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