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