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