp_seq, struct timespec *ctime, | |
703 | struct timespec *mtime, struct timespec *atime); | |
704 | extern int ceph_fill_trace(struct super_block *sb, | |
705 | struct ceph_mds_request *req, | |
706 | struct ceph_mds_session *session); | |
707 | extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, | |
708 | struct ceph_mds_session *session); | |
709 | ||
710 | extern int ceph_inode_holds_cap(struct inode *inode, int mask); | |
711 | ||
712 | extern int ceph_inode_set_size(struct inode *inode, loff_t size); | |
b415bf4f | 713 | extern void __ceph_do_pending_vmtruncate(struct inode *inode); |
3c6f6b79 SW |
714 | extern void ceph_queue_vmtruncate(struct inode *inode); |
715 | ||
716 | extern void ceph_queue_invalidate(struct inode *inode); | |
717 | extern void ceph_queue_writeback(struct inode *inode); | |
de57606c SW |
718 | |
719 | extern int ceph_do_getattr(struct inode *inode, int mask); | |
10556cb2 | 720 | extern int ceph_permission(struct inode *inode, int mask); |
de57606c SW |
721 | extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); |
722 | extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, | |
723 | struct kstat *stat); | |
724 | ||
725 | /* xattr.c */ | |
726 | extern int ceph_setxattr(struct dentry *, const char *, const void *, | |
727 | size_t, int); | |
7221fe4c GZ |
728 | int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int); |
729 | ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); | |
730 | int __ceph_removexattr(struct dentry *, const char *); | |
de57606c SW |
731 | extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); |
732 | extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); | |
733 | extern int ceph_removexattr(struct dentry *, const char *); | |
734 | extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); | |
735 | extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); | |
3ce6cd12 AE |
736 | extern void __init ceph_xattr_init(void); |
737 | extern void ceph_xattr_exit(void); | |
de57606c | 738 | |
7221fe4c | 739 | /* acl.c */ |
7221fe4c GZ |
740 | extern const struct xattr_handler *ceph_xattr_handlers[]; |
741 | ||
742 | #ifdef CONFIG_CEPH_FS_POSIX_ACL | |
743 | ||
744 | struct posix_acl *ceph_get_acl(struct inode *, int); | |
72466d0b | 745 | int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); |
7221fe4c | 746 | int ceph_init_acl(struct dentry *, struct inode *, struct inode *); |
c969d9bf GZ |
747 | |
748 | static inline void ceph_forget_all_cached_acls(struct inode *inode) | |
749 | { | |
750 | forget_all_cached_acls(inode); | |
751 | } | |
7221fe4c GZ |
752 | |
753 | #else | |
754 | ||
755 | #define ceph_get_acl NULL | |
72466d0b | 756 | #define ceph_set_acl NULL |
7221fe4c GZ |
757 | |
758 | static inline int ceph_init_acl(struct dentry *dentry, struct inode *inode, | |
759 | struct inode *dir) | |
760 | { | |
761 | return 0; | |
762 | } | |
763 | ||
764 | static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) | |
765 | { | |
766 | return 0; | |
767 | } | |
768 | ||
769 | static inline void ceph_forget_all_cached_acls(struct inode *inode) | |
770 | { | |
771 | } | |
772 | ||
773 | #endif | |
774 | ||
de57606c SW |
775 | /* caps.c */ |
776 | extern const char *ceph_cap_string(int c); | |
777 | extern void ceph_handle_caps(struct ceph_mds_session *session, | |
778 | struct ceph_msg *msg); | |
779 | extern int ceph_add_cap(struct inode *inode, | |
780 | struct ceph_mds_session *session, u64 cap_id, | |
781 | int fmode, unsigned issued, unsigned wanted, | |
782 | unsigned cap, unsigned seq, u64 realmino, int flags, | |
783 | struct ceph_cap_reservation *caps_reservation); | |
a096b09a | 784 | extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); |
37151668 YS |
785 | extern void ceph_put_cap(struct ceph_mds_client *mdsc, |
786 | struct ceph_cap *cap); | |
9215aeea | 787 | extern int ceph_is_any_caps(struct inode *inode); |
de57606c | 788 | |
d40ee0dc YZ |
789 | extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino, |
790 | u64 cap_id, u32 migrate_seq, u32 issue_seq); | |
de57606c | 791 | extern void ceph_queue_caps_release(struct inode *inode); |
f1a3d572 | 792 | extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); |
02c24a82 JB |
793 | extern int ceph_fsync(struct file *file, loff_t start, loff_t end, |
794 | int datasync); | |
de57606c SW |
795 | extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, |
796 | struct ceph_mds_session *session); | |
2bc50259 GF |
797 | extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, |
798 | int mds); | |
de57606c SW |
799 | extern int ceph_get_cap_mds(struct inode *inode); |
800 | extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); | |
801 | extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); | |
802 | extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, | |
803 | struct ceph_snap_context *snapc); | |
804 | extern void __ceph_flush_snaps(struct ceph_inode_info *ci, | |
e835124c SW |
805 | struct ceph_mds_session **psession, |
806 | int again); | |
de57606c SW |
807 | extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, |
808 | struct ceph_mds_session *session); | |
afcdaea3 SW |
809 | extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); |
810 | extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); | |
de57606c SW |
811 | |
812 | extern int ceph_encode_inode_release(void **p, struct inode *inode, | |
813 | int mds, int drop, int unless, int force); | |
814 | extern int ceph_encode_dentry_release(void **p, struct dentry *dn, | |
815 | int mds, int drop, int unless); | |
816 | ||
817 | extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, | |
818 | int *got, loff_t endoff); | |
819 | ||
820 | /* for counting open files by mode */ | |
821 | static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) | |
822 | { | |
823 | ci->i_nr_by_mode[mode]++; | |
824 | } | |
825 | extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); | |
826 | ||
827 | /* addr.c */ | |
828 | extern const struct address_space_operations ceph_aops; | |
829 | extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); | |
830 | ||
831 | /* file.c */ | |
832 | extern const struct file_operations ceph_file_fops; | |
833 | extern const struct address_space_operations ceph_aops; | |
9e0eb85d | 834 | |
de57606c | 835 | extern int ceph_open(struct inode *inode, struct file *file); |
5ef50c3b SW |
836 | extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, |
837 | struct file *file, unsigned flags, umode_t mode, | |
838 | int *opened); | |
de57606c | 839 | extern int ceph_release(struct inode *inode, struct file *filp); |
de57606c SW |
840 | |
841 | /* dir.c */ | |
842 | extern const struct file_operations ceph_dir_fops; | |
843 | extern const struct inode_operations ceph_dir_iops; | |
52dfb8ac | 844 | extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, |
de57606c SW |
845 | ceph_snapdir_dentry_ops; |
846 | ||
847 | extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); | |
468640e3 SW |
848 | extern int ceph_handle_snapdir(struct ceph_mds_request *req, |
849 | struct dentry *dentry, int err); | |
de57606c SW |
850 | extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, |
851 | struct dentry *dentry, int err); | |
852 | ||
853 | extern void ceph_dentry_lru_add(struct dentry *dn); | |
854 | extern void ceph_dentry_lru_touch(struct dentry *dn); | |
855 | extern void ceph_dentry_lru_del(struct dentry *dn); | |
81a6cf2d | 856 | extern void ceph_invalidate_dentry_lease(struct dentry *dentry); |
e5f86dc3 | 857 | extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); |
5f21c96d | 858 | extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry); |
de57606c SW |
859 | |
860 | /* | |
861 | * our d_ops vary depending on whether the inode is live, | |
862 | * snapshotted (read-only), or a virtual ".snap" directory. | |
863 | */ | |
864 | int ceph_init_dentry(struct dentry *dentry); | |
865 | ||
866 | ||
867 | /* ioctl.c */ | |
868 | extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); | |
869 | ||
870 | /* export.c */ | |
871 | extern const struct export_operations ceph_export_ops; | |
872 | ||
40819f6f | 873 | /* locks.c */ |
eb13e832 | 874 | extern __init void ceph_flock_init(void); |
40819f6f GF |
875 | extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); |
876 | extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); | |
877 | extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); | |
39be95e9 JS |
878 | extern int ceph_encode_locks_to_buffer(struct inode *inode, |
879 | struct ceph_filelock *flocks, | |
880 | int num_fcntl_locks, | |
881 | int num_flock_locks); | |
882 | extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, | |
883 | struct ceph_pagelist *pagelist, | |
884 | int num_fcntl_locks, int num_flock_locks); | |
40819f6f GF |
885 | extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); |
886 | ||
3d14c5d2 YS |
887 | /* debugfs.c */ |
888 | extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); | |
889 | extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); | |
890 | ||
de57606c | 891 | #endif /* _FS_CEPH_SUPER_H */ |