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 u64 cap_id; /* unique cap id (mds provided) */ 126 union { 127 /* in-use caps */ 128 struct { 129 int issued; /* latest, from the mds */ 130 int implemented; /* implemented superset of 131 issued (for revocation) */ 132 int mds, mds_wanted; 133 }; 134 /* caps to release */ 135 struct { 136 u64 cap_ino; 137 int queue_release; 138 }; 139 }; 140 u32 seq, issue_seq, mseq; 141 u32 cap_gen; /* active/stale cycle */ 142 unsigned long last_used; 143 struct list_head caps_item; 144 }; 145 146 #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ 147 #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ 148 #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ 149 150 /* 151 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 152 * we first complete any in-process sync writes and writeback any dirty 153 * data before flushing the snapped state (tracked here) back to the MDS. 154 */ 155 struct ceph_cap_snap { 156 atomic_t nref; 157 struct ceph_inode_info *ci; 158 struct list_head ci_item, flushing_item; 159 160 u64 follows, flush_tid; 161 int issued, dirty; 162 struct ceph_snap_context *context; 163 164 umode_t mode; 165 kuid_t uid; 166 kgid_t gid; 167 168 struct ceph_buffer *xattr_blob; 169 u64 xattr_version; 170 171 u64 size; 172 struct timespec mtime, atime, ctime; 173 u64 time_warp_seq; 174 int writing; /* a sync write is still in progress */ 175 int dirty_pages; /* dirty pages awaiting writeback */ 176 bool inline_data; 177 bool need_flush; 178 }; 179 180 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 181 { 182 if (atomic_dec_and_test(&capsnap->nref)) { 183 if (capsnap->xattr_blob) 184 ceph_buffer_put(capsnap->xattr_blob); 185 kfree(capsnap); 186 } 187 } 188 189 struct ceph_cap_flush { 190 u64 tid; 191 int caps; 192 bool kick; 193 struct rb_node g_node; // global 194 union { 195 struct rb_node i_node; // inode 196 struct list_head list; 197 }; 198 }; 199 200 /* 201 * The frag tree describes how a directory is fragmented, potentially across 202 * multiple metadata servers. It is also used to indicate points where 203 * metadata authority is delegated, and whether/where metadata is replicated. 204 * 205 * A _leaf_ frag will be present in the i_fragtree IFF there is 206 * delegation info. That is, if mds >= 0 || ndist > 0. 207 */ 208 #define CEPH_MAX_DIRFRAG_REP 4 209 210 struct ceph_inode_frag { 211 struct rb_node node; 212 213 /* fragtree state */ 214 u32 frag; 215 int split_by; /* i.e. 2^(split_by) children */ 216 217 /* delegation and replication info */ 218 int mds; /* -1 if same authority as parent */ 219 int ndist; /* >0 if replicated */ 220 int dist[CEPH_MAX_DIRFRAG_REP]; 221 }; 222 223 /* 224 * We cache inode xattrs as an encoded blob until they are first used, 225 * at which point we parse them into an rbtree. 226 */ 227 struct ceph_inode_xattr { 228 struct rb_node node; 229 230 const char *name; 231 int name_len; 232 const char *val; 233 int val_len; 234 int dirty; 235 236 int should_free_name; 237 int should_free_val; 238 }; 239 240 /* 241 * Ceph dentry state 242 */ 243 struct ceph_dentry_info { 244 struct ceph_mds_session *lease_session; 245 u32 lease_gen, lease_shared_gen; 246 u32 lease_seq; 247 unsigned long lease_renew_after, lease_renew_from; 248 struct list_head lru; 249 struct dentry *dentry; 250 u64 time; 251 u64 offset; 252 }; 253 254 struct ceph_inode_xattrs_info { 255 /* 256 * (still encoded) xattr blob. we avoid the overhead of parsing 257 * this until someone actually calls getxattr, etc. 258 * 259 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 260 * NULL means we don't know. 261 */ 262 struct ceph_buffer *blob, *prealloc_blob; 263 264 struct rb_root index; 265 bool dirty; 266 int count; 267 int names_size; 268 int vals_size; 269 u64 version, index_version; 270 }; 271 272 /* 273 * Ceph inode. 274 */ 275 struct ceph_inode_info { 276 struct ceph_vino i_vino; /* ceph ino + snap */ 277 278 spinlock_t i_ceph_lock; 279 280 u64 i_version; 281 u64 i_inline_version; 282 u32 i_time_warp_seq; 283 284 unsigned i_ceph_flags; 285 int i_ordered_count; 286 atomic_t i_release_count; 287 atomic_t i_complete_count; 288 289 struct ceph_dir_layout i_dir_layout; 290 struct ceph_file_layout i_layout; 291 char *i_symlink; 292 293 /* for dirs */ 294 struct timespec i_rctime; 295 u64 i_rbytes, i_rfiles, i_rsubdirs; 296 u64 i_files, i_subdirs; 297 298 struct rb_root i_fragtree; 299 struct mutex i_fragtree_mutex; 300 301 struct ceph_inode_xattrs_info i_xattrs; 302 303 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 304 * s_mutex. */ 305 struct rb_root i_caps; /* cap list */ 306 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 307 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 308 struct list_head i_dirty_item, i_flushing_item; 309 /* we need to track cap writeback on a per-cap-bit basis, to allow 310 * overlapping, pipelined cap flushes to the mds. we can probably 311 * reduce the tid to 8 bits if we're concerned about inode size. */ 312 struct ceph_cap_flush *i_prealloc_cap_flush; 313 struct rb_root i_cap_flush_tree; 314 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 315 unsigned long i_hold_caps_min; /* jiffies */ 316 unsigned long i_hold_caps_max; /* jiffies */ 317 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 318 struct ceph_cap_reservation i_cap_migration_resv; 319 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 320 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 321 dirty|flushing caps */ 322 unsigned i_snap_caps; /* cap bits for snapped files */ 323 324 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ 325 326 struct mutex i_truncate_mutex; 327 u32 i_truncate_seq; /* last truncate to smaller size */ 328 u64 i_truncate_size; /* and the size we last truncated down to */ 329 int i_truncate_pending; /* still need to call vmtruncate */ 330 331 u64 i_max_size; /* max file size authorized by mds */ 332 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 333 u64 i_wanted_max_size; /* offset we'd like to write too */ 334 u64 i_requested_max_size; /* max_size we've requested */ 335 336 /* held references to caps */ 337 int i_pin_ref; 338 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref; 339 int i_wrbuffer_ref, i_wrbuffer_ref_head; 340 u32 i_shared_gen; /* increment each time we get FILE_SHARED */ 341 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 342 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 343 344 struct list_head i_unsafe_writes; /* uncommitted sync writes */ 345 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 346 spinlock_t i_unsafe_lock; 347 348 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 349 int i_snap_realm_counter; /* snap realm (if caps) */ 350 struct list_head i_snap_realm_item; 351 struct list_head i_snap_flush_item; 352 353 struct work_struct i_wb_work; /* writeback work */ 354 struct work_struct i_pg_inv_work; /* page invalidation work */ 355 356 struct work_struct i_vmtruncate_work; 357 358 #ifdef CONFIG_CEPH_FSCACHE 359 struct fscache_cookie *fscache; 360 u32 i_fscache_gen; /* sequence, for delayed fscache validate */ 361 struct work_struct i_revalidate_work; 362 #endif 363 struct inode vfs_inode; /* at end */ 364 }; 365 366 static inline struct ceph_inode_info *ceph_inode(struct inode *inode) 367 { 368 return container_of(inode, struct ceph_inode_info, vfs_inode); 369 } 370 371 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) 372 { 373 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 374 } 375 376 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) 377 { 378 return (struct ceph_fs_client *)sb->s_fs_info; 379 } 380 381 static inline struct ceph_vino ceph_vino(struct inode *inode) 382 { 383 return ceph_inode(inode)->i_vino; 384 } 385 386 /* 387 * ino_t is <64 bits on many architectures, blech. 388 * 389 * i_ino (kernel inode) st_ino (userspace) 390 * i386 32 32 391 * x86_64+ino32 64 32 392 * x86_64 64 64 393 */ 394 static inline u32 ceph_ino_to_ino32(__u64 vino) 395 { 396 u32 ino = vino & 0xffffffff; 397 ino ^= vino >> 32; 398 if (!ino) 399 ino = 2; 400 return ino; 401 } 402 403 /* 404 * kernel i_ino value 405 */ 406 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) 407 { 408 #if BITS_PER_LONG == 32 409 return ceph_ino_to_ino32(vino.ino); 410 #else 411 return (ino_t)vino.ino; 412 #endif 413 } 414 415 /* 416 * user-visible ino (stat, filldir) 417 */ 418 #if BITS_PER_LONG == 32 419 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 420 { 421 return ino; 422 } 423 #else 424 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 425 { 426 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) 427 ino = ceph_ino_to_ino32(ino); 428 return ino; 429 } 430 #endif 431 432 433 /* for printf-style formatting */ 434 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 435 436 static inline u64 ceph_ino(struct inode *inode) 437 { 438 return ceph_inode(inode)->i_vino.ino; 439 } 440 static inline u64 ceph_snap(struct inode *inode) 441 { 442 return ceph_inode(inode)->i_vino.snap; 443 } 444 445 static inline int ceph_ino_compare(struct inode *inode, void *data) 446 { 447 struct ceph_vino *pvino = (struct ceph_vino *)data; 448 struct ceph_inode_info *ci = ceph_inode(inode); 449 return ci->i_vino.ino == pvino->ino && 450 ci->i_vino.snap == pvino->snap; 451 } 452 453 static inline struct inode *ceph_find_inode(struct super_block *sb, 454 struct ceph_vino vino) 455 { 456 ino_t t = ceph_vino_to_ino(vino); 457 return ilookup5(sb, t, ceph_ino_compare, &vino); 458 } 459 460 461 /* 462 * Ceph inode. 463 */ 464 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 465 #define CEPH_I_NODELAY (1 << 1) /* do not delay cap release */ 466 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 467 #define CEPH_I_NOFLUSH (1 << 3) /* do not flush dirty caps */ 468 #define CEPH_I_POOL_PERM (1 << 4) /* pool rd/wr bits are valid */ 469 #define CEPH_I_POOL_RD (1 << 5) /* can read from pool */ 470 #define CEPH_I_POOL_WR (1 << 6) /* can write to pool */ 471 472 473 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 474 int release_count, int ordered_count) 475 { 476 atomic_set(&ci->i_complete_count, release_count); 477 if (ci->i_ordered_count == ordered_count) 478 ci->i_ceph_flags |= CEPH_I_DIR_ORDERED; 479 else 480 ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED; 481 } 482 483 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 484 { 485 atomic_inc(&ci->i_release_count); 486 } 487 488 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 489 { 490 return atomic_read(&ci->i_complete_count) == 491 atomic_read(&ci->i_release_count); 492 } 493 494 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 495 { 496 return __ceph_dir_is_complete(ci) && 497 (ci->i_ceph_flags & CEPH_I_DIR_ORDERED); 498 } 499 500 static inline void ceph_dir_clear_complete(struct inode *inode) 501 { 502 __ceph_dir_clear_complete(ceph_inode(inode)); 503 } 504 505 static inline void ceph_dir_clear_ordered(struct inode *inode) 506 { 507 struct ceph_inode_info *ci = ceph_inode(inode); 508 spin_lock(&ci->i_ceph_lock); 509 ci->i_ordered_count++; 510 ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED; 511 spin_unlock(&ci->i_ceph_lock); 512 } 513 514 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 515 { 516 struct ceph_inode_info *ci = ceph_inode(inode); 517 bool ret; 518 spin_lock(&ci->i_ceph_lock); 519 ret = __ceph_dir_is_complete_ordered(ci); 520 spin_unlock(&ci->i_ceph_lock); 521 return ret; 522 } 523 524 /* find a specific frag @f */ 525 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 526 u32 f); 527 528 /* 529 * choose fragment for value @v. copy frag content to pfrag, if leaf 530 * exists 531 */ 532 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 533 struct ceph_inode_frag *pfrag, 534 int *found); 535 536 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) 537 { 538 return (struct ceph_dentry_info *)dentry->d_fsdata; 539 } 540 541 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) 542 { 543 return ((loff_t)frag << 32) | (loff_t)off; 544 } 545 546 /* 547 * caps helpers 548 */ 549 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 550 { 551 return !RB_EMPTY_ROOT(&ci->i_caps); 552 } 553 554 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 555 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 556 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 557 struct ceph_cap *cap); 558 559 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 560 { 561 int issued; 562 spin_lock(&ci->i_ceph_lock); 563 issued = __ceph_caps_issued(ci, NULL); 564 spin_unlock(&ci->i_ceph_lock); 565 return issued; 566 } 567 568 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, 569 int touch) 570 { 571 int r; 572 spin_lock(&ci->i_ceph_lock); 573 r = __ceph_caps_issued_mask(ci, mask, touch); 574 spin_unlock(&ci->i_ceph_lock); 575 return r; 576 } 577 578 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 579 { 580 return ci->i_dirty_caps | ci->i_flushing_caps; 581 } 582 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 583 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 584 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 585 struct ceph_cap_flush **pcf); 586 587 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 588 struct ceph_cap *ocap, int mask); 589 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 590 extern int __ceph_caps_used(struct ceph_inode_info *ci); 591 592 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 593 594 /* 595 * wanted, by virtue of open file modes AND cap refs (buffered/cached data) 596 */ 597 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) 598 { 599 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); 600 if (w & CEPH_CAP_FILE_BUFFER) 601 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ 602 return w; 603 } 604 605 /* what the mds thinks we want */ 606 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); 607 608 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 609 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 610 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); 611 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc, 612 struct ceph_cap_reservation *ctx, int need); 613 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, 614 struct ceph_cap_reservation *ctx); 615 extern void ceph_reservation_status(struct ceph_fs_client *client, 616 int *total, int *avail, int *used, 617 int *reserved, int *min); 618 619 620 621 /* 622 * we keep buffered readdir results attached to file->private_data 623 */ 624 #define CEPH_F_SYNC 1 625 #define CEPH_F_ATEND 2 626 627 struct ceph_file_info { 628 short fmode; /* initialized on open */ 629 short flags; /* CEPH_F_* */ 630 631 /* readdir: position within the dir */ 632 u32 frag; 633 struct ceph_mds_request *last_readdir; 634 635 /* readdir: position within a frag */ 636 unsigned offset; /* offset of last chunk, adjusted for . and .. */ 637 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 638 char *last_name; /* last entry in previous chunk */ 639 struct dentry *dentry; /* next dentry (for dcache readdir) */ 640 int dir_release_count; 641 int dir_ordered_count; 642 643 /* used for -o dirstat read() on directory thing */ 644 char *dir_info; 645 int dir_info_len; 646 }; 647 648 649 650 /* 651 * A "snap realm" describes a subset of the file hierarchy sharing 652 * the same set of snapshots that apply to it. The realms themselves 653 * are organized into a hierarchy, such that children inherit (some of) 654 * the snapshots of their parents. 655 * 656 * All inodes within the realm that have capabilities are linked into a 657 * per-realm list. 658 */ 659 struct ceph_snap_realm { 660 u64 ino; 661 atomic_t nref; 662 struct rb_node node; 663 664 u64 created, seq; 665 u64 parent_ino; 666 u64 parent_since; /* snapid when our current parent became so */ 667 668 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 669 u32 num_prior_parent_snaps; /* had prior to parent_since */ 670 u64 *snaps; /* snaps specific to this realm */ 671 u32 num_snaps; 672 673 struct ceph_snap_realm *parent; 674 struct list_head children; /* list of child realms */ 675 struct list_head child_item; 676 677 struct list_head empty_item; /* if i have ref==0 */ 678 679 struct list_head dirty_item; /* if realm needs new context */ 680 681 /* the current set of snaps for this realm */ 682 struct ceph_snap_context *cached_context; 683 684 struct list_head inodes_with_caps; 685 spinlock_t inodes_with_caps_lock; 686 }; 687 688 static inline int default_congestion_kb(void) 689 { 690 int congestion_kb; 691 692 /* 693 * Copied from NFS 694 * 695 * congestion size, scale with available memory. 696 * 697 * 64MB: 8192k 698 * 128MB: 11585k 699 * 256MB: 16384k 700 * 512MB: 23170k 701 * 1GB: 32768k 702 * 2GB: 46340k 703 * 4GB: 65536k 704 * 8GB: 92681k 705 * 16GB: 131072k 706 * 707 * This allows larger machines to have larger/more transfers. 708 * Limit the default to 256M 709 */ 710 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); 711 if (congestion_kb > 256*1024) 712 congestion_kb = 256*1024; 713 714 return congestion_kb; 715 } 716 717 718 719 /* snap.c */ 720 extern struct ceph_snap_context *ceph_empty_snapc; 721 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 722 u64 ino); 723 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 724 struct ceph_snap_realm *realm); 725 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 726 struct ceph_snap_realm *realm); 727 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 728 void *p, void *e, bool deletion, 729 struct ceph_snap_realm **realm_ret); 730 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 731 struct ceph_mds_session *session, 732 struct ceph_msg *msg); 733 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); 734 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 735 struct ceph_cap_snap *capsnap); 736 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 737 extern int ceph_snap_init(void); 738 extern void ceph_snap_exit(void); 739 740 /* 741 * a cap_snap is "pending" if it is still awaiting an in-progress 742 * sync write (that may/may not still update size, mtime, etc.). 743 */ 744 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 745 { 746 return !list_empty(&ci->i_cap_snaps) && 747 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 748 ci_item)->writing; 749 } 750 751 /* inode.c */ 752 extern const struct inode_operations ceph_file_iops; 753 754 extern struct inode *ceph_alloc_inode(struct super_block *sb); 755 extern void ceph_destroy_inode(struct inode *inode); 756 extern int ceph_drop_inode(struct inode *inode); 757 758 extern struct inode *ceph_get_inode(struct super_block *sb, 759 struct ceph_vino vino); 760 extern struct inode *ceph_get_snapdir(struct inode *parent); 761 extern int ceph_fill_file_size(struct inode *inode, int issued, 762 u32 truncate_seq, u64 truncate_size, u64 size); 763 extern void ceph_fill_file_time(struct inode *inode, int issued, 764 u64 time_warp_seq, struct timespec *ctime, 765 struct timespec *mtime, struct timespec *atime); 766 extern int ceph_fill_trace(struct super_block *sb, 767 struct ceph_mds_request *req, 768 struct ceph_mds_session *session); 769 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 770 struct ceph_mds_session *session); 771 772 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 773 774 extern int ceph_inode_set_size(struct inode *inode, loff_t size); 775 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 776 extern void ceph_queue_vmtruncate(struct inode *inode); 777 778 extern void ceph_queue_invalidate(struct inode *inode); 779 extern void ceph_queue_writeback(struct inode *inode); 780 781 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 782 int mask, bool force); 783 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 784 { 785 return __ceph_do_getattr(inode, NULL, mask, force); 786 } 787 extern int ceph_permission(struct inode *inode, int mask); 788 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); 789 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, 790 struct kstat *stat); 791 792 /* xattr.c */ 793 extern int ceph_setxattr(struct dentry *, const char *, const void *, 794 size_t, int); 795 int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int); 796 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 797 int __ceph_removexattr(struct dentry *, const char *); 798 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); 799 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 800 extern int ceph_removexattr(struct dentry *, const char *); 801 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); 802 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 803 extern void __init ceph_xattr_init(void); 804 extern void ceph_xattr_exit(void); 805 extern const struct xattr_handler *ceph_xattr_handlers[]; 806 807 /* acl.c */ 808 struct ceph_acls_info { 809 void *default_acl; 810 void *acl; 811 struct ceph_pagelist *pagelist; 812 }; 813 814 #ifdef CONFIG_CEPH_FS_POSIX_ACL 815 816 struct posix_acl *ceph_get_acl(struct inode *, int); 817 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); 818 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 819 struct ceph_acls_info *info); 820 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info); 821 void ceph_release_acls_info(struct ceph_acls_info *info); 822 823 static inline void ceph_forget_all_cached_acls(struct inode *inode) 824 { 825 forget_all_cached_acls(inode); 826 } 827 828 #else 829 830 #define ceph_get_acl NULL 831 #define ceph_set_acl NULL 832 833 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 834 struct ceph_acls_info *info) 835 { 836 return 0; 837 } 838 static inline void ceph_init_inode_acls(struct inode *inode, 839 struct ceph_acls_info *info) 840 { 841 } 842 static inline void ceph_release_acls_info(struct ceph_acls_info *info) 843 { 844 } 845 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 846 { 847 return 0; 848 } 849 850 static inline void ceph_forget_all_cached_acls(struct inode *inode) 851 { 852 } 853 854 #endif 855 856 /* caps.c */ 857 extern const char *ceph_cap_string(int c); 858 extern void ceph_handle_caps(struct ceph_mds_session *session, 859 struct ceph_msg *msg); 860 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 861 struct ceph_cap_reservation *ctx); 862 extern void ceph_add_cap(struct inode *inode, 863 struct ceph_mds_session *session, u64 cap_id, 864 int fmode, unsigned issued, unsigned wanted, 865 unsigned cap, unsigned seq, u64 realmino, int flags, 866 struct ceph_cap **new_cap); 867 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 868 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 869 struct ceph_cap *cap); 870 extern int ceph_is_any_caps(struct inode *inode); 871 872 extern void ceph_queue_caps_release(struct inode *inode); 873 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 874 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 875 int datasync); 876 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 877 struct ceph_mds_session *session); 878 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 879 struct ceph_mds_session *session); 880 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 881 int mds); 882 extern int ceph_get_cap_mds(struct inode *inode); 883 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 884 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 885 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 886 struct ceph_snap_context *snapc); 887 extern void __ceph_flush_snaps(struct ceph_inode_info *ci, 888 struct ceph_mds_session **psession, 889 int again); 890 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 891 struct ceph_mds_session *session); 892 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 893 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 894 895 extern int ceph_encode_inode_release(void **p, struct inode *inode, 896 int mds, int drop, int unless, int force); 897 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 898 int mds, int drop, int unless); 899 900 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, 901 loff_t endoff, int *got, struct page **pinned_page); 902 903 /* for counting open files by mode */ 904 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) 905 { 906 ci->i_nr_by_mode[mode]++; 907 } 908 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); 909 910 /* addr.c */ 911 extern const struct address_space_operations ceph_aops; 912 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 913 extern int ceph_uninline_data(struct file *filp, struct page *locked_page); 914 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need); 915 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 916 917 /* file.c */ 918 extern const struct file_operations ceph_file_fops; 919 920 extern int ceph_open(struct inode *inode, struct file *file); 921 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 922 struct file *file, unsigned flags, umode_t mode, 923 int *opened); 924 extern int ceph_release(struct inode *inode, struct file *filp); 925 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 926 char *data, size_t len); 927 /* dir.c */ 928 extern const struct file_operations ceph_dir_fops; 929 extern const struct file_operations ceph_snapdir_fops; 930 extern const struct inode_operations ceph_dir_iops; 931 extern const struct inode_operations ceph_snapdir_iops; 932 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, 933 ceph_snapdir_dentry_ops; 934 935 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 936 extern int ceph_handle_snapdir(struct ceph_mds_request *req, 937 struct dentry *dentry, int err); 938 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 939 struct dentry *dentry, int err); 940 941 extern void ceph_dentry_lru_add(struct dentry *dn); 942 extern void ceph_dentry_lru_touch(struct dentry *dn); 943 extern void ceph_dentry_lru_del(struct dentry *dn); 944 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 945 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 946 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry); 947 948 /* 949 * our d_ops vary depending on whether the inode is live, 950 * snapshotted (read-only), or a virtual ".snap" directory. 951 */ 952 int ceph_init_dentry(struct dentry *dentry); 953 954 955 /* ioctl.c */ 956 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 957 958 /* export.c */ 959 extern const struct export_operations ceph_export_ops; 960 961 /* locks.c */ 962 extern __init void ceph_flock_init(void); 963 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 964 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 965 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 966 extern int ceph_encode_locks_to_buffer(struct inode *inode, 967 struct ceph_filelock *flocks, 968 int num_fcntl_locks, 969 int num_flock_locks); 970 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 971 struct ceph_pagelist *pagelist, 972 int num_fcntl_locks, int num_flock_locks); 973 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); 974 975 /* debugfs.c */ 976 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); 977 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 978 979 #endif /* _FS_CEPH_SUPER_H */ 980