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 #include <linux/posix_acl.h> 17 18 #include <linux/ceph/libceph.h> 19 20 #ifdef CONFIG_CEPH_FSCACHE 21 #include <linux/fscache.h> 22 #endif 23 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. */ 29 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ 30 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 31 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 */ 35 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ 36 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ 37 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ 38 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */ 39 40 #define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES | \ 41 CEPH_MOUNT_OPT_DCACHE) 42 43 #define ceph_set_mount_opt(fsc, opt) \ 44 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt; 45 #define ceph_test_mount_opt(fsc, opt) \ 46 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 47 48 #define CEPH_RSIZE_DEFAULT 0 /* max read size */ 49 #define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */ 50 #define CEPH_MAX_READDIR_DEFAULT 1024 51 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 52 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 53 54 struct ceph_mount_options { 55 int flags; 56 int sb_flags; 57 58 int wsize; /* max write size */ 59 int rsize; /* max read size */ 60 int rasize; /* max readahead */ 61 int congestion_kb; /* max writeback in flight */ 62 int caps_wanted_delay_min, caps_wanted_delay_max; 63 int cap_release_safety; 64 int max_readdir; /* max readdir result (entires) */ 65 int max_readdir_bytes; /* max readdir result (bytes) */ 66 67 /* 68 * everything above this point can be memcmp'd; everything below 69 * is handled in compare_mount_options() 70 */ 71 72 char *snapdir_name; /* default ".snap" */ 73 }; 74 75 struct ceph_fs_client { 76 struct super_block *sb; 77 78 struct ceph_mount_options *mount_options; 79 struct ceph_client *client; 80 81 unsigned long mount_state; 82 int min_caps; /* min caps i added */ 83 84 struct ceph_mds_client *mdsc; 85 86 /* writeback */ 87 mempool_t *wb_pagevec_pool; 88 struct workqueue_struct *wb_wq; 89 struct workqueue_struct *pg_inv_wq; 90 struct workqueue_struct *trunc_wq; 91 atomic_long_t writeback_count; 92 93 struct backing_dev_info backing_dev_info; 94 95 #ifdef CONFIG_DEBUG_FS 96 struct dentry *debugfs_dentry_lru, *debugfs_caps; 97 struct dentry *debugfs_congestion_kb; 98 struct dentry *debugfs_bdi; 99 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 100 struct dentry *debugfs_mds_sessions; 101 #endif 102 103 #ifdef CONFIG_CEPH_FSCACHE 104 struct fscache_cookie *fscache; 105 struct workqueue_struct *revalidate_wq; 106 #endif 107 }; 108 109 110 /* 111 * File i/o capability. This tracks shared state with the metadata 112 * server that allows us to cache or writeback attributes or to read 113 * and write data. For any given inode, we should have one or more 114 * capabilities, one issued by each metadata server, and our 115 * cumulative access is the OR of all issued capabilities. 116 * 117 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 118 * session capability lists. 119 */ 120 struct ceph_cap { 121 struct ceph_inode_info *ci; 122 struct rb_node ci_node; /* per-ci cap tree */ 123 struct ceph_mds_session *session; 124 struct list_head session_caps; /* per-session caplist */ 125 int mds; 126 u64 cap_id; /* unique cap id (mds provided) */ 127 int issued; /* latest, from the mds */ 128 int implemented; /* implemented superset of issued (for revocation) */ 129 int mds_wanted; 130 u32 seq, issue_seq, mseq; 131 u32 cap_gen; /* active/stale cycle */ 132 unsigned long last_used; 133 struct list_head caps_item; 134 }; 135 136 #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ 137 #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ 138 #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ 139 140 /* 141 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 142 * we first complete any in-process sync writes and writeback any dirty 143 * data before flushing the snapped state (tracked here) back to the MDS. 144 */ 145 struct ceph_cap_snap { 146 atomic_t nref; 147 struct ceph_inode_info *ci; 148 struct list_head ci_item, flushing_item; 149 150 u64 follows, flush_tid; 151 int issued, dirty; 152 struct ceph_snap_context *context; 153 154 umode_t mode; 155 kuid_t uid; 156 kgid_t gid; 157 158 struct ceph_buffer *xattr_blob; 159 u64 xattr_version; 160 161 u64 size; 162 struct timespec mtime, atime, ctime; 163 u64 time_warp_seq; 164 int writing; /* a sync write is still in progress */ 165 int dirty_pages; /* dirty pages awaiting writeback */ 166 bool inline_data; 167 }; 168 169 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 170 { 171 if (atomic_dec_and_test(&capsnap->nref)) { 172 if (capsnap->xattr_blob) 173 ceph_buffer_put(capsnap->xattr_blob); 174 kfree(capsnap); 175 } 176 } 177 178 /* 179 * The frag tree describes how a directory is fragmented, potentially across 180 * multiple metadata servers. It is also used to indicate points where 181 * metadata authority is delegated, and whether/where metadata is replicated. 182 * 183 * A _leaf_ frag will be present in the i_fragtree IFF there is 184 * delegation info. That is, if mds >= 0 || ndist > 0. 185 */ 186 #define CEPH_MAX_DIRFRAG_REP 4 187 188 struct ceph_inode_frag { 189 struct rb_node node; 190 191 /* fragtree state */ 192 u32 frag; 193 int split_by; /* i.e. 2^(split_by) children */ 194 195 /* delegation and replication info */ 196 int mds; /* -1 if same authority as parent */ 197 int ndist; /* >0 if replicated */ 198 int dist[CEPH_MAX_DIRFRAG_REP]; 199 }; 200 201 /* 202 * We cache inode xattrs as an encoded blob until they are first used, 203 * at which point we parse them into an rbtree. 204 */ 205 struct ceph_inode_xattr { 206 struct rb_node node; 207 208 const char *name; 209 int name_len; 210 const char *val; 211 int val_len; 212 int dirty; 213 214 int should_free_name; 215 int should_free_val; 216 }; 217 218 /* 219 * Ceph dentry state 220 */ 221 struct ceph_dentry_info { 222 struct ceph_mds_session *lease_session; 223 u32 lease_gen, lease_shared_gen; 224 u32 lease_seq; 225 unsigned long lease_renew_after, lease_renew_from; 226 struct list_head lru; 227 struct dentry *dentry; 228 u64 time; 229 u64 offset; 230 }; 231 232 struct ceph_inode_xattrs_info { 233 /* 234 * (still encoded) xattr blob. we avoid the overhead of parsing 235 * this until someone actually calls getxattr, etc. 236 * 237 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 238 * NULL means we don't know. 239 */ 240 struct ceph_buffer *blob, *prealloc_blob; 241 242 struct rb_root index; 243 bool dirty; 244 int count; 245 int names_size; 246 int vals_size; 247 u64 version, index_version; 248 }; 249 250 /* 251 * Ceph inode. 252 */ 253 struct ceph_inode_info { 254 struct ceph_vino i_vino; /* ceph ino + snap */ 255 256 spinlock_t i_ceph_lock; 257 258 u64 i_version; 259 u64 i_inline_version; 260 u32 i_time_warp_seq; 261 262 unsigned i_ceph_flags; 263 int i_ordered_count; 264 atomic_t i_release_count; 265 atomic_t i_complete_count; 266 267 struct ceph_dir_layout i_dir_layout; 268 struct ceph_file_layout i_layout; 269 char *i_symlink; 270 271 /* for dirs */ 272 struct timespec i_rctime; 273 u64 i_rbytes, i_rfiles, i_rsubdirs; 274 u64 i_files, i_subdirs; 275 276 struct rb_root i_fragtree; 277 struct mutex i_fragtree_mutex; 278 279 struct ceph_inode_xattrs_info i_xattrs; 280 281 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 282 * s_mutex. */ 283 struct rb_root i_caps; /* cap list */ 284 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 285 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 286 struct list_head i_dirty_item, i_flushing_item; 287 u64 i_cap_flush_seq; 288 /* we need to track cap writeback on a per-cap-bit basis, to allow 289 * overlapping, pipelined cap flushes to the mds. we can probably 290 * reduce the tid to 8 bits if we're concerned about inode size. */ 291 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; 292 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 293 unsigned long i_hold_caps_min; /* jiffies */ 294 unsigned long i_hold_caps_max; /* jiffies */ 295 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 296 struct ceph_cap_reservation i_cap_migration_resv; 297 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 298 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 299 dirty|flushing caps */ 300 unsigned i_snap_caps; /* cap bits for snapped files */ 301 302 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ 303 304 struct mutex i_truncate_mutex; 305 u32 i_truncate_seq; /* last truncate to smaller size */ 306 u64 i_truncate_size; /* and the size we last truncated down to */ 307 int i_truncate_pending; /* still need to call vmtruncate */ 308 309 u64 i_max_size; /* max file size authorized by mds */ 310 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 311 u64 i_wanted_max_size; /* offset we'd like to write too */ 312 u64 i_requested_max_size; /* max_size we've requested */ 313 314 /* held references to caps */ 315 int i_pin_ref; 316 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref; 317 int i_wrbuffer_ref, i_wrbuffer_ref_head; 318 u32 i_shared_gen; /* increment each time we get FILE_SHARED */ 319 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 320 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 321 322 struct list_head i_unsafe_writes; /* uncommitted sync writes */ 323 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 324 spinlock_t i_unsafe_lock; 325 326 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 327 int i_snap_realm_counter; /* snap realm (if caps) */ 328 struct list_head i_snap_realm_item; 329 struct list_head i_snap_flush_item; 330 331 struct work_struct i_wb_work; /* writeback work */ 332 struct work_struct i_pg_inv_work; /* page invalidation work */ 333 334 struct work_struct i_vmtruncate_work; 335 336 #ifdef CONFIG_CEPH_FSCACHE 337 struct fscache_cookie *fscache; 338 u32 i_fscache_gen; /* sequence, for delayed fscache validate */ 339 struct work_struct i_revalidate_work; 340 #endif 341 struct inode vfs_inode; /* at end */ 342 }; 343 344 static inline struct ceph_inode_info *ceph_inode(struct inode *inode) 345 { 346 return container_of(inode, struct ceph_inode_info, vfs_inode); 347 } 348 349 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) 350 { 351 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 352 } 353 354 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) 355 { 356 return (struct ceph_fs_client *)sb->s_fs_info; 357 } 358 359 static inline struct ceph_vino ceph_vino(struct inode *inode) 360 { 361 return ceph_inode(inode)->i_vino; 362 } 363 364 /* 365 * ino_t is <64 bits on many architectures, blech. 366 * 367 * i_ino (kernel inode) st_ino (userspace) 368 * i386 32 32 369 * x86_64+ino32 64 32 370 * x86_64 64 64 371 */ 372 static inline u32 ceph_ino_to_ino32(__u64 vino) 373 { 374 u32 ino = vino & 0xffffffff; 375 ino ^= vino >> 32; 376 if (!ino) 377 ino = 2; 378 return ino; 379 } 380 381 /* 382 * kernel i_ino value 383 */ 384 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) 385 { 386 #if BITS_PER_LONG == 32 387 return ceph_ino_to_ino32(vino.ino); 388 #else 389 return (ino_t)vino.ino; 390 #endif 391 } 392 393 /* 394 * user-visible ino (stat, filldir) 395 */ 396 #if BITS_PER_LONG == 32 397 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 398 { 399 return ino; 400 } 401 #else 402 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 403 { 404 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) 405 ino = ceph_ino_to_ino32(ino); 406 return ino; 407 } 408 #endif 409 410 411 /* for printf-style formatting */ 412 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 413 414 static inline u64 ceph_ino(struct inode *inode) 415 { 416 return ceph_inode(inode)->i_vino.ino; 417 } 418 static inline u64 ceph_snap(struct inode *inode) 419 { 420 return ceph_inode(inode)->i_vino.snap; 421 } 422 423 static inline int ceph_ino_compare(struct inode *inode, void *data) 424 { 425 struct ceph_vino *pvino = (struct ceph_vino *)data; 426 struct ceph_inode_info *ci = ceph_inode(inode); 427 return ci->i_vino.ino == pvino->ino && 428 ci->i_vino.snap == pvino->snap; 429 } 430 431 static inline struct inode *ceph_find_inode(struct super_block *sb, 432 struct ceph_vino vino) 433 { 434 ino_t t = ceph_vino_to_ino(vino); 435 return ilookup5(sb, t, ceph_ino_compare, &vino); 436 } 437 438 439 /* 440 * Ceph inode. 441 */ 442 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 443 #define CEPH_I_NODELAY (1 << 1) /* do not delay cap release */ 444 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 445 #define CEPH_I_NOFLUSH (1 << 3) /* do not flush dirty caps */ 446 #define CEPH_I_POOL_PERM (1 << 4) /* pool rd/wr bits are valid */ 447 #define CEPH_I_POOL_RD (1 << 5) /* can read from pool */ 448 #define CEPH_I_POOL_WR (1 << 6) /* can write to pool */ 449 450 451 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 452 int release_count, int ordered_count) 453 { 454 atomic_set(&ci->i_complete_count, release_count); 455 if (ci->i_ordered_count == ordered_count) 456 ci->i_ceph_flags |= CEPH_I_DIR_ORDERED; 457 else 458 ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED; 459 } 460 461 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 462 { 463 atomic_inc(&ci->i_release_count); 464 } 465 466 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 467 { 468 return atomic_read(&ci->i_complete_count) == 469 atomic_read(&ci->i_release_count); 470 } 471 472 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 473 { 474 return __ceph_dir_is_complete(ci) && 475 (ci->i_ceph_flags & CEPH_I_DIR_ORDERED); 476 } 477 478 static inline void ceph_dir_clear_complete(struct inode *inode) 479 { 480 __ceph_dir_clear_complete(ceph_inode(inode)); 481 } 482 483 static inline void ceph_dir_clear_ordered(struct inode *inode) 484 { 485 struct ceph_inode_info *ci = ceph_inode(inode); 486 spin_lock(&ci->i_ceph_lock); 487 ci->i_ordered_count++; 488 ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED; 489 spin_unlock(&ci->i_ceph_lock); 490 } 491 492 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 493 { 494 struct ceph_inode_info *ci = ceph_inode(inode); 495 bool ret; 496 spin_lock(&ci->i_ceph_lock); 497 ret = __ceph_dir_is_complete_ordered(ci); 498 spin_unlock(&ci->i_ceph_lock); 499 return ret; 500 } 501 502 /* find a specific frag @f */ 503 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 504 u32 f); 505 506 /* 507 * choose fragment for value @v. copy frag content to pfrag, if leaf 508 * exists 509 */ 510 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 511 struct ceph_inode_frag *pfrag, 512 int *found); 513 514 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) 515 { 516 return (struct ceph_dentry_info *)dentry->d_fsdata; 517 } 518 519 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) 520 { 521 return ((loff_t)frag << 32) | (loff_t)off; 522 } 523 524 /* 525 * caps helpers 526 */ 527 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 528 { 529 return !RB_EMPTY_ROOT(&ci->i_caps); 530 } 531 532 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 533 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 534 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 535 struct ceph_cap *cap); 536 537 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 538 { 539 int issued; 540 spin_lock(&ci->i_ceph_lock); 541 issued = __ceph_caps_issued(ci, NULL); 542 spin_unlock(&ci->i_ceph_lock); 543 return issued; 544 } 545 546 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, 547 int touch) 548 { 549 int r; 550 spin_lock(&ci->i_ceph_lock); 551 r = __ceph_caps_issued_mask(ci, mask, touch); 552 spin_unlock(&ci->i_ceph_lock); 553 return r; 554 } 555 556 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 557 { 558 return ci->i_dirty_caps | ci->i_flushing_caps; 559 } 560 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); 561 562 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 563 struct ceph_cap *ocap, int mask); 564 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 565 extern int __ceph_caps_used(struct ceph_inode_info *ci); 566 567 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 568 569 /* 570 * wanted, by virtue of open file modes AND cap refs (buffered/cached data) 571 */ 572 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) 573 { 574 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); 575 if (w & CEPH_CAP_FILE_BUFFER) 576 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ 577 return w; 578 } 579 580 /* what the mds thinks we want */ 581 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); 582 583 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 584 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 585 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); 586 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc, 587 struct ceph_cap_reservation *ctx, int need); 588 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, 589 struct ceph_cap_reservation *ctx); 590 extern void ceph_reservation_status(struct ceph_fs_client *client, 591 int *total, int *avail, int *used, 592 int *reserved, int *min); 593 594 595 596 /* 597 * we keep buffered readdir results attached to file->private_data 598 */ 599 #define CEPH_F_SYNC 1 600 #define CEPH_F_ATEND 2 601 602 struct ceph_file_info { 603 short fmode; /* initialized on open */ 604 short flags; /* CEPH_F_* */ 605 606 /* readdir: position within the dir */ 607 u32 frag; 608 struct ceph_mds_request *last_readdir; 609 610 /* readdir: position within a frag */ 611 unsigned offset; /* offset of last chunk, adjusted for . and .. */ 612 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 613 char *last_name; /* last entry in previous chunk */ 614 struct dentry *dentry; /* next dentry (for dcache readdir) */ 615 int dir_release_count; 616 int dir_ordered_count; 617 618 /* used for -o dirstat read() on directory thing */ 619 char *dir_info; 620 int dir_info_len; 621 }; 622 623 624 625 /* 626 * A "snap realm" describes a subset of the file hierarchy sharing 627 * the same set of snapshots that apply to it. The realms themselves 628 * are organized into a hierarchy, such that children inherit (some of) 629 * the snapshots of their parents. 630 * 631 * All inodes within the realm that have capabilities are linked into a 632 * per-realm list. 633 */ 634 struct ceph_snap_realm { 635 u64 ino; 636 atomic_t nref; 637 struct rb_node node; 638 639 u64 created, seq; 640 u64 parent_ino; 641 u64 parent_since; /* snapid when our current parent became so */ 642 643 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 644 u32 num_prior_parent_snaps; /* had prior to parent_since */ 645 u64 *snaps; /* snaps specific to this realm */ 646 u32 num_snaps; 647 648 struct ceph_snap_realm *parent; 649 struct list_head children; /* list of child realms */ 650 struct list_head child_item; 651 652 struct list_head empty_item; /* if i have ref==0 */ 653 654 struct list_head dirty_item; /* if realm needs new context */ 655 656 /* the current set of snaps for this realm */ 657 struct ceph_snap_context *cached_context; 658 659 struct list_head inodes_with_caps; 660 spinlock_t inodes_with_caps_lock; 661 }; 662 663 static inline int default_congestion_kb(void) 664 { 665 int congestion_kb; 666 667 /* 668 * Copied from NFS 669 * 670 * congestion size, scale with available memory. 671 * 672 * 64MB: 8192k 673 * 128MB: 11585k 674 * 256MB: 16384k 675 * 512MB: 23170k 676 * 1GB: 32768k 677 * 2GB: 46340k 678 * 4GB: 65536k 679 * 8GB: 92681k 680 * 16GB: 131072k 681 * 682 * This allows larger machines to have larger/more transfers. 683 * Limit the default to 256M 684 */ 685 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); 686 if (congestion_kb > 256*1024) 687 congestion_kb = 256*1024; 688 689 return congestion_kb; 690 } 691 692 693 694 /* snap.c */ 695 extern struct ceph_snap_context *ceph_empty_snapc; 696 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 697 u64 ino); 698 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 699 struct ceph_snap_realm *realm); 700 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 701 struct ceph_snap_realm *realm); 702 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 703 void *p, void *e, bool deletion, 704 struct ceph_snap_realm **realm_ret); 705 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 706 struct ceph_mds_session *session, 707 struct ceph_msg *msg); 708 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); 709 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 710 struct ceph_cap_snap *capsnap); 711 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 712 extern int ceph_snap_init(void); 713 extern void ceph_snap_exit(void); 714 715 /* 716 * a cap_snap is "pending" if it is still awaiting an in-progress 717 * sync write (that may/may not still update size, mtime, etc.). 718 */ 719 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 720 { 721 return !list_empty(&ci->i_cap_snaps) && 722 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, 723 ci_item)->writing; 724 } 725 726 /* inode.c */ 727 extern const struct inode_operations ceph_file_iops; 728 729 extern struct inode *ceph_alloc_inode(struct super_block *sb); 730 extern void ceph_destroy_inode(struct inode *inode); 731 extern int ceph_drop_inode(struct inode *inode); 732 733 extern struct inode *ceph_get_inode(struct super_block *sb, 734 struct ceph_vino vino); 735 extern struct inode *ceph_get_snapdir(struct inode *parent); 736 extern int ceph_fill_file_size(struct inode *inode, int issued, 737 u32 truncate_seq, u64 truncate_size, u64 size); 738 extern void ceph_fill_file_time(struct inode *inode, int issued, 739 u64 time_warp_seq, struct timespec *ctime, 740 struct timespec *mtime, struct timespec *atime); 741 extern int ceph_fill_trace(struct super_block *sb, 742 struct ceph_mds_request *req, 743 struct ceph_mds_session *session); 744 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 745 struct ceph_mds_session *session); 746 747 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 748 749 extern int ceph_inode_set_size(struct inode *inode, loff_t size); 750 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 751 extern void ceph_queue_vmtruncate(struct inode *inode); 752 753 extern void ceph_queue_invalidate(struct inode *inode); 754 extern void ceph_queue_writeback(struct inode *inode); 755 756 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 757 int mask, bool force); 758 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 759 { 760 return __ceph_do_getattr(inode, NULL, mask, force); 761 } 762 extern int ceph_permission(struct inode *inode, int mask); 763 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); 764 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, 765 struct kstat *stat); 766 767 /* xattr.c */ 768 extern int ceph_setxattr(struct dentry *, const char *, const void *, 769 size_t, int); 770 int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int); 771 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 772 int __ceph_removexattr(struct dentry *, const char *); 773 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); 774 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 775 extern int ceph_removexattr(struct dentry *, const char *); 776 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); 777 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 778 extern void __init ceph_xattr_init(void); 779 extern void ceph_xattr_exit(void); 780 extern const struct xattr_handler *ceph_xattr_handlers[]; 781 782 /* acl.c */ 783 struct ceph_acls_info { 784 void *default_acl; 785 void *acl; 786 struct ceph_pagelist *pagelist; 787 }; 788 789 #ifdef CONFIG_CEPH_FS_POSIX_ACL 790 791 struct posix_acl *ceph_get_acl(struct inode *, int); 792 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); 793 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 794 struct ceph_acls_info *info); 795 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info); 796 void ceph_release_acls_info(struct ceph_acls_info *info); 797 798 static inline void ceph_forget_all_cached_acls(struct inode *inode) 799 { 800 forget_all_cached_acls(inode); 801 } 802 803 #else 804 805 #define ceph_get_acl NULL 806 #define ceph_set_acl NULL 807 808 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 809 struct ceph_acls_info *info) 810 { 811 return 0; 812 } 813 static inline void ceph_init_inode_acls(struct inode *inode, 814 struct ceph_acls_info *info) 815 { 816 } 817 static inline void ceph_release_acls_info(struct ceph_acls_info *info) 818 { 819 } 820 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 821 { 822 return 0; 823 } 824 825 static inline void ceph_forget_all_cached_acls(struct inode *inode) 826 { 827 } 828 829 #endif 830 831 /* caps.c */ 832 extern const char *ceph_cap_string(int c); 833 extern void ceph_handle_caps(struct ceph_mds_session *session, 834 struct ceph_msg *msg); 835 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 836 struct ceph_cap_reservation *ctx); 837 extern void ceph_add_cap(struct inode *inode, 838 struct ceph_mds_session *session, u64 cap_id, 839 int fmode, unsigned issued, unsigned wanted, 840 unsigned cap, unsigned seq, u64 realmino, int flags, 841 struct ceph_cap **new_cap); 842 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 843 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 844 struct ceph_cap *cap); 845 extern int ceph_is_any_caps(struct inode *inode); 846 847 extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino, 848 u64 cap_id, u32 migrate_seq, u32 issue_seq); 849 extern void ceph_queue_caps_release(struct inode *inode); 850 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 851 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 852 int datasync); 853 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 854 struct ceph_mds_session *session); 855 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 856 int mds); 857 extern int ceph_get_cap_mds(struct inode *inode); 858 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 859 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 860 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 861 struct ceph_snap_context *snapc); 862 extern void __ceph_flush_snaps(struct ceph_inode_info *ci, 863 struct ceph_mds_session **psession, 864 int again); 865 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 866 struct ceph_mds_session *session); 867 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 868 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 869 870 extern int ceph_encode_inode_release(void **p, struct inode *inode, 871 int mds, int drop, int unless, int force); 872 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 873 int mds, int drop, int unless); 874 875 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, 876 loff_t endoff, int *got, struct page **pinned_page); 877 878 /* for counting open files by mode */ 879 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) 880 { 881 ci->i_nr_by_mode[mode]++; 882 } 883 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); 884 885 /* addr.c */ 886 extern const struct address_space_operations ceph_aops; 887 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 888 extern int ceph_uninline_data(struct file *filp, struct page *locked_page); 889 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need); 890 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 891 892 /* file.c */ 893 extern const struct file_operations ceph_file_fops; 894 895 extern int ceph_open(struct inode *inode, struct file *file); 896 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 897 struct file *file, unsigned flags, umode_t mode, 898 int *opened); 899 extern int ceph_release(struct inode *inode, struct file *filp); 900 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 901 char *data, size_t len); 902 /* dir.c */ 903 extern const struct file_operations ceph_dir_fops; 904 extern const struct file_operations ceph_snapdir_fops; 905 extern const struct inode_operations ceph_dir_iops; 906 extern const struct inode_operations ceph_snapdir_iops; 907 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, 908 ceph_snapdir_dentry_ops; 909 910 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 911 extern int ceph_handle_snapdir(struct ceph_mds_request *req, 912 struct dentry *dentry, int err); 913 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 914 struct dentry *dentry, int err); 915 916 extern void ceph_dentry_lru_add(struct dentry *dn); 917 extern void ceph_dentry_lru_touch(struct dentry *dn); 918 extern void ceph_dentry_lru_del(struct dentry *dn); 919 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 920 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 921 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry); 922 923 /* 924 * our d_ops vary depending on whether the inode is live, 925 * snapshotted (read-only), or a virtual ".snap" directory. 926 */ 927 int ceph_init_dentry(struct dentry *dentry); 928 929 930 /* ioctl.c */ 931 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 932 933 /* export.c */ 934 extern const struct export_operations ceph_export_ops; 935 936 /* locks.c */ 937 extern __init void ceph_flock_init(void); 938 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 939 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 940 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 941 extern int ceph_encode_locks_to_buffer(struct inode *inode, 942 struct ceph_filelock *flocks, 943 int num_fcntl_locks, 944 int num_flock_locks); 945 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 946 struct ceph_pagelist *pagelist, 947 int num_fcntl_locks, int num_flock_locks); 948 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); 949 950 /* debugfs.c */ 951 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); 952 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 953 954 #endif /* _FS_CEPH_SUPER_H */ 955