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