1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _FS_CEPH_SUPER_H 3 #define _FS_CEPH_SUPER_H 4 5 #include <linux/ceph/ceph_debug.h> 6 7 #include <asm/unaligned.h> 8 #include <linux/backing-dev.h> 9 #include <linux/completion.h> 10 #include <linux/exportfs.h> 11 #include <linux/fs.h> 12 #include <linux/mempool.h> 13 #include <linux/pagemap.h> 14 #include <linux/wait.h> 15 #include <linux/writeback.h> 16 #include <linux/slab.h> 17 #include <linux/posix_acl.h> 18 #include <linux/refcount.h> 19 #include <linux/security.h> 20 21 #include <linux/ceph/libceph.h> 22 23 #ifdef CONFIG_CEPH_FSCACHE 24 #include <linux/fscache.h> 25 #endif 26 27 /* f_type in struct statfs */ 28 #define CEPH_SUPER_MAGIC 0x00c36400 29 30 /* large granularity for statfs utilization stats to facilitate 31 * large volume sizes on 32-bit machines. */ 32 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ 33 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 34 35 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blacklisted */ 36 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ 37 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ 38 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ 39 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ 40 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ 41 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ 42 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */ 43 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */ 44 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */ 45 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */ 46 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */ 47 48 #define CEPH_MOUNT_OPT_DEFAULT \ 49 (CEPH_MOUNT_OPT_DCACHE | \ 50 CEPH_MOUNT_OPT_NOCOPYFROM) 51 52 #define ceph_set_mount_opt(fsc, opt) \ 53 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt 54 #define ceph_clear_mount_opt(fsc, opt) \ 55 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt 56 #define ceph_test_mount_opt(fsc, opt) \ 57 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 58 59 /* max size of osd read request, limited by libceph */ 60 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN 61 /* osd has a configurable limitaion of max write size. 62 * CEPH_MSG_MAX_DATA_LEN should be small enough. */ 63 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN 64 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */ 65 #define CEPH_MAX_READDIR_DEFAULT 1024 66 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 67 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 68 69 /* 70 * Delay telling the MDS we no longer want caps, in case we reopen 71 * the file. Delay a minimum amount of time, even if we send a cap 72 * message for some other reason. Otherwise, take the oppotunity to 73 * update the mds to avoid sending another message later. 74 */ 75 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ 76 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ 77 78 struct ceph_mount_options { 79 unsigned int flags; 80 81 unsigned int wsize; /* max write size */ 82 unsigned int rsize; /* max read size */ 83 unsigned int rasize; /* max readahead */ 84 unsigned int congestion_kb; /* max writeback in flight */ 85 unsigned int caps_wanted_delay_min, caps_wanted_delay_max; 86 int caps_max; 87 unsigned int max_readdir; /* max readdir result (entries) */ 88 unsigned int max_readdir_bytes; /* max readdir result (bytes) */ 89 90 /* 91 * everything above this point can be memcmp'd; everything below 92 * is handled in compare_mount_options() 93 */ 94 95 char *snapdir_name; /* default ".snap" */ 96 char *mds_namespace; /* default NULL */ 97 char *server_path; /* default NULL (means "/") */ 98 char *fscache_uniq; /* default NULL */ 99 }; 100 101 struct ceph_fs_client { 102 struct super_block *sb; 103 104 struct ceph_mount_options *mount_options; 105 struct ceph_client *client; 106 107 unsigned long mount_state; 108 109 unsigned long last_auto_reconnect; 110 bool blacklisted; 111 112 bool have_copy_from2; 113 114 u32 filp_gen; 115 loff_t max_file_size; 116 117 struct ceph_mds_client *mdsc; 118 119 /* writeback */ 120 mempool_t *wb_pagevec_pool; 121 atomic_long_t writeback_count; 122 123 struct workqueue_struct *inode_wq; 124 struct workqueue_struct *cap_wq; 125 126 #ifdef CONFIG_DEBUG_FS 127 struct dentry *debugfs_dentry_lru, *debugfs_caps; 128 struct dentry *debugfs_congestion_kb; 129 struct dentry *debugfs_bdi; 130 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 131 struct dentry *debugfs_mds_sessions; 132 #endif 133 134 #ifdef CONFIG_CEPH_FSCACHE 135 struct fscache_cookie *fscache; 136 #endif 137 }; 138 139 140 /* 141 * File i/o capability. This tracks shared state with the metadata 142 * server that allows us to cache or writeback attributes or to read 143 * and write data. For any given inode, we should have one or more 144 * capabilities, one issued by each metadata server, and our 145 * cumulative access is the OR of all issued capabilities. 146 * 147 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 148 * session capability lists. 149 */ 150 struct ceph_cap { 151 struct ceph_inode_info *ci; 152 struct rb_node ci_node; /* per-ci cap tree */ 153 struct ceph_mds_session *session; 154 struct list_head session_caps; /* per-session caplist */ 155 u64 cap_id; /* unique cap id (mds provided) */ 156 union { 157 /* in-use caps */ 158 struct { 159 int issued; /* latest, from the mds */ 160 int implemented; /* implemented superset of 161 issued (for revocation) */ 162 int mds, mds_wanted; 163 }; 164 /* caps to release */ 165 struct { 166 u64 cap_ino; 167 int queue_release; 168 }; 169 }; 170 u32 seq, issue_seq, mseq; 171 u32 cap_gen; /* active/stale cycle */ 172 unsigned long last_used; 173 struct list_head caps_item; 174 }; 175 176 #define CHECK_CAPS_NODELAY 1 /* do not delay any further */ 177 #define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ 178 #define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ 179 180 struct ceph_cap_flush { 181 u64 tid; 182 int caps; /* 0 means capsnap */ 183 bool wake; /* wake up flush waiters when finish ? */ 184 struct list_head g_list; // global 185 struct list_head i_list; // per inode 186 }; 187 188 /* 189 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 190 * we first complete any in-process sync writes and writeback any dirty 191 * data before flushing the snapped state (tracked here) back to the MDS. 192 */ 193 struct ceph_cap_snap { 194 refcount_t nref; 195 struct list_head ci_item; 196 197 struct ceph_cap_flush cap_flush; 198 199 u64 follows; 200 int issued, dirty; 201 struct ceph_snap_context *context; 202 203 umode_t mode; 204 kuid_t uid; 205 kgid_t gid; 206 207 struct ceph_buffer *xattr_blob; 208 u64 xattr_version; 209 210 u64 size; 211 u64 change_attr; 212 struct timespec64 mtime, atime, ctime, btime; 213 u64 time_warp_seq; 214 u64 truncate_size; 215 u32 truncate_seq; 216 int writing; /* a sync write is still in progress */ 217 int dirty_pages; /* dirty pages awaiting writeback */ 218 bool inline_data; 219 bool need_flush; 220 }; 221 222 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 223 { 224 if (refcount_dec_and_test(&capsnap->nref)) { 225 if (capsnap->xattr_blob) 226 ceph_buffer_put(capsnap->xattr_blob); 227 kfree(capsnap); 228 } 229 } 230 231 /* 232 * The frag tree describes how a directory is fragmented, potentially across 233 * multiple metadata servers. It is also used to indicate points where 234 * metadata authority is delegated, and whether/where metadata is replicated. 235 * 236 * A _leaf_ frag will be present in the i_fragtree IFF there is 237 * delegation info. That is, if mds >= 0 || ndist > 0. 238 */ 239 #define CEPH_MAX_DIRFRAG_REP 4 240 241 struct ceph_inode_frag { 242 struct rb_node node; 243 244 /* fragtree state */ 245 u32 frag; 246 int split_by; /* i.e. 2^(split_by) children */ 247 248 /* delegation and replication info */ 249 int mds; /* -1 if same authority as parent */ 250 int ndist; /* >0 if replicated */ 251 int dist[CEPH_MAX_DIRFRAG_REP]; 252 }; 253 254 /* 255 * We cache inode xattrs as an encoded blob until they are first used, 256 * at which point we parse them into an rbtree. 257 */ 258 struct ceph_inode_xattr { 259 struct rb_node node; 260 261 const char *name; 262 int name_len; 263 const char *val; 264 int val_len; 265 int dirty; 266 267 int should_free_name; 268 int should_free_val; 269 }; 270 271 /* 272 * Ceph dentry state 273 */ 274 struct ceph_dentry_info { 275 struct dentry *dentry; 276 struct ceph_mds_session *lease_session; 277 struct list_head lease_list; 278 unsigned flags; 279 int lease_shared_gen; 280 u32 lease_gen; 281 u32 lease_seq; 282 unsigned long lease_renew_after, lease_renew_from; 283 unsigned long time; 284 u64 offset; 285 }; 286 287 #define CEPH_DENTRY_REFERENCED 1 288 #define CEPH_DENTRY_LEASE_LIST 2 289 #define CEPH_DENTRY_SHRINK_LIST 4 290 #define CEPH_DENTRY_PRIMARY_LINK 8 291 292 struct ceph_inode_xattrs_info { 293 /* 294 * (still encoded) xattr blob. we avoid the overhead of parsing 295 * this until someone actually calls getxattr, etc. 296 * 297 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 298 * NULL means we don't know. 299 */ 300 struct ceph_buffer *blob, *prealloc_blob; 301 302 struct rb_root index; 303 bool dirty; 304 int count; 305 int names_size; 306 int vals_size; 307 u64 version, index_version; 308 }; 309 310 /* 311 * Ceph inode. 312 */ 313 struct ceph_inode_info { 314 struct ceph_vino i_vino; /* ceph ino + snap */ 315 316 spinlock_t i_ceph_lock; 317 318 u64 i_version; 319 u64 i_inline_version; 320 u32 i_time_warp_seq; 321 322 unsigned long i_ceph_flags; 323 atomic64_t i_release_count; 324 atomic64_t i_ordered_count; 325 atomic64_t i_complete_seq[2]; 326 327 struct ceph_dir_layout i_dir_layout; 328 struct ceph_file_layout i_layout; 329 char *i_symlink; 330 331 /* for dirs */ 332 struct timespec64 i_rctime; 333 u64 i_rbytes, i_rfiles, i_rsubdirs; 334 u64 i_files, i_subdirs; 335 336 /* quotas */ 337 u64 i_max_bytes, i_max_files; 338 339 s32 i_dir_pin; 340 341 struct rb_root i_fragtree; 342 int i_fragtree_nsplits; 343 struct mutex i_fragtree_mutex; 344 345 struct ceph_inode_xattrs_info i_xattrs; 346 347 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 348 * s_mutex. */ 349 struct rb_root i_caps; /* cap list */ 350 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 351 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 352 struct list_head i_dirty_item, i_flushing_item; 353 /* we need to track cap writeback on a per-cap-bit basis, to allow 354 * overlapping, pipelined cap flushes to the mds. we can probably 355 * reduce the tid to 8 bits if we're concerned about inode size. */ 356 struct ceph_cap_flush *i_prealloc_cap_flush; 357 struct list_head i_cap_flush_list; 358 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 359 unsigned long i_hold_caps_min; /* jiffies */ 360 unsigned long i_hold_caps_max; /* jiffies */ 361 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 362 struct ceph_cap_reservation i_cap_migration_resv; 363 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 364 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 365 dirty|flushing caps */ 366 unsigned i_snap_caps; /* cap bits for snapped files */ 367 368 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ 369 370 struct mutex i_truncate_mutex; 371 u32 i_truncate_seq; /* last truncate to smaller size */ 372 u64 i_truncate_size; /* and the size we last truncated down to */ 373 int i_truncate_pending; /* still need to call vmtruncate */ 374 375 u64 i_max_size; /* max file size authorized by mds */ 376 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 377 u64 i_wanted_max_size; /* offset we'd like to write too */ 378 u64 i_requested_max_size; /* max_size we've requested */ 379 380 /* held references to caps */ 381 int i_pin_ref; 382 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; 383 int i_wrbuffer_ref, i_wrbuffer_ref_head; 384 atomic_t i_filelock_ref; 385 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ 386 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 387 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 388 389 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 390 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 391 spinlock_t i_unsafe_lock; 392 393 union { 394 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 395 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ 396 }; 397 int i_snap_realm_counter; /* snap realm (if caps) */ 398 struct list_head i_snap_realm_item; 399 struct list_head i_snap_flush_item; 400 struct timespec64 i_btime; 401 struct timespec64 i_snap_btime; 402 403 struct work_struct i_work; 404 unsigned long i_work_mask; 405 406 #ifdef CONFIG_CEPH_FSCACHE 407 struct fscache_cookie *fscache; 408 u32 i_fscache_gen; 409 #endif 410 errseq_t i_meta_err; 411 412 struct inode vfs_inode; /* at end */ 413 }; 414 415 static inline struct ceph_inode_info * 416 ceph_inode(const struct inode *inode) 417 { 418 return container_of(inode, struct ceph_inode_info, vfs_inode); 419 } 420 421 static inline struct ceph_fs_client * 422 ceph_inode_to_client(const struct inode *inode) 423 { 424 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 425 } 426 427 static inline struct ceph_fs_client * 428 ceph_sb_to_client(const struct super_block *sb) 429 { 430 return (struct ceph_fs_client *)sb->s_fs_info; 431 } 432 433 static inline struct ceph_vino 434 ceph_vino(const struct inode *inode) 435 { 436 return ceph_inode(inode)->i_vino; 437 } 438 439 /* 440 * ino_t is <64 bits on many architectures, blech. 441 * 442 * i_ino (kernel inode) st_ino (userspace) 443 * i386 32 32 444 * x86_64+ino32 64 32 445 * x86_64 64 64 446 */ 447 static inline u32 ceph_ino_to_ino32(__u64 vino) 448 { 449 u32 ino = vino & 0xffffffff; 450 ino ^= vino >> 32; 451 if (!ino) 452 ino = 2; 453 return ino; 454 } 455 456 /* 457 * kernel i_ino value 458 */ 459 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) 460 { 461 #if BITS_PER_LONG == 32 462 return ceph_ino_to_ino32(vino.ino); 463 #else 464 return (ino_t)vino.ino; 465 #endif 466 } 467 468 /* 469 * user-visible ino (stat, filldir) 470 */ 471 #if BITS_PER_LONG == 32 472 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 473 { 474 return ino; 475 } 476 #else 477 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 478 { 479 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) 480 ino = ceph_ino_to_ino32(ino); 481 return ino; 482 } 483 #endif 484 485 486 /* for printf-style formatting */ 487 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 488 489 static inline u64 ceph_ino(struct inode *inode) 490 { 491 return ceph_inode(inode)->i_vino.ino; 492 } 493 static inline u64 ceph_snap(struct inode *inode) 494 { 495 return ceph_inode(inode)->i_vino.snap; 496 } 497 498 static inline int ceph_ino_compare(struct inode *inode, void *data) 499 { 500 struct ceph_vino *pvino = (struct ceph_vino *)data; 501 struct ceph_inode_info *ci = ceph_inode(inode); 502 return ci->i_vino.ino == pvino->ino && 503 ci->i_vino.snap == pvino->snap; 504 } 505 506 static inline struct inode *ceph_find_inode(struct super_block *sb, 507 struct ceph_vino vino) 508 { 509 ino_t t = ceph_vino_to_ino(vino); 510 return ilookup5(sb, t, ceph_ino_compare, &vino); 511 } 512 513 514 /* 515 * Ceph inode. 516 */ 517 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 518 #define CEPH_I_NODELAY (1 << 1) /* do not delay cap release */ 519 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 520 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */ 521 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */ 522 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */ 523 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */ 524 #define CEPH_I_CAP_DROPPED (1 << 7) /* caps were forcibly dropped */ 525 #define CEPH_I_KICK_FLUSH (1 << 8) /* kick flushing caps */ 526 #define CEPH_I_FLUSH_SNAPS (1 << 9) /* need flush snapss */ 527 #define CEPH_I_ERROR_WRITE (1 << 10) /* have seen write errors */ 528 #define CEPH_I_ERROR_FILELOCK (1 << 11) /* have seen file lock errors */ 529 #define CEPH_I_ODIRECT (1 << 12) /* inode in direct I/O mode */ 530 #define CEPH_ASYNC_CREATE_BIT (13) /* async create in flight for this */ 531 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT) 532 533 /* 534 * Masks of ceph inode work. 535 */ 536 #define CEPH_I_WORK_WRITEBACK 0 /* writeback */ 537 #define CEPH_I_WORK_INVALIDATE_PAGES 1 /* invalidate pages */ 538 #define CEPH_I_WORK_VMTRUNCATE 2 /* vmtruncate */ 539 540 /* 541 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 542 * and then clear it when they start succeeding. Note that we do a lockless 543 * check first, and only take the lock if it looks like it needs to be changed. 544 * The write submission code just takes this as a hint, so we're not too 545 * worried if a few slip through in either direction. 546 */ 547 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 548 { 549 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { 550 spin_lock(&ci->i_ceph_lock); 551 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; 552 spin_unlock(&ci->i_ceph_lock); 553 } 554 } 555 556 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 557 { 558 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { 559 spin_lock(&ci->i_ceph_lock); 560 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; 561 spin_unlock(&ci->i_ceph_lock); 562 } 563 } 564 565 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 566 long long release_count, 567 long long ordered_count) 568 { 569 /* 570 * Makes sure operations that setup readdir cache (update page 571 * cache and i_size) are strongly ordered w.r.t. the following 572 * atomic64_set() operations. 573 */ 574 smp_mb(); 575 atomic64_set(&ci->i_complete_seq[0], release_count); 576 atomic64_set(&ci->i_complete_seq[1], ordered_count); 577 } 578 579 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 580 { 581 atomic64_inc(&ci->i_release_count); 582 } 583 584 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 585 { 586 atomic64_inc(&ci->i_ordered_count); 587 } 588 589 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 590 { 591 return atomic64_read(&ci->i_complete_seq[0]) == 592 atomic64_read(&ci->i_release_count); 593 } 594 595 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 596 { 597 return atomic64_read(&ci->i_complete_seq[0]) == 598 atomic64_read(&ci->i_release_count) && 599 atomic64_read(&ci->i_complete_seq[1]) == 600 atomic64_read(&ci->i_ordered_count); 601 } 602 603 static inline void ceph_dir_clear_complete(struct inode *inode) 604 { 605 __ceph_dir_clear_complete(ceph_inode(inode)); 606 } 607 608 static inline void ceph_dir_clear_ordered(struct inode *inode) 609 { 610 __ceph_dir_clear_ordered(ceph_inode(inode)); 611 } 612 613 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 614 { 615 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 616 smp_rmb(); 617 return ret; 618 } 619 620 /* find a specific frag @f */ 621 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 622 u32 f); 623 624 /* 625 * choose fragment for value @v. copy frag content to pfrag, if leaf 626 * exists 627 */ 628 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 629 struct ceph_inode_frag *pfrag, 630 int *found); 631 632 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) 633 { 634 return (struct ceph_dentry_info *)dentry->d_fsdata; 635 } 636 637 /* 638 * caps helpers 639 */ 640 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 641 { 642 return !RB_EMPTY_ROOT(&ci->i_caps); 643 } 644 645 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 646 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 647 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 648 struct ceph_cap *cap); 649 650 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 651 { 652 int issued; 653 spin_lock(&ci->i_ceph_lock); 654 issued = __ceph_caps_issued(ci, NULL); 655 spin_unlock(&ci->i_ceph_lock); 656 return issued; 657 } 658 659 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, 660 int touch) 661 { 662 int r; 663 spin_lock(&ci->i_ceph_lock); 664 r = __ceph_caps_issued_mask(ci, mask, touch); 665 spin_unlock(&ci->i_ceph_lock); 666 return r; 667 } 668 669 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 670 { 671 return ci->i_dirty_caps | ci->i_flushing_caps; 672 } 673 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 674 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 675 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 676 struct ceph_cap_flush **pcf); 677 678 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 679 struct ceph_cap *ocap, int mask); 680 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 681 extern int __ceph_caps_used(struct ceph_inode_info *ci); 682 683 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 684 extern int __ceph_caps_wanted(struct ceph_inode_info *ci); 685 686 /* what the mds thinks we want */ 687 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 688 689 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 690 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 691 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, 692 struct ceph_mount_options *fsopt); 693 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 694 struct ceph_cap_reservation *ctx, int need); 695 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, 696 struct ceph_cap_reservation *ctx); 697 extern void ceph_reservation_status(struct ceph_fs_client *client, 698 int *total, int *avail, int *used, 699 int *reserved, int *min); 700 701 702 703 /* 704 * we keep buffered readdir results attached to file->private_data 705 */ 706 #define CEPH_F_SYNC 1 707 #define CEPH_F_ATEND 2 708 709 struct ceph_file_info { 710 short fmode; /* initialized on open */ 711 short flags; /* CEPH_F_* */ 712 713 spinlock_t rw_contexts_lock; 714 struct list_head rw_contexts; 715 716 errseq_t meta_err; 717 u32 filp_gen; 718 atomic_t num_locks; 719 }; 720 721 struct ceph_dir_file_info { 722 struct ceph_file_info file_info; 723 724 /* readdir: position within the dir */ 725 u32 frag; 726 struct ceph_mds_request *last_readdir; 727 728 /* readdir: position within a frag */ 729 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 730 char *last_name; /* last entry in previous chunk */ 731 long long dir_release_count; 732 long long dir_ordered_count; 733 int readdir_cache_idx; 734 735 /* used for -o dirstat read() on directory thing */ 736 char *dir_info; 737 int dir_info_len; 738 }; 739 740 struct ceph_rw_context { 741 struct list_head list; 742 struct task_struct *thread; 743 int caps; 744 }; 745 746 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ 747 struct ceph_rw_context _name = { \ 748 .thread = current, \ 749 .caps = _caps, \ 750 } 751 752 static inline void ceph_add_rw_context(struct ceph_file_info *cf, 753 struct ceph_rw_context *ctx) 754 { 755 spin_lock(&cf->rw_contexts_lock); 756 list_add(&ctx->list, &cf->rw_contexts); 757 spin_unlock(&cf->rw_contexts_lock); 758 } 759 760 static inline void ceph_del_rw_context(struct ceph_file_info *cf, 761 struct ceph_rw_context *ctx) 762 { 763 spin_lock(&cf->rw_contexts_lock); 764 list_del(&ctx->list); 765 spin_unlock(&cf->rw_contexts_lock); 766 } 767 768 static inline struct ceph_rw_context* 769 ceph_find_rw_context(struct ceph_file_info *cf) 770 { 771 struct ceph_rw_context *ctx, *found = NULL; 772 spin_lock(&cf->rw_contexts_lock); 773 list_for_each_entry(ctx, &cf->rw_contexts, list) { 774 if (ctx->thread == current) { 775 found = ctx; 776 break; 777 } 778 } 779 spin_unlock(&cf->rw_contexts_lock); 780 return found; 781 } 782 783 struct ceph_readdir_cache_control { 784 struct page *page; 785 struct dentry **dentries; 786 int index; 787 }; 788 789 /* 790 * A "snap realm" describes a subset of the file hierarchy sharing 791 * the same set of snapshots that apply to it. The realms themselves 792 * are organized into a hierarchy, such that children inherit (some of) 793 * the snapshots of their parents. 794 * 795 * All inodes within the realm that have capabilities are linked into a 796 * per-realm list. 797 */ 798 struct ceph_snap_realm { 799 u64 ino; 800 struct inode *inode; 801 atomic_t nref; 802 struct rb_node node; 803 804 u64 created, seq; 805 u64 parent_ino; 806 u64 parent_since; /* snapid when our current parent became so */ 807 808 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 809 u32 num_prior_parent_snaps; /* had prior to parent_since */ 810 u64 *snaps; /* snaps specific to this realm */ 811 u32 num_snaps; 812 813 struct ceph_snap_realm *parent; 814 struct list_head children; /* list of child realms */ 815 struct list_head child_item; 816 817 struct list_head empty_item; /* if i have ref==0 */ 818 819 struct list_head dirty_item; /* if realm needs new context */ 820 821 /* the current set of snaps for this realm */ 822 struct ceph_snap_context *cached_context; 823 824 struct list_head inodes_with_caps; 825 spinlock_t inodes_with_caps_lock; 826 }; 827 828 static inline int default_congestion_kb(void) 829 { 830 int congestion_kb; 831 832 /* 833 * Copied from NFS 834 * 835 * congestion size, scale with available memory. 836 * 837 * 64MB: 8192k 838 * 128MB: 11585k 839 * 256MB: 16384k 840 * 512MB: 23170k 841 * 1GB: 32768k 842 * 2GB: 46340k 843 * 4GB: 65536k 844 * 8GB: 92681k 845 * 16GB: 131072k 846 * 847 * This allows larger machines to have larger/more transfers. 848 * Limit the default to 256M 849 */ 850 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); 851 if (congestion_kb > 256*1024) 852 congestion_kb = 256*1024; 853 854 return congestion_kb; 855 } 856 857 858 /* super.c */ 859 extern int ceph_force_reconnect(struct super_block *sb); 860 /* snap.c */ 861 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 862 u64 ino); 863 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 864 struct ceph_snap_realm *realm); 865 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 866 struct ceph_snap_realm *realm); 867 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 868 void *p, void *e, bool deletion, 869 struct ceph_snap_realm **realm_ret); 870 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 871 struct ceph_mds_session *session, 872 struct ceph_msg *msg); 873 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); 874 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 875 struct ceph_cap_snap *capsnap); 876 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 877 878 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, 879 u64 snap); 880 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, 881 struct ceph_snapid_map *sm); 882 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); 883 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); 884 885 886 /* 887 * a cap_snap is "pending" if it is still awaiting an in-progress 888 * sync write (that may/may not still update size, mtime, etc.). 889 */ 890 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 891 { 892 return !list_empty(&ci->i_cap_snaps) && 893 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 894 ci_item)->writing; 895 } 896 897 /* inode.c */ 898 extern const struct inode_operations ceph_file_iops; 899 900 extern struct inode *ceph_alloc_inode(struct super_block *sb); 901 extern void ceph_evict_inode(struct inode *inode); 902 extern void ceph_free_inode(struct inode *inode); 903 904 extern struct inode *ceph_get_inode(struct super_block *sb, 905 struct ceph_vino vino); 906 extern struct inode *ceph_get_snapdir(struct inode *parent); 907 extern int ceph_fill_file_size(struct inode *inode, int issued, 908 u32 truncate_seq, u64 truncate_size, u64 size); 909 extern void ceph_fill_file_time(struct inode *inode, int issued, 910 u64 time_warp_seq, struct timespec64 *ctime, 911 struct timespec64 *mtime, 912 struct timespec64 *atime); 913 extern int ceph_fill_trace(struct super_block *sb, 914 struct ceph_mds_request *req); 915 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 916 struct ceph_mds_session *session); 917 918 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 919 920 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 921 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 922 extern void ceph_queue_vmtruncate(struct inode *inode); 923 extern void ceph_queue_invalidate(struct inode *inode); 924 extern void ceph_queue_writeback(struct inode *inode); 925 extern void ceph_async_iput(struct inode *inode); 926 927 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 928 int mask, bool force); 929 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 930 { 931 return __ceph_do_getattr(inode, NULL, mask, force); 932 } 933 extern int ceph_permission(struct inode *inode, int mask); 934 extern int __ceph_setattr(struct inode *inode, struct iattr *attr); 935 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); 936 extern int ceph_getattr(const struct path *path, struct kstat *stat, 937 u32 request_mask, unsigned int flags); 938 939 /* xattr.c */ 940 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 941 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 942 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 943 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); 944 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 945 extern const struct xattr_handler *ceph_xattr_handlers[]; 946 947 struct ceph_acl_sec_ctx { 948 #ifdef CONFIG_CEPH_FS_POSIX_ACL 949 void *default_acl; 950 void *acl; 951 #endif 952 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 953 void *sec_ctx; 954 u32 sec_ctxlen; 955 #endif 956 struct ceph_pagelist *pagelist; 957 }; 958 959 #ifdef CONFIG_SECURITY 960 extern bool ceph_security_xattr_deadlock(struct inode *in); 961 extern bool ceph_security_xattr_wanted(struct inode *in); 962 #else 963 static inline bool ceph_security_xattr_deadlock(struct inode *in) 964 { 965 return false; 966 } 967 static inline bool ceph_security_xattr_wanted(struct inode *in) 968 { 969 return false; 970 } 971 #endif 972 973 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 974 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 975 struct ceph_acl_sec_ctx *ctx); 976 static inline void ceph_security_invalidate_secctx(struct inode *inode) 977 { 978 security_inode_invalidate_secctx(inode); 979 } 980 #else 981 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 982 struct ceph_acl_sec_ctx *ctx) 983 { 984 return 0; 985 } 986 static inline void ceph_security_invalidate_secctx(struct inode *inode) 987 { 988 } 989 #endif 990 991 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); 992 993 /* acl.c */ 994 #ifdef CONFIG_CEPH_FS_POSIX_ACL 995 996 struct posix_acl *ceph_get_acl(struct inode *, int); 997 int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); 998 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 999 struct ceph_acl_sec_ctx *as_ctx); 1000 void ceph_init_inode_acls(struct inode *inode, 1001 struct ceph_acl_sec_ctx *as_ctx); 1002 1003 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1004 { 1005 forget_all_cached_acls(inode); 1006 } 1007 1008 #else 1009 1010 #define ceph_get_acl NULL 1011 #define ceph_set_acl NULL 1012 1013 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1014 struct ceph_acl_sec_ctx *as_ctx) 1015 { 1016 return 0; 1017 } 1018 static inline void ceph_init_inode_acls(struct inode *inode, 1019 struct ceph_acl_sec_ctx *as_ctx) 1020 { 1021 } 1022 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 1023 { 1024 return 0; 1025 } 1026 1027 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1028 { 1029 } 1030 1031 #endif 1032 1033 /* caps.c */ 1034 extern const char *ceph_cap_string(int c); 1035 extern void ceph_handle_caps(struct ceph_mds_session *session, 1036 struct ceph_msg *msg); 1037 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 1038 struct ceph_cap_reservation *ctx); 1039 extern void ceph_add_cap(struct inode *inode, 1040 struct ceph_mds_session *session, u64 cap_id, 1041 int fmode, unsigned issued, unsigned wanted, 1042 unsigned cap, unsigned seq, u64 realmino, int flags, 1043 struct ceph_cap **new_cap); 1044 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1045 extern void __ceph_remove_caps(struct ceph_inode_info *ci); 1046 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 1047 struct ceph_cap *cap); 1048 extern int ceph_is_any_caps(struct inode *inode); 1049 1050 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 1051 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 1052 int datasync); 1053 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 1054 struct ceph_mds_session *session); 1055 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 1056 struct ceph_mds_session *session); 1057 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, 1058 struct ceph_inode_info *ci); 1059 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 1060 int mds); 1061 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, 1062 bool snap_rwsem_locked); 1063 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 1064 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 1065 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 1066 struct ceph_snap_context *snapc); 1067 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 1068 struct ceph_mds_session **psession); 1069 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 1070 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 1071 struct ceph_mds_session *session); 1072 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 1073 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 1074 extern int ceph_drop_caps_for_unlink(struct inode *inode); 1075 extern int ceph_encode_inode_release(void **p, struct inode *inode, 1076 int mds, int drop, int unless, int force); 1077 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 1078 struct inode *dir, 1079 int mds, int drop, int unless); 1080 1081 extern int ceph_get_caps(struct file *filp, int need, int want, 1082 loff_t endoff, int *got, struct page **pinned_page); 1083 extern int ceph_try_get_caps(struct inode *inode, 1084 int need, int want, bool nonblock, int *got); 1085 1086 /* for counting open files by mode */ 1087 extern void __ceph_get_fmode(struct ceph_inode_info *ci, int mode); 1088 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); 1089 1090 /* addr.c */ 1091 extern const struct address_space_operations ceph_aops; 1092 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 1093 extern int ceph_uninline_data(struct file *filp, struct page *locked_page); 1094 extern int ceph_pool_perm_check(struct inode *inode, int need); 1095 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 1096 1097 /* file.c */ 1098 extern const struct file_operations ceph_file_fops; 1099 1100 extern int ceph_renew_caps(struct inode *inode); 1101 extern int ceph_open(struct inode *inode, struct file *file); 1102 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 1103 struct file *file, unsigned flags, umode_t mode); 1104 extern int ceph_release(struct inode *inode, struct file *filp); 1105 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 1106 char *data, size_t len); 1107 1108 /* dir.c */ 1109 extern const struct file_operations ceph_dir_fops; 1110 extern const struct file_operations ceph_snapdir_fops; 1111 extern const struct inode_operations ceph_dir_iops; 1112 extern const struct inode_operations ceph_snapdir_iops; 1113 extern const struct dentry_operations ceph_dentry_ops; 1114 1115 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 1116 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 1117 extern int ceph_handle_snapdir(struct ceph_mds_request *req, 1118 struct dentry *dentry, int err); 1119 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 1120 struct dentry *dentry, int err); 1121 1122 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); 1123 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); 1124 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 1125 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); 1126 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 1127 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 1128 1129 /* ioctl.c */ 1130 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1131 1132 /* export.c */ 1133 extern const struct export_operations ceph_export_ops; 1134 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); 1135 1136 /* locks.c */ 1137 extern __init void ceph_flock_init(void); 1138 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 1139 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 1140 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 1141 extern int ceph_encode_locks_to_buffer(struct inode *inode, 1142 struct ceph_filelock *flocks, 1143 int num_fcntl_locks, 1144 int num_flock_locks); 1145 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 1146 struct ceph_pagelist *pagelist, 1147 int num_fcntl_locks, int num_flock_locks); 1148 1149 /* debugfs.c */ 1150 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); 1151 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1152 1153 /* quota.c */ 1154 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci) 1155 { 1156 return ci->i_max_files || ci->i_max_bytes; 1157 } 1158 1159 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); 1160 1161 static inline void __ceph_update_quota(struct ceph_inode_info *ci, 1162 u64 max_bytes, u64 max_files) 1163 { 1164 bool had_quota, has_quota; 1165 had_quota = __ceph_has_any_quota(ci); 1166 ci->i_max_bytes = max_bytes; 1167 ci->i_max_files = max_files; 1168 has_quota = __ceph_has_any_quota(ci); 1169 1170 if (had_quota != has_quota) 1171 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota); 1172 } 1173 1174 extern void ceph_handle_quota(struct ceph_mds_client *mdsc, 1175 struct ceph_mds_session *session, 1176 struct ceph_msg *msg); 1177 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); 1178 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); 1179 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, 1180 loff_t newlen); 1181 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, 1182 loff_t newlen); 1183 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, 1184 struct kstatfs *buf); 1185 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); 1186 1187 #endif /* _FS_CEPH_SUPER_H */ 1188