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 #include <linux/refcount.h> 18 19 #include <linux/ceph/libceph.h> 20 21 #ifdef CONFIG_CEPH_FSCACHE 22 #include <linux/fscache.h> 23 #endif 24 25 /* f_type in struct statfs */ 26 #define CEPH_SUPER_MAGIC 0x00c36400 27 28 /* large granularity for statfs utilization stats to facilitate 29 * large volume sizes on 32-bit machines. */ 30 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ 31 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 32 33 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ 34 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ 35 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ 36 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ 37 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ 38 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ 39 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */ 40 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */ 41 42 #define CEPH_MOUNT_OPT_DEFAULT CEPH_MOUNT_OPT_DCACHE 43 44 #define ceph_set_mount_opt(fsc, opt) \ 45 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt; 46 #define ceph_test_mount_opt(fsc, opt) \ 47 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 48 49 #define CEPH_RSIZE_DEFAULT (64*1024*1024) /* max read size */ 50 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */ 51 #define CEPH_MAX_READDIR_DEFAULT 1024 52 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 53 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 54 55 struct ceph_mount_options { 56 int flags; 57 int sb_flags; 58 59 int wsize; /* max write size */ 60 int rsize; /* max read size */ 61 int rasize; /* max readahead */ 62 int congestion_kb; /* max writeback in flight */ 63 int caps_wanted_delay_min, caps_wanted_delay_max; 64 int cap_release_safety; 65 int max_readdir; /* max readdir result (entires) */ 66 int max_readdir_bytes; /* max readdir result (bytes) */ 67 68 /* 69 * everything above this point can be memcmp'd; everything below 70 * is handled in compare_mount_options() 71 */ 72 73 char *snapdir_name; /* default ".snap" */ 74 char *mds_namespace; /* default NULL */ 75 char *server_path; /* default "/" */ 76 char *fscache_uniq; /* default NULL */ 77 }; 78 79 struct ceph_fs_client { 80 struct super_block *sb; 81 82 struct ceph_mount_options *mount_options; 83 struct ceph_client *client; 84 85 unsigned long mount_state; 86 int min_caps; /* min caps i added */ 87 88 struct ceph_mds_client *mdsc; 89 90 /* writeback */ 91 mempool_t *wb_pagevec_pool; 92 struct workqueue_struct *wb_wq; 93 struct workqueue_struct *pg_inv_wq; 94 struct workqueue_struct *trunc_wq; 95 atomic_long_t writeback_count; 96 97 #ifdef CONFIG_DEBUG_FS 98 struct dentry *debugfs_dentry_lru, *debugfs_caps; 99 struct dentry *debugfs_congestion_kb; 100 struct dentry *debugfs_bdi; 101 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 102 struct dentry *debugfs_mds_sessions; 103 #endif 104 105 #ifdef CONFIG_CEPH_FSCACHE 106 struct fscache_cookie *fscache; 107 #endif 108 }; 109 110 111 /* 112 * File i/o capability. This tracks shared state with the metadata 113 * server that allows us to cache or writeback attributes or to read 114 * and write data. For any given inode, we should have one or more 115 * capabilities, one issued by each metadata server, and our 116 * cumulative access is the OR of all issued capabilities. 117 * 118 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 119 * session capability lists. 120 */ 121 struct ceph_cap { 122 struct ceph_inode_info *ci; 123 struct rb_node ci_node; /* per-ci cap tree */ 124 struct ceph_mds_session *session; 125 struct list_head session_caps; /* per-session caplist */ 126 u64 cap_id; /* unique cap id (mds provided) */ 127 union { 128 /* in-use caps */ 129 struct { 130 int issued; /* latest, from the mds */ 131 int implemented; /* implemented superset of 132 issued (for revocation) */ 133 int mds, mds_wanted; 134 }; 135 /* caps to release */ 136 struct { 137 u64 cap_ino; 138 int queue_release; 139 }; 140 }; 141 u32 seq, issue_seq, mseq; 142 u32 cap_gen; /* active/stale cycle */ 143 unsigned long last_used; 144 struct list_head caps_item; 145 }; 146 147 #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ 148 #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ 149 #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ 150 151 struct ceph_cap_flush { 152 u64 tid; 153 int caps; /* 0 means capsnap */ 154 bool wake; /* wake up flush waiters when finish ? */ 155 struct list_head g_list; // global 156 struct list_head i_list; // per inode 157 }; 158 159 /* 160 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 161 * we first complete any in-process sync writes and writeback any dirty 162 * data before flushing the snapped state (tracked here) back to the MDS. 163 */ 164 struct ceph_cap_snap { 165 refcount_t nref; 166 struct list_head ci_item; 167 168 struct ceph_cap_flush cap_flush; 169 170 u64 follows; 171 int issued, dirty; 172 struct ceph_snap_context *context; 173 174 umode_t mode; 175 kuid_t uid; 176 kgid_t gid; 177 178 struct ceph_buffer *xattr_blob; 179 u64 xattr_version; 180 181 u64 size; 182 struct timespec mtime, atime, ctime; 183 u64 time_warp_seq; 184 u64 truncate_size; 185 u32 truncate_seq; 186 int writing; /* a sync write is still in progress */ 187 int dirty_pages; /* dirty pages awaiting writeback */ 188 bool inline_data; 189 bool need_flush; 190 }; 191 192 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 193 { 194 if (refcount_dec_and_test(&capsnap->nref)) { 195 if (capsnap->xattr_blob) 196 ceph_buffer_put(capsnap->xattr_blob); 197 kfree(capsnap); 198 } 199 } 200 201 /* 202 * The frag tree describes how a directory is fragmented, potentially across 203 * multiple metadata servers. It is also used to indicate points where 204 * metadata authority is delegated, and whether/where metadata is replicated. 205 * 206 * A _leaf_ frag will be present in the i_fragtree IFF there is 207 * delegation info. That is, if mds >= 0 || ndist > 0. 208 */ 209 #define CEPH_MAX_DIRFRAG_REP 4 210 211 struct ceph_inode_frag { 212 struct rb_node node; 213 214 /* fragtree state */ 215 u32 frag; 216 int split_by; /* i.e. 2^(split_by) children */ 217 218 /* delegation and replication info */ 219 int mds; /* -1 if same authority as parent */ 220 int ndist; /* >0 if replicated */ 221 int dist[CEPH_MAX_DIRFRAG_REP]; 222 }; 223 224 /* 225 * We cache inode xattrs as an encoded blob until they are first used, 226 * at which point we parse them into an rbtree. 227 */ 228 struct ceph_inode_xattr { 229 struct rb_node node; 230 231 const char *name; 232 int name_len; 233 const char *val; 234 int val_len; 235 int dirty; 236 237 int should_free_name; 238 int should_free_val; 239 }; 240 241 /* 242 * Ceph dentry state 243 */ 244 struct ceph_dentry_info { 245 struct ceph_mds_session *lease_session; 246 u32 lease_gen, lease_shared_gen; 247 u32 lease_seq; 248 unsigned long lease_renew_after, lease_renew_from; 249 struct list_head lru; 250 struct dentry *dentry; 251 unsigned long time; 252 u64 offset; 253 }; 254 255 struct ceph_inode_xattrs_info { 256 /* 257 * (still encoded) xattr blob. we avoid the overhead of parsing 258 * this until someone actually calls getxattr, etc. 259 * 260 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 261 * NULL means we don't know. 262 */ 263 struct ceph_buffer *blob, *prealloc_blob; 264 265 struct rb_root index; 266 bool dirty; 267 int count; 268 int names_size; 269 int vals_size; 270 u64 version, index_version; 271 }; 272 273 /* 274 * Ceph inode. 275 */ 276 struct ceph_inode_info { 277 struct ceph_vino i_vino; /* ceph ino + snap */ 278 279 spinlock_t i_ceph_lock; 280 281 u64 i_version; 282 u64 i_inline_version; 283 u32 i_time_warp_seq; 284 285 unsigned i_ceph_flags; 286 atomic64_t i_release_count; 287 atomic64_t i_ordered_count; 288 atomic64_t i_complete_seq[2]; 289 290 struct ceph_dir_layout i_dir_layout; 291 struct ceph_file_layout i_layout; 292 char *i_symlink; 293 294 /* for dirs */ 295 struct timespec i_rctime; 296 u64 i_rbytes, i_rfiles, i_rsubdirs; 297 u64 i_files, i_subdirs; 298 299 struct rb_root i_fragtree; 300 int i_fragtree_nsplits; 301 struct mutex i_fragtree_mutex; 302 303 struct ceph_inode_xattrs_info i_xattrs; 304 305 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 306 * s_mutex. */ 307 struct rb_root i_caps; /* cap list */ 308 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 309 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 310 struct list_head i_dirty_item, i_flushing_item; 311 /* we need to track cap writeback on a per-cap-bit basis, to allow 312 * overlapping, pipelined cap flushes to the mds. we can probably 313 * reduce the tid to 8 bits if we're concerned about inode size. */ 314 struct ceph_cap_flush *i_prealloc_cap_flush; 315 struct list_head i_cap_flush_list; 316 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 317 unsigned long i_hold_caps_min; /* jiffies */ 318 unsigned long i_hold_caps_max; /* jiffies */ 319 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 320 struct ceph_cap_reservation i_cap_migration_resv; 321 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 322 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 323 dirty|flushing caps */ 324 unsigned i_snap_caps; /* cap bits for snapped files */ 325 326 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ 327 328 struct mutex i_truncate_mutex; 329 u32 i_truncate_seq; /* last truncate to smaller size */ 330 u64 i_truncate_size; /* and the size we last truncated down to */ 331 int i_truncate_pending; /* still need to call vmtruncate */ 332 333 u64 i_max_size; /* max file size authorized by mds */ 334 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 335 u64 i_wanted_max_size; /* offset we'd like to write too */ 336 u64 i_requested_max_size; /* max_size we've requested */ 337 338 /* held references to caps */ 339 int i_pin_ref; 340 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref; 341 int i_wrbuffer_ref, i_wrbuffer_ref_head; 342 u32 i_shared_gen; /* increment each time we get FILE_SHARED */ 343 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 344 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 345 346 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 347 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 348 spinlock_t i_unsafe_lock; 349 350 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 351 int i_snap_realm_counter; /* snap realm (if caps) */ 352 struct list_head i_snap_realm_item; 353 struct list_head i_snap_flush_item; 354 355 struct work_struct i_wb_work; /* writeback work */ 356 struct work_struct i_pg_inv_work; /* page invalidation work */ 357 358 struct work_struct i_vmtruncate_work; 359 360 #ifdef CONFIG_CEPH_FSCACHE 361 struct fscache_cookie *fscache; 362 u32 i_fscache_gen; 363 #endif 364 struct inode vfs_inode; /* at end */ 365 }; 366 367 static inline struct ceph_inode_info *ceph_inode(struct inode *inode) 368 { 369 return container_of(inode, struct ceph_inode_info, vfs_inode); 370 } 371 372 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) 373 { 374 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 375 } 376 377 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) 378 { 379 return (struct ceph_fs_client *)sb->s_fs_info; 380 } 381 382 static inline struct ceph_vino ceph_vino(struct inode *inode) 383 { 384 return ceph_inode(inode)->i_vino; 385 } 386 387 /* 388 * ino_t is <64 bits on many architectures, blech. 389 * 390 * i_ino (kernel inode) st_ino (userspace) 391 * i386 32 32 392 * x86_64+ino32 64 32 393 * x86_64 64 64 394 */ 395 static inline u32 ceph_ino_to_ino32(__u64 vino) 396 { 397 u32 ino = vino & 0xffffffff; 398 ino ^= vino >> 32; 399 if (!ino) 400 ino = 2; 401 return ino; 402 } 403 404 /* 405 * kernel i_ino value 406 */ 407 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) 408 { 409 #if BITS_PER_LONG == 32 410 return ceph_ino_to_ino32(vino.ino); 411 #else 412 return (ino_t)vino.ino; 413 #endif 414 } 415 416 /* 417 * user-visible ino (stat, filldir) 418 */ 419 #if BITS_PER_LONG == 32 420 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 421 { 422 return ino; 423 } 424 #else 425 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 426 { 427 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) 428 ino = ceph_ino_to_ino32(ino); 429 return ino; 430 } 431 #endif 432 433 434 /* for printf-style formatting */ 435 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 436 437 static inline u64 ceph_ino(struct inode *inode) 438 { 439 return ceph_inode(inode)->i_vino.ino; 440 } 441 static inline u64 ceph_snap(struct inode *inode) 442 { 443 return ceph_inode(inode)->i_vino.snap; 444 } 445 446 static inline int ceph_ino_compare(struct inode *inode, void *data) 447 { 448 struct ceph_vino *pvino = (struct ceph_vino *)data; 449 struct ceph_inode_info *ci = ceph_inode(inode); 450 return ci->i_vino.ino == pvino->ino && 451 ci->i_vino.snap == pvino->snap; 452 } 453 454 static inline struct inode *ceph_find_inode(struct super_block *sb, 455 struct ceph_vino vino) 456 { 457 ino_t t = ceph_vino_to_ino(vino); 458 return ilookup5(sb, t, ceph_ino_compare, &vino); 459 } 460 461 462 /* 463 * Ceph inode. 464 */ 465 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 466 #define CEPH_I_NODELAY (1 << 1) /* do not delay cap release */ 467 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 468 #define CEPH_I_NOFLUSH (1 << 3) /* do not flush dirty caps */ 469 #define CEPH_I_POOL_PERM (1 << 4) /* pool rd/wr bits are valid */ 470 #define CEPH_I_POOL_RD (1 << 5) /* can read from pool */ 471 #define CEPH_I_POOL_WR (1 << 6) /* can write to pool */ 472 #define CEPH_I_SEC_INITED (1 << 7) /* security initialized */ 473 #define CEPH_I_CAP_DROPPED (1 << 8) /* caps were forcibly dropped */ 474 #define CEPH_I_KICK_FLUSH (1 << 9) /* kick flushing caps */ 475 #define CEPH_I_FLUSH_SNAPS (1 << 10) /* need flush snapss */ 476 #define CEPH_I_ERROR_WRITE (1 << 11) /* have seen write errors */ 477 478 /* 479 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 480 * and then clear it when they start succeeding. Note that we do a lockless 481 * check first, and only take the lock if it looks like it needs to be changed. 482 * The write submission code just takes this as a hint, so we're not too 483 * worried if a few slip through in either direction. 484 */ 485 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 486 { 487 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { 488 spin_lock(&ci->i_ceph_lock); 489 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; 490 spin_unlock(&ci->i_ceph_lock); 491 } 492 } 493 494 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 495 { 496 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { 497 spin_lock(&ci->i_ceph_lock); 498 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; 499 spin_unlock(&ci->i_ceph_lock); 500 } 501 } 502 503 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 504 long long release_count, 505 long long ordered_count) 506 { 507 smp_mb__before_atomic(); 508 atomic64_set(&ci->i_complete_seq[0], release_count); 509 atomic64_set(&ci->i_complete_seq[1], ordered_count); 510 } 511 512 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 513 { 514 atomic64_inc(&ci->i_release_count); 515 } 516 517 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 518 { 519 atomic64_inc(&ci->i_ordered_count); 520 } 521 522 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 523 { 524 return atomic64_read(&ci->i_complete_seq[0]) == 525 atomic64_read(&ci->i_release_count); 526 } 527 528 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 529 { 530 return atomic64_read(&ci->i_complete_seq[0]) == 531 atomic64_read(&ci->i_release_count) && 532 atomic64_read(&ci->i_complete_seq[1]) == 533 atomic64_read(&ci->i_ordered_count); 534 } 535 536 static inline void ceph_dir_clear_complete(struct inode *inode) 537 { 538 __ceph_dir_clear_complete(ceph_inode(inode)); 539 } 540 541 static inline void ceph_dir_clear_ordered(struct inode *inode) 542 { 543 __ceph_dir_clear_ordered(ceph_inode(inode)); 544 } 545 546 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 547 { 548 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 549 smp_rmb(); 550 return ret; 551 } 552 553 /* find a specific frag @f */ 554 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 555 u32 f); 556 557 /* 558 * choose fragment for value @v. copy frag content to pfrag, if leaf 559 * exists 560 */ 561 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 562 struct ceph_inode_frag *pfrag, 563 int *found); 564 565 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) 566 { 567 return (struct ceph_dentry_info *)dentry->d_fsdata; 568 } 569 570 /* 571 * caps helpers 572 */ 573 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 574 { 575 return !RB_EMPTY_ROOT(&ci->i_caps); 576 } 577 578 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 579 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 580 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 581 struct ceph_cap *cap); 582 583 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 584 { 585 int issued; 586 spin_lock(&ci->i_ceph_lock); 587 issued = __ceph_caps_issued(ci, NULL); 588 spin_unlock(&ci->i_ceph_lock); 589 return issued; 590 } 591 592 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, 593 int touch) 594 { 595 int r; 596 spin_lock(&ci->i_ceph_lock); 597 r = __ceph_caps_issued_mask(ci, mask, touch); 598 spin_unlock(&ci->i_ceph_lock); 599 return r; 600 } 601 602 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 603 { 604 return ci->i_dirty_caps | ci->i_flushing_caps; 605 } 606 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 607 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 608 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 609 struct ceph_cap_flush **pcf); 610 611 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 612 struct ceph_cap *ocap, int mask); 613 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 614 extern int __ceph_caps_used(struct ceph_inode_info *ci); 615 616 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 617 618 /* 619 * wanted, by virtue of open file modes AND cap refs (buffered/cached data) 620 */ 621 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) 622 { 623 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); 624 if (w & CEPH_CAP_FILE_BUFFER) 625 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ 626 return w; 627 } 628 629 /* what the mds thinks we want */ 630 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 631 632 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 633 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 634 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); 635 extern void ceph_reserve_caps(struct ceph_mds_client *mdsc, 636 struct ceph_cap_reservation *ctx, int need); 637 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, 638 struct ceph_cap_reservation *ctx); 639 extern void ceph_reservation_status(struct ceph_fs_client *client, 640 int *total, int *avail, int *used, 641 int *reserved, int *min); 642 643 644 645 /* 646 * we keep buffered readdir results attached to file->private_data 647 */ 648 #define CEPH_F_SYNC 1 649 #define CEPH_F_ATEND 2 650 651 struct ceph_file_info { 652 short fmode; /* initialized on open */ 653 short flags; /* CEPH_F_* */ 654 655 /* readdir: position within the dir */ 656 u32 frag; 657 struct ceph_mds_request *last_readdir; 658 659 /* readdir: position within a frag */ 660 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 661 char *last_name; /* last entry in previous chunk */ 662 long long dir_release_count; 663 long long dir_ordered_count; 664 int readdir_cache_idx; 665 666 /* used for -o dirstat read() on directory thing */ 667 char *dir_info; 668 int dir_info_len; 669 }; 670 671 struct ceph_readdir_cache_control { 672 struct page *page; 673 struct dentry **dentries; 674 int index; 675 }; 676 677 /* 678 * A "snap realm" describes a subset of the file hierarchy sharing 679 * the same set of snapshots that apply to it. The realms themselves 680 * are organized into a hierarchy, such that children inherit (some of) 681 * the snapshots of their parents. 682 * 683 * All inodes within the realm that have capabilities are linked into a 684 * per-realm list. 685 */ 686 struct ceph_snap_realm { 687 u64 ino; 688 atomic_t nref; 689 struct rb_node node; 690 691 u64 created, seq; 692 u64 parent_ino; 693 u64 parent_since; /* snapid when our current parent became so */ 694 695 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 696 u32 num_prior_parent_snaps; /* had prior to parent_since */ 697 u64 *snaps; /* snaps specific to this realm */ 698 u32 num_snaps; 699 700 struct ceph_snap_realm *parent; 701 struct list_head children; /* list of child realms */ 702 struct list_head child_item; 703 704 struct list_head empty_item; /* if i have ref==0 */ 705 706 struct list_head dirty_item; /* if realm needs new context */ 707 708 /* the current set of snaps for this realm */ 709 struct ceph_snap_context *cached_context; 710 711 struct list_head inodes_with_caps; 712 spinlock_t inodes_with_caps_lock; 713 }; 714 715 static inline int default_congestion_kb(void) 716 { 717 int congestion_kb; 718 719 /* 720 * Copied from NFS 721 * 722 * congestion size, scale with available memory. 723 * 724 * 64MB: 8192k 725 * 128MB: 11585k 726 * 256MB: 16384k 727 * 512MB: 23170k 728 * 1GB: 32768k 729 * 2GB: 46340k 730 * 4GB: 65536k 731 * 8GB: 92681k 732 * 16GB: 131072k 733 * 734 * This allows larger machines to have larger/more transfers. 735 * Limit the default to 256M 736 */ 737 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); 738 if (congestion_kb > 256*1024) 739 congestion_kb = 256*1024; 740 741 return congestion_kb; 742 } 743 744 745 746 /* snap.c */ 747 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 748 u64 ino); 749 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 750 struct ceph_snap_realm *realm); 751 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 752 struct ceph_snap_realm *realm); 753 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 754 void *p, void *e, bool deletion, 755 struct ceph_snap_realm **realm_ret); 756 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 757 struct ceph_mds_session *session, 758 struct ceph_msg *msg); 759 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); 760 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 761 struct ceph_cap_snap *capsnap); 762 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 763 764 /* 765 * a cap_snap is "pending" if it is still awaiting an in-progress 766 * sync write (that may/may not still update size, mtime, etc.). 767 */ 768 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 769 { 770 return !list_empty(&ci->i_cap_snaps) && 771 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 772 ci_item)->writing; 773 } 774 775 /* inode.c */ 776 extern const struct inode_operations ceph_file_iops; 777 778 extern struct inode *ceph_alloc_inode(struct super_block *sb); 779 extern void ceph_destroy_inode(struct inode *inode); 780 extern int ceph_drop_inode(struct inode *inode); 781 782 extern struct inode *ceph_get_inode(struct super_block *sb, 783 struct ceph_vino vino); 784 extern struct inode *ceph_get_snapdir(struct inode *parent); 785 extern int ceph_fill_file_size(struct inode *inode, int issued, 786 u32 truncate_seq, u64 truncate_size, u64 size); 787 extern void ceph_fill_file_time(struct inode *inode, int issued, 788 u64 time_warp_seq, struct timespec *ctime, 789 struct timespec *mtime, struct timespec *atime); 790 extern int ceph_fill_trace(struct super_block *sb, 791 struct ceph_mds_request *req); 792 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 793 struct ceph_mds_session *session); 794 795 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 796 797 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 798 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 799 extern void ceph_queue_vmtruncate(struct inode *inode); 800 801 extern void ceph_queue_invalidate(struct inode *inode); 802 extern void ceph_queue_writeback(struct inode *inode); 803 804 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 805 int mask, bool force); 806 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 807 { 808 return __ceph_do_getattr(inode, NULL, mask, force); 809 } 810 extern int ceph_permission(struct inode *inode, int mask); 811 extern int __ceph_setattr(struct inode *inode, struct iattr *attr); 812 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); 813 extern int ceph_getattr(const struct path *path, struct kstat *stat, 814 u32 request_mask, unsigned int flags); 815 816 /* xattr.c */ 817 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 818 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 819 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 820 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); 821 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 822 extern void __init ceph_xattr_init(void); 823 extern void ceph_xattr_exit(void); 824 extern const struct xattr_handler *ceph_xattr_handlers[]; 825 826 #ifdef CONFIG_SECURITY 827 extern bool ceph_security_xattr_deadlock(struct inode *in); 828 extern bool ceph_security_xattr_wanted(struct inode *in); 829 #else 830 static inline bool ceph_security_xattr_deadlock(struct inode *in) 831 { 832 return false; 833 } 834 static inline bool ceph_security_xattr_wanted(struct inode *in) 835 { 836 return false; 837 } 838 #endif 839 840 /* acl.c */ 841 struct ceph_acls_info { 842 void *default_acl; 843 void *acl; 844 struct ceph_pagelist *pagelist; 845 }; 846 847 #ifdef CONFIG_CEPH_FS_POSIX_ACL 848 849 struct posix_acl *ceph_get_acl(struct inode *, int); 850 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); 851 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 852 struct ceph_acls_info *info); 853 void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info); 854 void ceph_release_acls_info(struct ceph_acls_info *info); 855 856 static inline void ceph_forget_all_cached_acls(struct inode *inode) 857 { 858 forget_all_cached_acls(inode); 859 } 860 861 #else 862 863 #define ceph_get_acl NULL 864 #define ceph_set_acl NULL 865 866 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 867 struct ceph_acls_info *info) 868 { 869 return 0; 870 } 871 static inline void ceph_init_inode_acls(struct inode *inode, 872 struct ceph_acls_info *info) 873 { 874 } 875 static inline void ceph_release_acls_info(struct ceph_acls_info *info) 876 { 877 } 878 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 879 { 880 return 0; 881 } 882 883 static inline void ceph_forget_all_cached_acls(struct inode *inode) 884 { 885 } 886 887 #endif 888 889 /* caps.c */ 890 extern const char *ceph_cap_string(int c); 891 extern void ceph_handle_caps(struct ceph_mds_session *session, 892 struct ceph_msg *msg); 893 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 894 struct ceph_cap_reservation *ctx); 895 extern void ceph_add_cap(struct inode *inode, 896 struct ceph_mds_session *session, u64 cap_id, 897 int fmode, unsigned issued, unsigned wanted, 898 unsigned cap, unsigned seq, u64 realmino, int flags, 899 struct ceph_cap **new_cap); 900 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 901 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 902 struct ceph_cap *cap); 903 extern int ceph_is_any_caps(struct inode *inode); 904 905 extern void ceph_queue_caps_release(struct inode *inode); 906 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 907 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 908 int datasync); 909 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 910 struct ceph_mds_session *session); 911 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 912 struct ceph_mds_session *session); 913 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 914 int mds); 915 extern int ceph_get_cap_mds(struct inode *inode); 916 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 917 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 918 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 919 struct ceph_snap_context *snapc); 920 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 921 struct ceph_mds_session **psession); 922 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 923 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 924 struct ceph_mds_session *session); 925 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 926 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 927 928 extern int ceph_encode_inode_release(void **p, struct inode *inode, 929 int mds, int drop, int unless, int force); 930 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 931 struct inode *dir, 932 int mds, int drop, int unless); 933 934 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, 935 loff_t endoff, int *got, struct page **pinned_page); 936 extern int ceph_try_get_caps(struct ceph_inode_info *ci, 937 int need, int want, int *got); 938 939 /* for counting open files by mode */ 940 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode); 941 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); 942 943 /* addr.c */ 944 extern const struct address_space_operations ceph_aops; 945 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 946 extern int ceph_uninline_data(struct file *filp, struct page *locked_page); 947 extern int ceph_pool_perm_check(struct ceph_inode_info *ci, int need); 948 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 949 950 /* file.c */ 951 extern const struct file_operations ceph_file_fops; 952 953 extern int ceph_renew_caps(struct inode *inode); 954 extern int ceph_open(struct inode *inode, struct file *file); 955 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 956 struct file *file, unsigned flags, umode_t mode, 957 int *opened); 958 extern int ceph_release(struct inode *inode, struct file *filp); 959 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 960 char *data, size_t len); 961 962 /* dir.c */ 963 extern const struct file_operations ceph_dir_fops; 964 extern const struct file_operations ceph_snapdir_fops; 965 extern const struct inode_operations ceph_dir_iops; 966 extern const struct inode_operations ceph_snapdir_iops; 967 extern const struct dentry_operations ceph_dentry_ops; 968 969 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 970 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 971 extern int ceph_handle_snapdir(struct ceph_mds_request *req, 972 struct dentry *dentry, int err); 973 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 974 struct dentry *dentry, int err); 975 976 extern void ceph_dentry_lru_add(struct dentry *dn); 977 extern void ceph_dentry_lru_touch(struct dentry *dn); 978 extern void ceph_dentry_lru_del(struct dentry *dn); 979 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 980 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 981 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 982 983 /* ioctl.c */ 984 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 985 986 /* export.c */ 987 extern const struct export_operations ceph_export_ops; 988 989 /* locks.c */ 990 extern __init void ceph_flock_init(void); 991 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 992 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 993 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 994 extern int ceph_encode_locks_to_buffer(struct inode *inode, 995 struct ceph_filelock *flocks, 996 int num_fcntl_locks, 997 int num_flock_locks); 998 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 999 struct ceph_pagelist *pagelist, 1000 int num_fcntl_locks, int num_flock_locks); 1001 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); 1002 1003 /* debugfs.c */ 1004 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); 1005 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1006 1007 #endif /* _FS_CEPH_SUPER_H */ 1008