1 #ifndef _FS_CEPH_SUPER_H 2 #define _FS_CEPH_SUPER_H 3 4 #include "ceph_debug.h" 5 6 #include <asm/unaligned.h> 7 #include <linux/backing-dev.h> 8 #include <linux/completion.h> 9 #include <linux/exportfs.h> 10 #include <linux/fs.h> 11 #include <linux/mempool.h> 12 #include <linux/pagemap.h> 13 #include <linux/wait.h> 14 #include <linux/writeback.h> 15 #include <linux/slab.h> 16 17 #include "types.h" 18 #include "messenger.h" 19 #include "msgpool.h" 20 #include "mon_client.h" 21 #include "mds_client.h" 22 #include "osd_client.h" 23 #include "ceph_fs.h" 24 25 /* f_type in struct statfs */ 26 #define CEPH_SUPER_MAGIC 0x00c36400 27 28 /* large granularity for statfs utilization stats to facilitate 29 * large volume sizes on 32-bit machines. */ 30 #define CEPH_BLOCK_SHIFT 20 /* 1 MB */ 31 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 32 33 /* 34 * Supported features 35 */ 36 #define CEPH_FEATURE_SUPPORTED CEPH_FEATURE_NOSRCADDR | CEPH_FEATURE_FLOCK 37 #define CEPH_FEATURE_REQUIRED CEPH_FEATURE_NOSRCADDR 38 39 /* 40 * mount options 41 */ 42 #define CEPH_OPT_FSID (1<<0) 43 #define CEPH_OPT_NOSHARE (1<<1) /* don't share client with other sbs */ 44 #define CEPH_OPT_MYIP (1<<2) /* specified my ip */ 45 #define CEPH_OPT_DIRSTAT (1<<4) /* funky `cat dirname` for stats */ 46 #define CEPH_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ 47 #define CEPH_OPT_NOCRC (1<<6) /* no data crc on writes */ 48 #define CEPH_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ 49 50 #define CEPH_OPT_DEFAULT (CEPH_OPT_RBYTES) 51 52 #define ceph_set_opt(client, opt) \ 53 (client)->mount_args->flags |= CEPH_OPT_##opt; 54 #define ceph_test_opt(client, opt) \ 55 (!!((client)->mount_args->flags & CEPH_OPT_##opt)) 56 57 58 struct ceph_mount_args { 59 int sb_flags; 60 int flags; 61 struct ceph_fsid fsid; 62 struct ceph_entity_addr my_addr; 63 int num_mon; 64 struct ceph_entity_addr *mon_addr; 65 int mount_timeout; 66 int osd_idle_ttl; 67 int osd_timeout; 68 int osd_keepalive_timeout; 69 int wsize; 70 int rsize; /* max readahead */ 71 int congestion_kb; /* max writeback in flight */ 72 int caps_wanted_delay_min, caps_wanted_delay_max; 73 int cap_release_safety; 74 int max_readdir; /* max readdir result (entires) */ 75 int max_readdir_bytes; /* max readdir result (bytes) */ 76 char *snapdir_name; /* default ".snap" */ 77 char *name; 78 char *secret; 79 }; 80 81 /* 82 * defaults 83 */ 84 #define CEPH_MOUNT_TIMEOUT_DEFAULT 60 85 #define CEPH_OSD_TIMEOUT_DEFAULT 60 /* seconds */ 86 #define CEPH_OSD_KEEPALIVE_DEFAULT 5 87 #define CEPH_OSD_IDLE_TTL_DEFAULT 60 88 #define CEPH_MOUNT_RSIZE_DEFAULT (512*1024) /* readahead */ 89 #define CEPH_MAX_READDIR_DEFAULT 1024 90 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 91 92 #define CEPH_MSG_MAX_FRONT_LEN (16*1024*1024) 93 #define CEPH_MSG_MAX_DATA_LEN (16*1024*1024) 94 95 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 96 #define CEPH_AUTH_NAME_DEFAULT "guest" 97 /* 98 * Delay telling the MDS we no longer want caps, in case we reopen 99 * the file. Delay a minimum amount of time, even if we send a cap 100 * message for some other reason. Otherwise, take the oppotunity to 101 * update the mds to avoid sending another message later. 102 */ 103 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ 104 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ 105 106 #define CEPH_CAP_RELEASE_SAFETY_DEFAULT (CEPH_CAPS_PER_RELEASE * 4) 107 108 /* mount state */ 109 enum { 110 CEPH_MOUNT_MOUNTING, 111 CEPH_MOUNT_MOUNTED, 112 CEPH_MOUNT_UNMOUNTING, 113 CEPH_MOUNT_UNMOUNTED, 114 CEPH_MOUNT_SHUTDOWN, 115 }; 116 117 /* 118 * subtract jiffies 119 */ 120 static inline unsigned long time_sub(unsigned long a, unsigned long b) 121 { 122 BUG_ON(time_after(b, a)); 123 return (long)a - (long)b; 124 } 125 126 /* 127 * per-filesystem client state 128 * 129 * possibly shared by multiple mount points, if they are 130 * mounting the same ceph filesystem/cluster. 131 */ 132 struct ceph_client { 133 struct ceph_fsid fsid; 134 bool have_fsid; 135 136 struct mutex mount_mutex; /* serialize mount attempts */ 137 struct ceph_mount_args *mount_args; 138 139 struct super_block *sb; 140 141 unsigned long mount_state; 142 wait_queue_head_t auth_wq; 143 144 int auth_err; 145 146 int min_caps; /* min caps i added */ 147 148 struct ceph_messenger *msgr; /* messenger instance */ 149 struct ceph_mon_client monc; 150 struct ceph_mds_client mdsc; 151 struct ceph_osd_client osdc; 152 153 /* writeback */ 154 mempool_t *wb_pagevec_pool; 155 struct workqueue_struct *wb_wq; 156 struct workqueue_struct *pg_inv_wq; 157 struct workqueue_struct *trunc_wq; 158 atomic_long_t writeback_count; 159 160 struct backing_dev_info backing_dev_info; 161 162 #ifdef CONFIG_DEBUG_FS 163 struct dentry *debugfs_monmap; 164 struct dentry *debugfs_mdsmap, *debugfs_osdmap; 165 struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps; 166 struct dentry *debugfs_congestion_kb; 167 struct dentry *debugfs_bdi; 168 #endif 169 }; 170 171 /* 172 * File i/o capability. This tracks shared state with the metadata 173 * server that allows us to cache or writeback attributes or to read 174 * and write data. For any given inode, we should have one or more 175 * capabilities, one issued by each metadata server, and our 176 * cumulative access is the OR of all issued capabilities. 177 * 178 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 179 * session capability lists. 180 */ 181 struct ceph_cap { 182 struct ceph_inode_info *ci; 183 struct rb_node ci_node; /* per-ci cap tree */ 184 struct ceph_mds_session *session; 185 struct list_head session_caps; /* per-session caplist */ 186 int mds; 187 u64 cap_id; /* unique cap id (mds provided) */ 188 int issued; /* latest, from the mds */ 189 int implemented; /* implemented superset of issued (for revocation) */ 190 int mds_wanted; 191 u32 seq, issue_seq, mseq; 192 u32 cap_gen; /* active/stale cycle */ 193 unsigned long last_used; 194 struct list_head caps_item; 195 }; 196 197 #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ 198 #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ 199 #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ 200 201 /* 202 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 203 * we first complete any in-process sync writes and writeback any dirty 204 * data before flushing the snapped state (tracked here) back to the MDS. 205 */ 206 struct ceph_cap_snap { 207 atomic_t nref; 208 struct ceph_inode_info *ci; 209 struct list_head ci_item, flushing_item; 210 211 u64 follows, flush_tid; 212 int issued, dirty; 213 struct ceph_snap_context *context; 214 215 mode_t mode; 216 uid_t uid; 217 gid_t gid; 218 219 void *xattr_blob; 220 int xattr_len; 221 u64 xattr_version; 222 223 u64 size; 224 struct timespec mtime, atime, ctime; 225 u64 time_warp_seq; 226 int writing; /* a sync write is still in progress */ 227 int dirty_pages; /* dirty pages awaiting writeback */ 228 }; 229 230 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 231 { 232 if (atomic_dec_and_test(&capsnap->nref)) 233 kfree(capsnap); 234 } 235 236 /* 237 * The frag tree describes how a directory is fragmented, potentially across 238 * multiple metadata servers. It is also used to indicate points where 239 * metadata authority is delegated, and whether/where metadata is replicated. 240 * 241 * A _leaf_ frag will be present in the i_fragtree IFF there is 242 * delegation info. That is, if mds >= 0 || ndist > 0. 243 */ 244 #define CEPH_MAX_DIRFRAG_REP 4 245 246 struct ceph_inode_frag { 247 struct rb_node node; 248 249 /* fragtree state */ 250 u32 frag; 251 int split_by; /* i.e. 2^(split_by) children */ 252 253 /* delegation and replication info */ 254 int mds; /* -1 if same authority as parent */ 255 int ndist; /* >0 if replicated */ 256 int dist[CEPH_MAX_DIRFRAG_REP]; 257 }; 258 259 /* 260 * We cache inode xattrs as an encoded blob until they are first used, 261 * at which point we parse them into an rbtree. 262 */ 263 struct ceph_inode_xattr { 264 struct rb_node node; 265 266 const char *name; 267 int name_len; 268 const char *val; 269 int val_len; 270 int dirty; 271 272 int should_free_name; 273 int should_free_val; 274 }; 275 276 struct ceph_inode_xattrs_info { 277 /* 278 * (still encoded) xattr blob. we avoid the overhead of parsing 279 * this until someone actually calls getxattr, etc. 280 * 281 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 282 * NULL means we don't know. 283 */ 284 struct ceph_buffer *blob, *prealloc_blob; 285 286 struct rb_root index; 287 bool dirty; 288 int count; 289 int names_size; 290 int vals_size; 291 u64 version, index_version; 292 }; 293 294 /* 295 * Ceph inode. 296 */ 297 #define CEPH_I_COMPLETE 1 /* we have complete directory cached */ 298 #define CEPH_I_NODELAY 4 /* do not delay cap release */ 299 #define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ 300 #define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ 301 302 struct ceph_inode_info { 303 struct ceph_vino i_vino; /* ceph ino + snap */ 304 305 u64 i_version; 306 u32 i_time_warp_seq; 307 308 unsigned i_ceph_flags; 309 unsigned long i_release_count; 310 311 struct ceph_file_layout i_layout; 312 char *i_symlink; 313 314 /* for dirs */ 315 struct timespec i_rctime; 316 u64 i_rbytes, i_rfiles, i_rsubdirs; 317 u64 i_files, i_subdirs; 318 u64 i_max_offset; /* largest readdir offset, set with I_COMPLETE */ 319 320 struct rb_root i_fragtree; 321 struct mutex i_fragtree_mutex; 322 323 struct ceph_inode_xattrs_info i_xattrs; 324 325 /* capabilities. protected _both_ by i_lock and cap->session's 326 * s_mutex. */ 327 struct rb_root i_caps; /* cap list */ 328 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 329 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 330 struct list_head i_dirty_item, i_flushing_item; 331 u64 i_cap_flush_seq; 332 /* we need to track cap writeback on a per-cap-bit basis, to allow 333 * overlapping, pipelined cap flushes to the mds. we can probably 334 * reduce the tid to 8 bits if we're concerned about inode size. */ 335 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; 336 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 337 unsigned long i_hold_caps_min; /* jiffies */ 338 unsigned long i_hold_caps_max; /* jiffies */ 339 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 340 int i_cap_exporting_mds; /* to handle cap migration between */ 341 unsigned i_cap_exporting_mseq; /* mds's. */ 342 unsigned i_cap_exporting_issued; 343 struct ceph_cap_reservation i_cap_migration_resv; 344 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 345 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 */ 346 unsigned i_snap_caps; /* cap bits for snapped files */ 347 348 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ 349 350 u32 i_truncate_seq; /* last truncate to smaller size */ 351 u64 i_truncate_size; /* and the size we last truncated down to */ 352 int i_truncate_pending; /* still need to call vmtruncate */ 353 354 u64 i_max_size; /* max file size authorized by mds */ 355 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 356 u64 i_wanted_max_size; /* offset we'd like to write too */ 357 u64 i_requested_max_size; /* max_size we've requested */ 358 359 /* held references to caps */ 360 int i_pin_ref; 361 int i_rd_ref, i_rdcache_ref, i_wr_ref; 362 int i_wrbuffer_ref, i_wrbuffer_ref_head; 363 u32 i_shared_gen; /* increment each time we get FILE_SHARED */ 364 u32 i_rdcache_gen; /* we increment this each time we get 365 FILE_CACHE. If it's non-zero, we 366 _may_ have cached pages. */ 367 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 368 369 struct list_head i_unsafe_writes; /* uncommitted sync writes */ 370 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 371 spinlock_t i_unsafe_lock; 372 373 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 374 int i_snap_realm_counter; /* snap realm (if caps) */ 375 struct list_head i_snap_realm_item; 376 struct list_head i_snap_flush_item; 377 378 struct work_struct i_wb_work; /* writeback work */ 379 struct work_struct i_pg_inv_work; /* page invalidation work */ 380 381 struct work_struct i_vmtruncate_work; 382 383 struct inode vfs_inode; /* at end */ 384 }; 385 386 static inline struct ceph_inode_info *ceph_inode(struct inode *inode) 387 { 388 return container_of(inode, struct ceph_inode_info, vfs_inode); 389 } 390 391 static inline void ceph_i_clear(struct inode *inode, unsigned mask) 392 { 393 struct ceph_inode_info *ci = ceph_inode(inode); 394 395 spin_lock(&inode->i_lock); 396 ci->i_ceph_flags &= ~mask; 397 spin_unlock(&inode->i_lock); 398 } 399 400 static inline void ceph_i_set(struct inode *inode, unsigned mask) 401 { 402 struct ceph_inode_info *ci = ceph_inode(inode); 403 404 spin_lock(&inode->i_lock); 405 ci->i_ceph_flags |= mask; 406 spin_unlock(&inode->i_lock); 407 } 408 409 static inline bool ceph_i_test(struct inode *inode, unsigned mask) 410 { 411 struct ceph_inode_info *ci = ceph_inode(inode); 412 bool r; 413 414 smp_mb(); 415 r = (ci->i_ceph_flags & mask) == mask; 416 return r; 417 } 418 419 420 /* find a specific frag @f */ 421 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 422 u32 f); 423 424 /* 425 * choose fragment for value @v. copy frag content to pfrag, if leaf 426 * exists 427 */ 428 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 429 struct ceph_inode_frag *pfrag, 430 int *found); 431 432 /* 433 * Ceph dentry state 434 */ 435 struct ceph_dentry_info { 436 struct ceph_mds_session *lease_session; 437 u32 lease_gen, lease_shared_gen; 438 u32 lease_seq; 439 unsigned long lease_renew_after, lease_renew_from; 440 struct list_head lru; 441 struct dentry *dentry; 442 u64 time; 443 u64 offset; 444 }; 445 446 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) 447 { 448 return (struct ceph_dentry_info *)dentry->d_fsdata; 449 } 450 451 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) 452 { 453 return ((loff_t)frag << 32) | (loff_t)off; 454 } 455 456 /* 457 * ino_t is <64 bits on many architectures, blech. 458 * 459 * don't include snap in ino hash, at least for now. 460 */ 461 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) 462 { 463 ino_t ino = (ino_t)vino.ino; /* ^ (vino.snap << 20); */ 464 #if BITS_PER_LONG == 32 465 ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8; 466 if (!ino) 467 ino = 1; 468 #endif 469 return ino; 470 } 471 472 static inline int ceph_set_ino_cb(struct inode *inode, void *data) 473 { 474 ceph_inode(inode)->i_vino = *(struct ceph_vino *)data; 475 inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data); 476 return 0; 477 } 478 479 static inline struct ceph_vino ceph_vino(struct inode *inode) 480 { 481 return ceph_inode(inode)->i_vino; 482 } 483 484 /* for printf-style formatting */ 485 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 486 487 static inline u64 ceph_ino(struct inode *inode) 488 { 489 return ceph_inode(inode)->i_vino.ino; 490 } 491 static inline u64 ceph_snap(struct inode *inode) 492 { 493 return ceph_inode(inode)->i_vino.snap; 494 } 495 496 static inline int ceph_ino_compare(struct inode *inode, void *data) 497 { 498 struct ceph_vino *pvino = (struct ceph_vino *)data; 499 struct ceph_inode_info *ci = ceph_inode(inode); 500 return ci->i_vino.ino == pvino->ino && 501 ci->i_vino.snap == pvino->snap; 502 } 503 504 static inline struct inode *ceph_find_inode(struct super_block *sb, 505 struct ceph_vino vino) 506 { 507 ino_t t = ceph_vino_to_ino(vino); 508 return ilookup5(sb, t, ceph_ino_compare, &vino); 509 } 510 511 512 /* 513 * caps helpers 514 */ 515 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 516 { 517 return !RB_EMPTY_ROOT(&ci->i_caps); 518 } 519 520 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 521 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 522 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 523 struct ceph_cap *cap); 524 525 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 526 { 527 int issued; 528 spin_lock(&ci->vfs_inode.i_lock); 529 issued = __ceph_caps_issued(ci, NULL); 530 spin_unlock(&ci->vfs_inode.i_lock); 531 return issued; 532 } 533 534 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, 535 int touch) 536 { 537 int r; 538 spin_lock(&ci->vfs_inode.i_lock); 539 r = __ceph_caps_issued_mask(ci, mask, touch); 540 spin_unlock(&ci->vfs_inode.i_lock); 541 return r; 542 } 543 544 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 545 { 546 return ci->i_dirty_caps | ci->i_flushing_caps; 547 } 548 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); 549 550 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 551 extern int __ceph_caps_used(struct ceph_inode_info *ci); 552 553 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 554 555 /* 556 * wanted, by virtue of open file modes AND cap refs (buffered/cached data) 557 */ 558 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) 559 { 560 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); 561 if (w & CEPH_CAP_FILE_BUFFER) 562 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ 563 return w; 564 } 565 566 /* what the mds thinks we want */ 567 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); 568 569 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 570 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 571 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); 572 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 573 struct ceph_cap_reservation *ctx, int need); 574 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, 575 struct ceph_cap_reservation *ctx); 576 extern void ceph_reservation_status(struct ceph_client *client, 577 int *total, int *avail, int *used, 578 int *reserved, int *min); 579 580 static inline struct ceph_client *ceph_inode_to_client(struct inode *inode) 581 { 582 return (struct ceph_client *)inode->i_sb->s_fs_info; 583 } 584 585 static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb) 586 { 587 return (struct ceph_client *)sb->s_fs_info; 588 } 589 590 591 /* 592 * we keep buffered readdir results attached to file->private_data 593 */ 594 struct ceph_file_info { 595 int fmode; /* initialized on open */ 596 597 /* readdir: position within the dir */ 598 u32 frag; 599 struct ceph_mds_request *last_readdir; 600 int at_end; 601 602 /* readdir: position within a frag */ 603 unsigned offset; /* offset of last chunk, adjusted for . and .. */ 604 u64 next_offset; /* offset of next chunk (last_name's + 1) */ 605 char *last_name; /* last entry in previous chunk */ 606 struct dentry *dentry; /* next dentry (for dcache readdir) */ 607 unsigned long dir_release_count; 608 609 /* used for -o dirstat read() on directory thing */ 610 char *dir_info; 611 int dir_info_len; 612 }; 613 614 615 616 /* 617 * snapshots 618 */ 619 620 /* 621 * A "snap context" is the set of existing snapshots when we 622 * write data. It is used by the OSD to guide its COW behavior. 623 * 624 * The ceph_snap_context is refcounted, and attached to each dirty 625 * page, indicating which context the dirty data belonged when it was 626 * dirtied. 627 */ 628 struct ceph_snap_context { 629 atomic_t nref; 630 u64 seq; 631 int num_snaps; 632 u64 snaps[]; 633 }; 634 635 static inline struct ceph_snap_context * 636 ceph_get_snap_context(struct ceph_snap_context *sc) 637 { 638 /* 639 printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), 640 atomic_read(&sc->nref)+1); 641 */ 642 if (sc) 643 atomic_inc(&sc->nref); 644 return sc; 645 } 646 647 static inline void ceph_put_snap_context(struct ceph_snap_context *sc) 648 { 649 if (!sc) 650 return; 651 /* 652 printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref), 653 atomic_read(&sc->nref)-1); 654 */ 655 if (atomic_dec_and_test(&sc->nref)) { 656 /*printk(" deleting snap_context %p\n", sc);*/ 657 kfree(sc); 658 } 659 } 660 661 /* 662 * A "snap realm" describes a subset of the file hierarchy sharing 663 * the same set of snapshots that apply to it. The realms themselves 664 * are organized into a hierarchy, such that children inherit (some of) 665 * the snapshots of their parents. 666 * 667 * All inodes within the realm that have capabilities are linked into a 668 * per-realm list. 669 */ 670 struct ceph_snap_realm { 671 u64 ino; 672 atomic_t nref; 673 struct rb_node node; 674 675 u64 created, seq; 676 u64 parent_ino; 677 u64 parent_since; /* snapid when our current parent became so */ 678 679 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 680 int num_prior_parent_snaps; /* had prior to parent_since */ 681 u64 *snaps; /* snaps specific to this realm */ 682 int num_snaps; 683 684 struct ceph_snap_realm *parent; 685 struct list_head children; /* list of child realms */ 686 struct list_head child_item; 687 688 struct list_head empty_item; /* if i have ref==0 */ 689 690 /* the current set of snaps for this realm */ 691 struct ceph_snap_context *cached_context; 692 693 struct list_head inodes_with_caps; 694 spinlock_t inodes_with_caps_lock; 695 }; 696 697 698 699 /* 700 * calculate the number of pages a given length and offset map onto, 701 * if we align the data. 702 */ 703 static inline int calc_pages_for(u64 off, u64 len) 704 { 705 return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) - 706 (off >> PAGE_CACHE_SHIFT); 707 } 708 709 710 711 /* snap.c */ 712 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 713 u64 ino); 714 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 715 struct ceph_snap_realm *realm); 716 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 717 struct ceph_snap_realm *realm); 718 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 719 void *p, void *e, bool deletion); 720 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 721 struct ceph_mds_session *session, 722 struct ceph_msg *msg); 723 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); 724 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 725 struct ceph_cap_snap *capsnap); 726 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 727 728 /* 729 * a cap_snap is "pending" if it is still awaiting an in-progress 730 * sync write (that may/may not still update size, mtime, etc.). 731 */ 732 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 733 { 734 return !list_empty(&ci->i_cap_snaps) && 735 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, 736 ci_item)->writing; 737 } 738 739 740 /* super.c */ 741 extern struct kmem_cache *ceph_inode_cachep; 742 extern struct kmem_cache *ceph_cap_cachep; 743 extern struct kmem_cache *ceph_dentry_cachep; 744 extern struct kmem_cache *ceph_file_cachep; 745 746 extern const char *ceph_msg_type_name(int type); 747 extern int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid); 748 749 /* inode.c */ 750 extern const struct inode_operations ceph_file_iops; 751 752 extern struct inode *ceph_alloc_inode(struct super_block *sb); 753 extern void ceph_destroy_inode(struct inode *inode); 754 755 extern struct inode *ceph_get_inode(struct super_block *sb, 756 struct ceph_vino vino); 757 extern struct inode *ceph_get_snapdir(struct inode *parent); 758 extern int ceph_fill_file_size(struct inode *inode, int issued, 759 u32 truncate_seq, u64 truncate_size, u64 size); 760 extern void ceph_fill_file_time(struct inode *inode, int issued, 761 u64 time_warp_seq, struct timespec *ctime, 762 struct timespec *mtime, struct timespec *atime); 763 extern int ceph_fill_trace(struct super_block *sb, 764 struct ceph_mds_request *req, 765 struct ceph_mds_session *session); 766 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 767 struct ceph_mds_session *session); 768 769 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 770 771 extern int ceph_inode_set_size(struct inode *inode, loff_t size); 772 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 773 extern void ceph_queue_vmtruncate(struct inode *inode); 774 775 extern void ceph_queue_invalidate(struct inode *inode); 776 extern void ceph_queue_writeback(struct inode *inode); 777 778 extern int ceph_do_getattr(struct inode *inode, int mask); 779 extern int ceph_permission(struct inode *inode, int mask); 780 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); 781 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, 782 struct kstat *stat); 783 784 /* xattr.c */ 785 extern int ceph_setxattr(struct dentry *, const char *, const void *, 786 size_t, int); 787 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); 788 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 789 extern int ceph_removexattr(struct dentry *, const char *); 790 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); 791 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 792 793 /* caps.c */ 794 extern const char *ceph_cap_string(int c); 795 extern void ceph_handle_caps(struct ceph_mds_session *session, 796 struct ceph_msg *msg); 797 extern int ceph_add_cap(struct inode *inode, 798 struct ceph_mds_session *session, u64 cap_id, 799 int fmode, unsigned issued, unsigned wanted, 800 unsigned cap, unsigned seq, u64 realmino, int flags, 801 struct ceph_cap_reservation *caps_reservation); 802 extern void __ceph_remove_cap(struct ceph_cap *cap); 803 static inline void ceph_remove_cap(struct ceph_cap *cap) 804 { 805 struct inode *inode = &cap->ci->vfs_inode; 806 spin_lock(&inode->i_lock); 807 __ceph_remove_cap(cap); 808 spin_unlock(&inode->i_lock); 809 } 810 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 811 struct ceph_cap *cap); 812 813 extern void ceph_queue_caps_release(struct inode *inode); 814 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 815 extern int ceph_fsync(struct file *file, int datasync); 816 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 817 struct ceph_mds_session *session); 818 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 819 int mds); 820 extern int ceph_get_cap_mds(struct inode *inode); 821 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 822 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 823 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 824 struct ceph_snap_context *snapc); 825 extern void __ceph_flush_snaps(struct ceph_inode_info *ci, 826 struct ceph_mds_session **psession); 827 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 828 struct ceph_mds_session *session); 829 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 830 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 831 832 extern int ceph_encode_inode_release(void **p, struct inode *inode, 833 int mds, int drop, int unless, int force); 834 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 835 int mds, int drop, int unless); 836 837 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, 838 int *got, loff_t endoff); 839 840 /* for counting open files by mode */ 841 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) 842 { 843 ci->i_nr_by_mode[mode]++; 844 } 845 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); 846 847 /* addr.c */ 848 extern const struct address_space_operations ceph_aops; 849 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 850 851 /* file.c */ 852 extern const struct file_operations ceph_file_fops; 853 extern const struct address_space_operations ceph_aops; 854 extern int ceph_open(struct inode *inode, struct file *file); 855 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry, 856 struct nameidata *nd, int mode, 857 int locked_dir); 858 extern int ceph_release(struct inode *inode, struct file *filp); 859 extern void ceph_release_page_vector(struct page **pages, int num_pages); 860 861 /* dir.c */ 862 extern const struct file_operations ceph_dir_fops; 863 extern const struct inode_operations ceph_dir_iops; 864 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, 865 ceph_snapdir_dentry_ops; 866 867 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 868 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 869 struct dentry *dentry, int err); 870 871 extern void ceph_dentry_lru_add(struct dentry *dn); 872 extern void ceph_dentry_lru_touch(struct dentry *dn); 873 extern void ceph_dentry_lru_del(struct dentry *dn); 874 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 875 876 /* 877 * our d_ops vary depending on whether the inode is live, 878 * snapshotted (read-only), or a virtual ".snap" directory. 879 */ 880 int ceph_init_dentry(struct dentry *dentry); 881 882 883 /* ioctl.c */ 884 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 885 886 /* export.c */ 887 extern const struct export_operations ceph_export_ops; 888 889 /* debugfs.c */ 890 extern int ceph_debugfs_init(void); 891 extern void ceph_debugfs_cleanup(void); 892 extern int ceph_debugfs_client_init(struct ceph_client *client); 893 extern void ceph_debugfs_client_cleanup(struct ceph_client *client); 894 895 /* locks.c */ 896 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 897 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 898 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 899 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p, 900 int p_locks, int f_locks); 901 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); 902 903 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry) 904 { 905 if (dentry && dentry->d_parent) 906 return dentry->d_parent->d_inode; 907 908 return NULL; 909 } 910 911 #endif /* _FS_CEPH_SUPER_H */ 912