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 /* large granularity for statfs utilization stats to facilitate 28 * large volume sizes on 32-bit machines. */ 29 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ 30 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 31 #define CEPH_4K_BLOCK_SHIFT 12 /* 4 KB */ 32 33 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */ 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 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */ 45 #define CEPH_MOUNT_OPT_NOPAGECACHE (1<<16) /* bypass pagecache altogether */ 46 47 #define CEPH_MOUNT_OPT_DEFAULT \ 48 (CEPH_MOUNT_OPT_DCACHE | \ 49 CEPH_MOUNT_OPT_NOCOPYFROM | \ 50 CEPH_MOUNT_OPT_ASYNC_DIROPS) 51 52 #define ceph_set_mount_opt(fsc, opt) \ 53 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt 54 #define ceph_clear_mount_opt(fsc, opt) \ 55 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt 56 #define ceph_test_mount_opt(fsc, opt) \ 57 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 58 59 /* max size of osd read request, limited by libceph */ 60 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN 61 /* osd has a configurable limitaion of max write size. 62 * CEPH_MSG_MAX_DATA_LEN should be small enough. */ 63 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN 64 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */ 65 #define CEPH_MAX_READDIR_DEFAULT 1024 66 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 67 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 68 69 /* 70 * Delay telling the MDS we no longer want caps, in case we reopen 71 * the file. Delay a minimum amount of time, even if we send a cap 72 * message for some other reason. Otherwise, take the oppotunity to 73 * update the mds to avoid sending another message later. 74 */ 75 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ 76 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ 77 78 struct ceph_mount_options { 79 unsigned int flags; 80 81 unsigned int wsize; /* max write size */ 82 unsigned int rsize; /* max read size */ 83 unsigned int rasize; /* max readahead */ 84 unsigned int congestion_kb; /* max writeback in flight */ 85 unsigned int caps_wanted_delay_min, caps_wanted_delay_max; 86 int caps_max; 87 unsigned int max_readdir; /* max readdir result (entries) */ 88 unsigned int max_readdir_bytes; /* max readdir result (bytes) */ 89 90 bool new_dev_syntax; 91 92 /* 93 * everything above this point can be memcmp'd; everything below 94 * is handled in compare_mount_options() 95 */ 96 97 char *snapdir_name; /* default ".snap" */ 98 char *mds_namespace; /* default NULL */ 99 char *server_path; /* default NULL (means "/") */ 100 char *fscache_uniq; /* default NULL */ 101 char *mon_addr; 102 }; 103 104 struct ceph_fs_client { 105 struct super_block *sb; 106 107 struct list_head metric_wakeup; 108 109 struct ceph_mount_options *mount_options; 110 struct ceph_client *client; 111 112 int mount_state; 113 114 bool blocklisted; 115 116 bool have_copy_from2; 117 118 u32 filp_gen; 119 loff_t max_file_size; 120 121 struct ceph_mds_client *mdsc; 122 123 atomic_long_t writeback_count; 124 125 struct workqueue_struct *inode_wq; 126 struct workqueue_struct *cap_wq; 127 128 #ifdef CONFIG_DEBUG_FS 129 struct dentry *debugfs_dentry_lru, *debugfs_caps; 130 struct dentry *debugfs_congestion_kb; 131 struct dentry *debugfs_bdi; 132 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 133 struct dentry *debugfs_status; 134 struct dentry *debugfs_mds_sessions; 135 struct dentry *debugfs_metrics_dir; 136 #endif 137 138 #ifdef CONFIG_CEPH_FSCACHE 139 struct fscache_volume *fscache; 140 #endif 141 }; 142 143 144 /* 145 * File i/o capability. This tracks shared state with the metadata 146 * server that allows us to cache or writeback attributes or to read 147 * and write data. For any given inode, we should have one or more 148 * capabilities, one issued by each metadata server, and our 149 * cumulative access is the OR of all issued capabilities. 150 * 151 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 152 * session capability lists. 153 */ 154 struct ceph_cap { 155 struct ceph_inode_info *ci; 156 struct rb_node ci_node; /* per-ci cap tree */ 157 struct ceph_mds_session *session; 158 struct list_head session_caps; /* per-session caplist */ 159 u64 cap_id; /* unique cap id (mds provided) */ 160 union { 161 /* in-use caps */ 162 struct { 163 int issued; /* latest, from the mds */ 164 int implemented; /* implemented superset of 165 issued (for revocation) */ 166 int mds; /* mds index for this cap */ 167 int mds_wanted; /* caps wanted from this mds */ 168 }; 169 /* caps to release */ 170 struct { 171 u64 cap_ino; 172 int queue_release; 173 }; 174 }; 175 u32 seq, issue_seq, mseq; 176 u32 cap_gen; /* active/stale cycle */ 177 unsigned long last_used; 178 struct list_head caps_item; 179 }; 180 181 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */ 182 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */ 183 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */ 184 185 struct ceph_cap_flush { 186 u64 tid; 187 int caps; 188 bool wake; /* wake up flush waiters when finish ? */ 189 bool is_capsnap; /* true means capsnap */ 190 struct list_head g_list; // global 191 struct list_head i_list; // per inode 192 }; 193 194 /* 195 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 196 * we first complete any in-process sync writes and writeback any dirty 197 * data before flushing the snapped state (tracked here) back to the MDS. 198 */ 199 struct ceph_cap_snap { 200 refcount_t nref; 201 struct list_head ci_item; 202 203 struct ceph_cap_flush cap_flush; 204 205 u64 follows; 206 int issued, dirty; 207 struct ceph_snap_context *context; 208 209 umode_t mode; 210 kuid_t uid; 211 kgid_t gid; 212 213 struct ceph_buffer *xattr_blob; 214 u64 xattr_version; 215 216 u64 size; 217 u64 change_attr; 218 struct timespec64 mtime, atime, ctime, btime; 219 u64 time_warp_seq; 220 u64 truncate_size; 221 u32 truncate_seq; 222 int writing; /* a sync write is still in progress */ 223 int dirty_pages; /* dirty pages awaiting writeback */ 224 bool inline_data; 225 bool need_flush; 226 }; 227 228 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 229 { 230 if (refcount_dec_and_test(&capsnap->nref)) { 231 if (capsnap->xattr_blob) 232 ceph_buffer_put(capsnap->xattr_blob); 233 kfree(capsnap); 234 } 235 } 236 237 /* 238 * The frag tree describes how a directory is fragmented, potentially across 239 * multiple metadata servers. It is also used to indicate points where 240 * metadata authority is delegated, and whether/where metadata is replicated. 241 * 242 * A _leaf_ frag will be present in the i_fragtree IFF there is 243 * delegation info. That is, if mds >= 0 || ndist > 0. 244 */ 245 #define CEPH_MAX_DIRFRAG_REP 4 246 247 struct ceph_inode_frag { 248 struct rb_node node; 249 250 /* fragtree state */ 251 u32 frag; 252 int split_by; /* i.e. 2^(split_by) children */ 253 254 /* delegation and replication info */ 255 int mds; /* -1 if same authority as parent */ 256 int ndist; /* >0 if replicated */ 257 int dist[CEPH_MAX_DIRFRAG_REP]; 258 }; 259 260 /* 261 * We cache inode xattrs as an encoded blob until they are first used, 262 * at which point we parse them into an rbtree. 263 */ 264 struct ceph_inode_xattr { 265 struct rb_node node; 266 267 const char *name; 268 int name_len; 269 const char *val; 270 int val_len; 271 int dirty; 272 273 int should_free_name; 274 int should_free_val; 275 }; 276 277 /* 278 * Ceph dentry state 279 */ 280 struct ceph_dentry_info { 281 struct dentry *dentry; 282 struct ceph_mds_session *lease_session; 283 struct list_head lease_list; 284 unsigned flags; 285 int lease_shared_gen; 286 u32 lease_gen; 287 u32 lease_seq; 288 unsigned long lease_renew_after, lease_renew_from; 289 unsigned long time; 290 u64 offset; 291 }; 292 293 #define CEPH_DENTRY_REFERENCED 1 294 #define CEPH_DENTRY_LEASE_LIST 2 295 #define CEPH_DENTRY_SHRINK_LIST 4 296 #define CEPH_DENTRY_PRIMARY_LINK 8 297 298 struct ceph_inode_xattrs_info { 299 /* 300 * (still encoded) xattr blob. we avoid the overhead of parsing 301 * this until someone actually calls getxattr, etc. 302 * 303 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 304 * NULL means we don't know. 305 */ 306 struct ceph_buffer *blob, *prealloc_blob; 307 308 struct rb_root index; 309 bool dirty; 310 int count; 311 int names_size; 312 int vals_size; 313 u64 version, index_version; 314 }; 315 316 /* 317 * Ceph inode. 318 */ 319 struct ceph_inode_info { 320 struct ceph_vino i_vino; /* ceph ino + snap */ 321 322 spinlock_t i_ceph_lock; 323 324 u64 i_version; 325 u64 i_inline_version; 326 u32 i_time_warp_seq; 327 328 unsigned long i_ceph_flags; 329 atomic64_t i_release_count; 330 atomic64_t i_ordered_count; 331 atomic64_t i_complete_seq[2]; 332 333 struct ceph_dir_layout i_dir_layout; 334 struct ceph_file_layout i_layout; 335 struct ceph_file_layout i_cached_layout; // for async creates 336 char *i_symlink; 337 338 /* for dirs */ 339 struct timespec64 i_rctime; 340 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps; 341 u64 i_files, i_subdirs; 342 343 /* quotas */ 344 u64 i_max_bytes, i_max_files; 345 346 s32 i_dir_pin; 347 348 struct rb_root i_fragtree; 349 int i_fragtree_nsplits; 350 struct mutex i_fragtree_mutex; 351 352 struct ceph_inode_xattrs_info i_xattrs; 353 354 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 355 * s_mutex. */ 356 struct rb_root i_caps; /* cap list */ 357 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 358 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 359 360 /* 361 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty 362 * is protected by the mdsc->cap_dirty_lock, but each individual item 363 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty 364 * requires the mdsc->cap_dirty_lock. List presence for an item can 365 * be tested under the i_ceph_lock. Changing anything requires both. 366 */ 367 struct list_head i_dirty_item; 368 369 /* 370 * Link to session's s_cap_flushing list. Protected in a similar 371 * fashion to i_dirty_item, but also by the s_mutex for changes. The 372 * s_cap_flushing list can be walked while holding either the s_mutex 373 * or msdc->cap_dirty_lock. List presence can also be checked while 374 * holding the i_ceph_lock for this inode. 375 */ 376 struct list_head i_flushing_item; 377 378 /* we need to track cap writeback on a per-cap-bit basis, to allow 379 * overlapping, pipelined cap flushes to the mds. we can probably 380 * reduce the tid to 8 bits if we're concerned about inode size. */ 381 struct ceph_cap_flush *i_prealloc_cap_flush; 382 struct list_head i_cap_flush_list; 383 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 384 unsigned long i_hold_caps_max; /* jiffies */ 385 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 386 struct ceph_cap_reservation i_cap_migration_resv; 387 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 388 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 389 dirty|flushing caps */ 390 unsigned i_snap_caps; /* cap bits for snapped files */ 391 392 unsigned long i_last_rd; 393 unsigned long i_last_wr; 394 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ 395 396 struct mutex i_truncate_mutex; 397 u32 i_truncate_seq; /* last truncate to smaller size */ 398 u64 i_truncate_size; /* and the size we last truncated down to */ 399 int i_truncate_pending; /* still need to call vmtruncate */ 400 401 u64 i_max_size; /* max file size authorized by mds */ 402 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 403 u64 i_wanted_max_size; /* offset we'd like to write too */ 404 u64 i_requested_max_size; /* max_size we've requested */ 405 406 /* held references to caps */ 407 int i_pin_ref; 408 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; 409 int i_wrbuffer_ref, i_wrbuffer_ref_head; 410 atomic_t i_filelock_ref; 411 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ 412 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 413 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 414 415 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 416 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 417 spinlock_t i_unsafe_lock; 418 419 union { 420 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 421 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ 422 }; 423 struct list_head i_snap_realm_item; 424 struct list_head i_snap_flush_item; 425 struct timespec64 i_btime; 426 struct timespec64 i_snap_btime; 427 428 struct work_struct i_work; 429 unsigned long i_work_mask; 430 431 #ifdef CONFIG_CEPH_FSCACHE 432 struct fscache_cookie *fscache; 433 #endif 434 struct inode vfs_inode; /* at end */ 435 }; 436 437 static inline struct ceph_inode_info * 438 ceph_inode(const struct inode *inode) 439 { 440 return container_of(inode, struct ceph_inode_info, vfs_inode); 441 } 442 443 static inline struct ceph_fs_client * 444 ceph_inode_to_client(const struct inode *inode) 445 { 446 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 447 } 448 449 static inline struct ceph_fs_client * 450 ceph_sb_to_client(const struct super_block *sb) 451 { 452 return (struct ceph_fs_client *)sb->s_fs_info; 453 } 454 455 static inline struct ceph_mds_client * 456 ceph_sb_to_mdsc(const struct super_block *sb) 457 { 458 return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc; 459 } 460 461 static inline struct ceph_vino 462 ceph_vino(const struct inode *inode) 463 { 464 return ceph_inode(inode)->i_vino; 465 } 466 467 static inline u32 ceph_ino_to_ino32(u64 vino) 468 { 469 u32 ino = vino & 0xffffffff; 470 ino ^= vino >> 32; 471 if (!ino) 472 ino = 2; 473 return ino; 474 } 475 476 /* 477 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on 478 * some arches. We generally do not use this value inside the ceph driver, but 479 * we do want to set it to something, so that generic vfs code has an 480 * appropriate value for tracepoints and the like. 481 */ 482 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino) 483 { 484 if (sizeof(ino_t) == sizeof(u32)) 485 return ceph_ino_to_ino32(vino.ino); 486 return (ino_t)vino.ino; 487 } 488 489 /* for printf-style formatting */ 490 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 491 492 static inline u64 ceph_ino(struct inode *inode) 493 { 494 return ceph_inode(inode)->i_vino.ino; 495 } 496 497 static inline u64 ceph_snap(struct inode *inode) 498 { 499 return ceph_inode(inode)->i_vino.snap; 500 } 501 502 /** 503 * ceph_present_ino - format an inode number for presentation to userland 504 * @sb: superblock where the inode lives 505 * @ino: inode number to (possibly) convert 506 * 507 * If the user mounted with the ino32 option, then the 64-bit value needs 508 * to be converted to something that can fit inside 32 bits. Note that 509 * internal kernel code never uses this value, so this is entirely for 510 * userland consumption. 511 */ 512 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino) 513 { 514 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))) 515 return ceph_ino_to_ino32(ino); 516 return ino; 517 } 518 519 static inline u64 ceph_present_inode(struct inode *inode) 520 { 521 return ceph_present_ino(inode->i_sb, ceph_ino(inode)); 522 } 523 524 static inline int ceph_ino_compare(struct inode *inode, void *data) 525 { 526 struct ceph_vino *pvino = (struct ceph_vino *)data; 527 struct ceph_inode_info *ci = ceph_inode(inode); 528 return ci->i_vino.ino == pvino->ino && 529 ci->i_vino.snap == pvino->snap; 530 } 531 532 /* 533 * The MDS reserves a set of inodes for its own usage. These should never 534 * be accessible by clients, and so the MDS has no reason to ever hand these 535 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE. 536 * 537 * These come from src/mds/mdstypes.h in the ceph sources. 538 */ 539 #define CEPH_MAX_MDS 0x100 540 #define CEPH_NUM_STRAY 10 541 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS) 542 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS) 543 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY)) 544 545 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino) 546 { 547 if (vino.ino >= CEPH_INO_SYSTEM_BASE || 548 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET) 549 return false; 550 551 /* Don't warn on mdsdirs */ 552 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET, 553 "Attempt to access reserved inode number 0x%llx", 554 vino.ino); 555 return true; 556 } 557 558 static inline struct inode *ceph_find_inode(struct super_block *sb, 559 struct ceph_vino vino) 560 { 561 if (ceph_vino_is_reserved(vino)) 562 return NULL; 563 564 /* 565 * NB: The hashval will be run through the fs/inode.c hash function 566 * anyway, so there is no need to squash the inode number down to 567 * 32-bits first. Just use low-order bits on arches with 32-bit long. 568 */ 569 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino); 570 } 571 572 573 /* 574 * Ceph inode. 575 */ 576 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 577 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 578 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */ 579 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */ 580 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */ 581 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */ 582 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */ 583 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */ 584 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */ 585 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */ 586 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */ 587 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */ 588 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT) 589 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */ 590 591 /* 592 * Masks of ceph inode work. 593 */ 594 #define CEPH_I_WORK_WRITEBACK 0 595 #define CEPH_I_WORK_INVALIDATE_PAGES 1 596 #define CEPH_I_WORK_VMTRUNCATE 2 597 #define CEPH_I_WORK_CHECK_CAPS 3 598 #define CEPH_I_WORK_FLUSH_SNAPS 4 599 600 /* 601 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 602 * and then clear it when they start succeeding. Note that we do a lockless 603 * check first, and only take the lock if it looks like it needs to be changed. 604 * The write submission code just takes this as a hint, so we're not too 605 * worried if a few slip through in either direction. 606 */ 607 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 608 { 609 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { 610 spin_lock(&ci->i_ceph_lock); 611 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; 612 spin_unlock(&ci->i_ceph_lock); 613 } 614 } 615 616 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 617 { 618 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { 619 spin_lock(&ci->i_ceph_lock); 620 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; 621 spin_unlock(&ci->i_ceph_lock); 622 } 623 } 624 625 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 626 long long release_count, 627 long long ordered_count) 628 { 629 /* 630 * Makes sure operations that setup readdir cache (update page 631 * cache and i_size) are strongly ordered w.r.t. the following 632 * atomic64_set() operations. 633 */ 634 smp_mb(); 635 atomic64_set(&ci->i_complete_seq[0], release_count); 636 atomic64_set(&ci->i_complete_seq[1], ordered_count); 637 } 638 639 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 640 { 641 atomic64_inc(&ci->i_release_count); 642 } 643 644 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 645 { 646 atomic64_inc(&ci->i_ordered_count); 647 } 648 649 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 650 { 651 return atomic64_read(&ci->i_complete_seq[0]) == 652 atomic64_read(&ci->i_release_count); 653 } 654 655 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 656 { 657 return atomic64_read(&ci->i_complete_seq[0]) == 658 atomic64_read(&ci->i_release_count) && 659 atomic64_read(&ci->i_complete_seq[1]) == 660 atomic64_read(&ci->i_ordered_count); 661 } 662 663 static inline void ceph_dir_clear_complete(struct inode *inode) 664 { 665 __ceph_dir_clear_complete(ceph_inode(inode)); 666 } 667 668 static inline void ceph_dir_clear_ordered(struct inode *inode) 669 { 670 __ceph_dir_clear_ordered(ceph_inode(inode)); 671 } 672 673 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 674 { 675 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 676 smp_rmb(); 677 return ret; 678 } 679 680 /* find a specific frag @f */ 681 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 682 u32 f); 683 684 /* 685 * choose fragment for value @v. copy frag content to pfrag, if leaf 686 * exists 687 */ 688 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 689 struct ceph_inode_frag *pfrag, 690 int *found); 691 692 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) 693 { 694 return (struct ceph_dentry_info *)dentry->d_fsdata; 695 } 696 697 /* 698 * caps helpers 699 */ 700 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 701 { 702 return !RB_EMPTY_ROOT(&ci->i_caps); 703 } 704 705 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 706 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 707 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, 708 int t); 709 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 710 struct ceph_cap *cap); 711 712 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 713 { 714 int issued; 715 spin_lock(&ci->i_ceph_lock); 716 issued = __ceph_caps_issued(ci, NULL); 717 spin_unlock(&ci->i_ceph_lock); 718 return issued; 719 } 720 721 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, 722 int mask, int touch) 723 { 724 int r; 725 spin_lock(&ci->i_ceph_lock); 726 r = __ceph_caps_issued_mask_metric(ci, mask, touch); 727 spin_unlock(&ci->i_ceph_lock); 728 return r; 729 } 730 731 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 732 { 733 return ci->i_dirty_caps | ci->i_flushing_caps; 734 } 735 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 736 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 737 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 738 struct ceph_cap_flush **pcf); 739 740 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 741 struct ceph_cap *ocap, int mask); 742 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 743 extern int __ceph_caps_used(struct ceph_inode_info *ci); 744 745 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci) 746 { 747 return ci->i_nr_by_mode[0]; 748 } 749 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 750 extern int __ceph_caps_wanted(struct ceph_inode_info *ci); 751 752 /* what the mds thinks we want */ 753 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 754 755 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 756 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 757 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, 758 struct ceph_mount_options *fsopt); 759 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 760 struct ceph_cap_reservation *ctx, int need); 761 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, 762 struct ceph_cap_reservation *ctx); 763 extern void ceph_reservation_status(struct ceph_fs_client *client, 764 int *total, int *avail, int *used, 765 int *reserved, int *min); 766 767 768 769 /* 770 * we keep buffered readdir results attached to file->private_data 771 */ 772 #define CEPH_F_SYNC 1 773 #define CEPH_F_ATEND 2 774 775 struct ceph_file_info { 776 short fmode; /* initialized on open */ 777 short flags; /* CEPH_F_* */ 778 779 spinlock_t rw_contexts_lock; 780 struct list_head rw_contexts; 781 782 u32 filp_gen; 783 atomic_t num_locks; 784 }; 785 786 struct ceph_dir_file_info { 787 struct ceph_file_info file_info; 788 789 /* readdir: position within the dir */ 790 u32 frag; 791 struct ceph_mds_request *last_readdir; 792 793 /* readdir: position within a frag */ 794 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 795 char *last_name; /* last entry in previous chunk */ 796 long long dir_release_count; 797 long long dir_ordered_count; 798 int readdir_cache_idx; 799 800 /* used for -o dirstat read() on directory thing */ 801 char *dir_info; 802 int dir_info_len; 803 }; 804 805 struct ceph_rw_context { 806 struct list_head list; 807 struct task_struct *thread; 808 int caps; 809 }; 810 811 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ 812 struct ceph_rw_context _name = { \ 813 .thread = current, \ 814 .caps = _caps, \ 815 } 816 817 static inline void ceph_add_rw_context(struct ceph_file_info *cf, 818 struct ceph_rw_context *ctx) 819 { 820 spin_lock(&cf->rw_contexts_lock); 821 list_add(&ctx->list, &cf->rw_contexts); 822 spin_unlock(&cf->rw_contexts_lock); 823 } 824 825 static inline void ceph_del_rw_context(struct ceph_file_info *cf, 826 struct ceph_rw_context *ctx) 827 { 828 spin_lock(&cf->rw_contexts_lock); 829 list_del(&ctx->list); 830 spin_unlock(&cf->rw_contexts_lock); 831 } 832 833 static inline struct ceph_rw_context* 834 ceph_find_rw_context(struct ceph_file_info *cf) 835 { 836 struct ceph_rw_context *ctx, *found = NULL; 837 spin_lock(&cf->rw_contexts_lock); 838 list_for_each_entry(ctx, &cf->rw_contexts, list) { 839 if (ctx->thread == current) { 840 found = ctx; 841 break; 842 } 843 } 844 spin_unlock(&cf->rw_contexts_lock); 845 return found; 846 } 847 848 struct ceph_readdir_cache_control { 849 struct page *page; 850 struct dentry **dentries; 851 int index; 852 }; 853 854 /* 855 * A "snap realm" describes a subset of the file hierarchy sharing 856 * the same set of snapshots that apply to it. The realms themselves 857 * are organized into a hierarchy, such that children inherit (some of) 858 * the snapshots of their parents. 859 * 860 * All inodes within the realm that have capabilities are linked into a 861 * per-realm list. 862 */ 863 struct ceph_snap_realm { 864 u64 ino; 865 struct inode *inode; 866 atomic_t nref; 867 struct rb_node node; 868 869 u64 created, seq; 870 u64 parent_ino; 871 u64 parent_since; /* snapid when our current parent became so */ 872 873 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 874 u32 num_prior_parent_snaps; /* had prior to parent_since */ 875 u64 *snaps; /* snaps specific to this realm */ 876 u32 num_snaps; 877 878 struct ceph_snap_realm *parent; 879 struct list_head children; /* list of child realms */ 880 struct list_head child_item; 881 882 struct list_head empty_item; /* if i have ref==0 */ 883 884 struct list_head dirty_item; /* if realm needs new context */ 885 886 /* the current set of snaps for this realm */ 887 struct ceph_snap_context *cached_context; 888 889 struct list_head inodes_with_caps; 890 spinlock_t inodes_with_caps_lock; 891 }; 892 893 static inline int default_congestion_kb(void) 894 { 895 int congestion_kb; 896 897 /* 898 * Copied from NFS 899 * 900 * congestion size, scale with available memory. 901 * 902 * 64MB: 8192k 903 * 128MB: 11585k 904 * 256MB: 16384k 905 * 512MB: 23170k 906 * 1GB: 32768k 907 * 2GB: 46340k 908 * 4GB: 65536k 909 * 8GB: 92681k 910 * 16GB: 131072k 911 * 912 * This allows larger machines to have larger/more transfers. 913 * Limit the default to 256M 914 */ 915 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); 916 if (congestion_kb > 256*1024) 917 congestion_kb = 256*1024; 918 919 return congestion_kb; 920 } 921 922 923 /* super.c */ 924 extern int ceph_force_reconnect(struct super_block *sb); 925 /* snap.c */ 926 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 927 u64 ino); 928 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 929 struct ceph_snap_realm *realm); 930 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 931 struct ceph_snap_realm *realm); 932 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 933 void *p, void *e, bool deletion, 934 struct ceph_snap_realm **realm_ret); 935 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm); 936 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 937 struct ceph_mds_session *session, 938 struct ceph_msg *msg); 939 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 940 struct ceph_cap_snap *capsnap); 941 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 942 943 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, 944 u64 snap); 945 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, 946 struct ceph_snapid_map *sm); 947 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); 948 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); 949 void ceph_umount_begin(struct super_block *sb); 950 951 952 /* 953 * a cap_snap is "pending" if it is still awaiting an in-progress 954 * sync write (that may/may not still update size, mtime, etc.). 955 */ 956 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 957 { 958 return !list_empty(&ci->i_cap_snaps) && 959 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 960 ci_item)->writing; 961 } 962 963 /* inode.c */ 964 struct ceph_mds_reply_info_in; 965 struct ceph_mds_reply_dirfrag; 966 967 extern const struct inode_operations ceph_file_iops; 968 969 extern struct inode *ceph_alloc_inode(struct super_block *sb); 970 extern void ceph_evict_inode(struct inode *inode); 971 extern void ceph_free_inode(struct inode *inode); 972 973 extern struct inode *ceph_get_inode(struct super_block *sb, 974 struct ceph_vino vino); 975 extern struct inode *ceph_get_snapdir(struct inode *parent); 976 extern int ceph_fill_file_size(struct inode *inode, int issued, 977 u32 truncate_seq, u64 truncate_size, u64 size); 978 extern void ceph_fill_file_time(struct inode *inode, int issued, 979 u64 time_warp_seq, struct timespec64 *ctime, 980 struct timespec64 *mtime, 981 struct timespec64 *atime); 982 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page, 983 struct ceph_mds_reply_info_in *iinfo, 984 struct ceph_mds_reply_dirfrag *dirinfo, 985 struct ceph_mds_session *session, int cap_fmode, 986 struct ceph_cap_reservation *caps_reservation); 987 extern int ceph_fill_trace(struct super_block *sb, 988 struct ceph_mds_request *req); 989 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 990 struct ceph_mds_session *session); 991 992 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 993 994 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 995 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 996 997 void ceph_queue_inode_work(struct inode *inode, int work_bit); 998 999 static inline void ceph_queue_vmtruncate(struct inode *inode) 1000 { 1001 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE); 1002 } 1003 1004 static inline void ceph_queue_invalidate(struct inode *inode) 1005 { 1006 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES); 1007 } 1008 1009 static inline void ceph_queue_writeback(struct inode *inode) 1010 { 1011 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK); 1012 } 1013 1014 static inline void ceph_queue_check_caps(struct inode *inode) 1015 { 1016 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS); 1017 } 1018 1019 static inline void ceph_queue_flush_snaps(struct inode *inode) 1020 { 1021 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS); 1022 } 1023 1024 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 1025 int mask, bool force); 1026 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 1027 { 1028 return __ceph_do_getattr(inode, NULL, mask, force); 1029 } 1030 extern int ceph_permission(struct user_namespace *mnt_userns, 1031 struct inode *inode, int mask); 1032 extern int __ceph_setattr(struct inode *inode, struct iattr *attr); 1033 extern int ceph_setattr(struct user_namespace *mnt_userns, 1034 struct dentry *dentry, struct iattr *attr); 1035 extern int ceph_getattr(struct user_namespace *mnt_userns, 1036 const struct path *path, struct kstat *stat, 1037 u32 request_mask, unsigned int flags); 1038 void ceph_inode_shutdown(struct inode *inode); 1039 1040 static inline bool ceph_inode_is_shutdown(struct inode *inode) 1041 { 1042 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags); 1043 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1044 int state = READ_ONCE(fsc->mount_state); 1045 1046 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN; 1047 } 1048 1049 /* xattr.c */ 1050 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 1051 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 1052 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 1053 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); 1054 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 1055 extern const struct xattr_handler *ceph_xattr_handlers[]; 1056 1057 struct ceph_acl_sec_ctx { 1058 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1059 void *default_acl; 1060 void *acl; 1061 #endif 1062 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1063 void *sec_ctx; 1064 u32 sec_ctxlen; 1065 #endif 1066 struct ceph_pagelist *pagelist; 1067 }; 1068 1069 #ifdef CONFIG_SECURITY 1070 extern bool ceph_security_xattr_deadlock(struct inode *in); 1071 extern bool ceph_security_xattr_wanted(struct inode *in); 1072 #else 1073 static inline bool ceph_security_xattr_deadlock(struct inode *in) 1074 { 1075 return false; 1076 } 1077 static inline bool ceph_security_xattr_wanted(struct inode *in) 1078 { 1079 return false; 1080 } 1081 #endif 1082 1083 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1084 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1085 struct ceph_acl_sec_ctx *ctx); 1086 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1087 { 1088 security_inode_invalidate_secctx(inode); 1089 } 1090 #else 1091 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1092 struct ceph_acl_sec_ctx *ctx) 1093 { 1094 return 0; 1095 } 1096 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1097 { 1098 } 1099 #endif 1100 1101 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); 1102 1103 /* acl.c */ 1104 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1105 1106 struct posix_acl *ceph_get_acl(struct inode *, int, bool); 1107 int ceph_set_acl(struct user_namespace *mnt_userns, 1108 struct inode *inode, struct posix_acl *acl, int type); 1109 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1110 struct ceph_acl_sec_ctx *as_ctx); 1111 void ceph_init_inode_acls(struct inode *inode, 1112 struct ceph_acl_sec_ctx *as_ctx); 1113 1114 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1115 { 1116 forget_all_cached_acls(inode); 1117 } 1118 1119 #else 1120 1121 #define ceph_get_acl NULL 1122 #define ceph_set_acl NULL 1123 1124 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1125 struct ceph_acl_sec_ctx *as_ctx) 1126 { 1127 return 0; 1128 } 1129 static inline void ceph_init_inode_acls(struct inode *inode, 1130 struct ceph_acl_sec_ctx *as_ctx) 1131 { 1132 } 1133 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 1134 { 1135 return 0; 1136 } 1137 1138 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1139 { 1140 } 1141 1142 #endif 1143 1144 /* caps.c */ 1145 extern const char *ceph_cap_string(int c); 1146 extern void ceph_handle_caps(struct ceph_mds_session *session, 1147 struct ceph_msg *msg); 1148 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 1149 struct ceph_cap_reservation *ctx); 1150 extern void ceph_add_cap(struct inode *inode, 1151 struct ceph_mds_session *session, u64 cap_id, 1152 unsigned issued, unsigned wanted, 1153 unsigned cap, unsigned seq, u64 realmino, int flags, 1154 struct ceph_cap **new_cap); 1155 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1156 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1157 extern void __ceph_remove_caps(struct ceph_inode_info *ci); 1158 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 1159 struct ceph_cap *cap); 1160 extern int ceph_is_any_caps(struct inode *inode); 1161 1162 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 1163 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 1164 int datasync); 1165 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 1166 struct ceph_mds_session *session); 1167 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 1168 struct ceph_mds_session *session); 1169 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, 1170 struct ceph_inode_info *ci); 1171 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 1172 int mds); 1173 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, 1174 bool snap_rwsem_locked); 1175 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 1176 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 1177 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had); 1178 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, 1179 int had); 1180 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 1181 struct ceph_snap_context *snapc); 1182 extern void __ceph_remove_capsnap(struct inode *inode, 1183 struct ceph_cap_snap *capsnap, 1184 bool *wake_ci, bool *wake_mdsc); 1185 extern void ceph_remove_capsnap(struct inode *inode, 1186 struct ceph_cap_snap *capsnap, 1187 bool *wake_ci, bool *wake_mdsc); 1188 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 1189 struct ceph_mds_session **psession); 1190 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 1191 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 1192 struct ceph_mds_session *session); 1193 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 1194 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 1195 extern int ceph_drop_caps_for_unlink(struct inode *inode); 1196 extern int ceph_encode_inode_release(void **p, struct inode *inode, 1197 int mds, int drop, int unless, int force); 1198 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 1199 struct inode *dir, 1200 int mds, int drop, int unless); 1201 1202 extern int ceph_get_caps(struct file *filp, int need, int want, 1203 loff_t endoff, int *got); 1204 extern int ceph_try_get_caps(struct inode *inode, 1205 int need, int want, bool nonblock, int *got); 1206 1207 /* for counting open files by mode */ 1208 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count); 1209 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count); 1210 extern void __ceph_touch_fmode(struct ceph_inode_info *ci, 1211 struct ceph_mds_client *mdsc, int fmode); 1212 1213 /* addr.c */ 1214 extern const struct address_space_operations ceph_aops; 1215 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 1216 extern int ceph_uninline_data(struct file *filp, struct page *locked_page); 1217 extern int ceph_pool_perm_check(struct inode *inode, int need); 1218 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 1219 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate); 1220 1221 /* file.c */ 1222 extern const struct file_operations ceph_file_fops; 1223 1224 extern int ceph_renew_caps(struct inode *inode, int fmode); 1225 extern int ceph_open(struct inode *inode, struct file *file); 1226 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 1227 struct file *file, unsigned flags, umode_t mode); 1228 extern int ceph_release(struct inode *inode, struct file *filp); 1229 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 1230 char *data, size_t len); 1231 1232 /* dir.c */ 1233 extern const struct file_operations ceph_dir_fops; 1234 extern const struct file_operations ceph_snapdir_fops; 1235 extern const struct inode_operations ceph_dir_iops; 1236 extern const struct inode_operations ceph_snapdir_iops; 1237 extern const struct dentry_operations ceph_dentry_ops; 1238 1239 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 1240 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 1241 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req, 1242 struct dentry *dentry); 1243 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 1244 struct dentry *dentry, int err); 1245 1246 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); 1247 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); 1248 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 1249 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); 1250 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 1251 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 1252 1253 /* ioctl.c */ 1254 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1255 1256 /* export.c */ 1257 extern const struct export_operations ceph_export_ops; 1258 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); 1259 1260 /* locks.c */ 1261 extern __init void ceph_flock_init(void); 1262 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 1263 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 1264 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 1265 extern int ceph_encode_locks_to_buffer(struct inode *inode, 1266 struct ceph_filelock *flocks, 1267 int num_fcntl_locks, 1268 int num_flock_locks); 1269 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 1270 struct ceph_pagelist *pagelist, 1271 int num_fcntl_locks, int num_flock_locks); 1272 1273 /* debugfs.c */ 1274 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); 1275 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1276 1277 /* quota.c */ 1278 static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci) 1279 { 1280 return ci->i_max_files || ci->i_max_bytes; 1281 } 1282 1283 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); 1284 1285 static inline void __ceph_update_quota(struct ceph_inode_info *ci, 1286 u64 max_bytes, u64 max_files) 1287 { 1288 bool had_quota, has_quota; 1289 had_quota = __ceph_has_any_quota(ci); 1290 ci->i_max_bytes = max_bytes; 1291 ci->i_max_files = max_files; 1292 has_quota = __ceph_has_any_quota(ci); 1293 1294 if (had_quota != has_quota) 1295 ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota); 1296 } 1297 1298 extern void ceph_handle_quota(struct ceph_mds_client *mdsc, 1299 struct ceph_mds_session *session, 1300 struct ceph_msg *msg); 1301 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); 1302 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); 1303 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, 1304 loff_t newlen); 1305 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, 1306 loff_t newlen); 1307 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, 1308 struct kstatfs *buf); 1309 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); 1310 1311 #endif /* _FS_CEPH_SUPER_H */ 1312