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