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 #include <linux/netfs.h> 21 #include <linux/fscache.h> 22 #include <linux/hashtable.h> 23 24 #include <linux/ceph/libceph.h> 25 26 /* large granularity for statfs utilization stats to facilitate 27 * large volume sizes on 32-bit machines. */ 28 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ 29 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 30 #define CEPH_4K_BLOCK_SHIFT 12 /* 4 KB */ 31 32 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */ 33 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ 34 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ 35 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ 36 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ 37 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ 38 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ 39 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */ 40 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */ 41 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */ 42 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */ 43 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */ 44 #define CEPH_MOUNT_OPT_NOPAGECACHE (1<<16) /* bypass pagecache altogether */ 45 #define CEPH_MOUNT_OPT_SPARSEREAD (1<<17) /* always do sparse reads */ 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 /* mount state */ 105 enum { 106 CEPH_MOUNT_MOUNTING, 107 CEPH_MOUNT_MOUNTED, 108 CEPH_MOUNT_UNMOUNTING, 109 CEPH_MOUNT_UNMOUNTED, 110 CEPH_MOUNT_SHUTDOWN, 111 CEPH_MOUNT_RECOVER, 112 CEPH_MOUNT_FENCE_IO, 113 }; 114 115 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8 116 117 struct ceph_fs_client { 118 struct super_block *sb; 119 120 struct list_head metric_wakeup; 121 122 struct ceph_mount_options *mount_options; 123 struct ceph_client *client; 124 125 int mount_state; 126 127 bool blocklisted; 128 129 bool have_copy_from2; 130 131 u32 filp_gen; 132 loff_t max_file_size; 133 134 struct ceph_mds_client *mdsc; 135 136 atomic_long_t writeback_count; 137 bool write_congested; 138 139 struct workqueue_struct *inode_wq; 140 struct workqueue_struct *cap_wq; 141 142 DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS); 143 spinlock_t async_unlink_conflict_lock; 144 145 #ifdef CONFIG_DEBUG_FS 146 struct dentry *debugfs_dentry_lru, *debugfs_caps; 147 struct dentry *debugfs_congestion_kb; 148 struct dentry *debugfs_bdi; 149 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 150 struct dentry *debugfs_status; 151 struct dentry *debugfs_mds_sessions; 152 struct dentry *debugfs_metrics_dir; 153 #endif 154 155 #ifdef CONFIG_CEPH_FSCACHE 156 struct fscache_volume *fscache; 157 #endif 158 }; 159 160 161 /* 162 * File i/o capability. This tracks shared state with the metadata 163 * server that allows us to cache or writeback attributes or to read 164 * and write data. For any given inode, we should have one or more 165 * capabilities, one issued by each metadata server, and our 166 * cumulative access is the OR of all issued capabilities. 167 * 168 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 169 * session capability lists. 170 */ 171 struct ceph_cap { 172 struct ceph_inode_info *ci; 173 struct rb_node ci_node; /* per-ci cap tree */ 174 struct ceph_mds_session *session; 175 struct list_head session_caps; /* per-session caplist */ 176 u64 cap_id; /* unique cap id (mds provided) */ 177 union { 178 /* in-use caps */ 179 struct { 180 int issued; /* latest, from the mds */ 181 int implemented; /* implemented superset of 182 issued (for revocation) */ 183 int mds; /* mds index for this cap */ 184 int mds_wanted; /* caps wanted from this mds */ 185 }; 186 /* caps to release */ 187 struct { 188 u64 cap_ino; 189 int queue_release; 190 }; 191 }; 192 u32 seq, issue_seq, mseq; 193 u32 cap_gen; /* active/stale cycle */ 194 unsigned long last_used; 195 struct list_head caps_item; 196 }; 197 198 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */ 199 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */ 200 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */ 201 202 struct ceph_cap_flush { 203 u64 tid; 204 int caps; 205 bool wake; /* wake up flush waiters when finish ? */ 206 bool is_capsnap; /* true means capsnap */ 207 struct list_head g_list; // global 208 struct list_head i_list; // per inode 209 }; 210 211 /* 212 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 213 * we first complete any in-process sync writes and writeback any dirty 214 * data before flushing the snapped state (tracked here) back to the MDS. 215 */ 216 struct ceph_cap_snap { 217 refcount_t nref; 218 struct list_head ci_item; 219 220 struct ceph_cap_flush cap_flush; 221 222 u64 follows; 223 int issued, dirty; 224 struct ceph_snap_context *context; 225 226 umode_t mode; 227 kuid_t uid; 228 kgid_t gid; 229 230 struct ceph_buffer *xattr_blob; 231 u64 xattr_version; 232 233 u64 size; 234 u64 change_attr; 235 struct timespec64 mtime, atime, ctime, btime; 236 u64 time_warp_seq; 237 u64 truncate_size; 238 u32 truncate_seq; 239 int writing; /* a sync write is still in progress */ 240 int dirty_pages; /* dirty pages awaiting writeback */ 241 bool inline_data; 242 bool need_flush; 243 }; 244 245 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 246 { 247 if (refcount_dec_and_test(&capsnap->nref)) { 248 if (capsnap->xattr_blob) 249 ceph_buffer_put(capsnap->xattr_blob); 250 kmem_cache_free(ceph_cap_snap_cachep, capsnap); 251 } 252 } 253 254 /* 255 * The frag tree describes how a directory is fragmented, potentially across 256 * multiple metadata servers. It is also used to indicate points where 257 * metadata authority is delegated, and whether/where metadata is replicated. 258 * 259 * A _leaf_ frag will be present in the i_fragtree IFF there is 260 * delegation info. That is, if mds >= 0 || ndist > 0. 261 */ 262 #define CEPH_MAX_DIRFRAG_REP 4 263 264 struct ceph_inode_frag { 265 struct rb_node node; 266 267 /* fragtree state */ 268 u32 frag; 269 int split_by; /* i.e. 2^(split_by) children */ 270 271 /* delegation and replication info */ 272 int mds; /* -1 if same authority as parent */ 273 int ndist; /* >0 if replicated */ 274 int dist[CEPH_MAX_DIRFRAG_REP]; 275 }; 276 277 /* 278 * We cache inode xattrs as an encoded blob until they are first used, 279 * at which point we parse them into an rbtree. 280 */ 281 struct ceph_inode_xattr { 282 struct rb_node node; 283 284 const char *name; 285 int name_len; 286 const char *val; 287 int val_len; 288 int dirty; 289 290 int should_free_name; 291 int should_free_val; 292 }; 293 294 /* 295 * Ceph dentry state 296 */ 297 struct ceph_dentry_info { 298 struct dentry *dentry; 299 struct ceph_mds_session *lease_session; 300 struct list_head lease_list; 301 struct hlist_node hnode; 302 unsigned long flags; 303 int lease_shared_gen; 304 u32 lease_gen; 305 u32 lease_seq; 306 unsigned long lease_renew_after, lease_renew_from; 307 unsigned long time; 308 u64 offset; 309 }; 310 311 #define CEPH_DENTRY_REFERENCED (1 << 0) 312 #define CEPH_DENTRY_LEASE_LIST (1 << 1) 313 #define CEPH_DENTRY_SHRINK_LIST (1 << 2) 314 #define CEPH_DENTRY_PRIMARY_LINK (1 << 3) 315 #define CEPH_DENTRY_ASYNC_UNLINK_BIT (4) 316 #define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT) 317 #define CEPH_DENTRY_ASYNC_CREATE_BIT (5) 318 #define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT) 319 320 struct ceph_inode_xattrs_info { 321 /* 322 * (still encoded) xattr blob. we avoid the overhead of parsing 323 * this until someone actually calls getxattr, etc. 324 * 325 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 326 * NULL means we don't know. 327 */ 328 struct ceph_buffer *blob, *prealloc_blob; 329 330 struct rb_root index; 331 bool dirty; 332 int count; 333 int names_size; 334 int vals_size; 335 u64 version, index_version; 336 }; 337 338 /* 339 * Ceph inode. 340 */ 341 struct ceph_inode_info { 342 struct netfs_inode netfs; /* Netfslib context and vfs inode */ 343 struct ceph_vino i_vino; /* ceph ino + snap */ 344 345 spinlock_t i_ceph_lock; 346 347 u64 i_version; 348 u64 i_inline_version; 349 u32 i_time_warp_seq; 350 351 unsigned long i_ceph_flags; 352 atomic64_t i_release_count; 353 atomic64_t i_ordered_count; 354 atomic64_t i_complete_seq[2]; 355 356 struct ceph_dir_layout i_dir_layout; 357 struct ceph_file_layout i_layout; 358 struct ceph_file_layout i_cached_layout; // for async creates 359 char *i_symlink; 360 361 /* for dirs */ 362 struct timespec64 i_rctime; 363 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps; 364 u64 i_files, i_subdirs; 365 366 /* quotas */ 367 u64 i_max_bytes, i_max_files; 368 369 s32 i_dir_pin; 370 371 struct rb_root i_fragtree; 372 int i_fragtree_nsplits; 373 struct mutex i_fragtree_mutex; 374 375 struct ceph_inode_xattrs_info i_xattrs; 376 377 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 378 * s_mutex. */ 379 struct rb_root i_caps; /* cap list */ 380 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 381 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 382 383 /* 384 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty 385 * is protected by the mdsc->cap_dirty_lock, but each individual item 386 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty 387 * requires the mdsc->cap_dirty_lock. List presence for an item can 388 * be tested under the i_ceph_lock. Changing anything requires both. 389 */ 390 struct list_head i_dirty_item; 391 392 /* 393 * Link to session's s_cap_flushing list. Protected in a similar 394 * fashion to i_dirty_item, but also by the s_mutex for changes. The 395 * s_cap_flushing list can be walked while holding either the s_mutex 396 * or msdc->cap_dirty_lock. List presence can also be checked while 397 * holding the i_ceph_lock for this inode. 398 */ 399 struct list_head i_flushing_item; 400 401 /* we need to track cap writeback on a per-cap-bit basis, to allow 402 * overlapping, pipelined cap flushes to the mds. we can probably 403 * reduce the tid to 8 bits if we're concerned about inode size. */ 404 struct ceph_cap_flush *i_prealloc_cap_flush; 405 struct list_head i_cap_flush_list; 406 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 407 unsigned long i_hold_caps_max; /* jiffies */ 408 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 409 struct ceph_cap_reservation i_cap_migration_resv; 410 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 411 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 412 dirty|flushing caps */ 413 unsigned i_snap_caps; /* cap bits for snapped files */ 414 415 unsigned long i_last_rd; 416 unsigned long i_last_wr; 417 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ 418 419 struct mutex i_truncate_mutex; 420 u32 i_truncate_seq; /* last truncate to smaller size */ 421 u64 i_truncate_size; /* and the size we last truncated down to */ 422 int i_truncate_pending; /* still need to call vmtruncate */ 423 424 u64 i_max_size; /* max file size authorized by mds */ 425 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 426 u64 i_wanted_max_size; /* offset we'd like to write too */ 427 u64 i_requested_max_size; /* max_size we've requested */ 428 429 /* held references to caps */ 430 int i_pin_ref; 431 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; 432 int i_wrbuffer_ref, i_wrbuffer_ref_head; 433 atomic_t i_filelock_ref; 434 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ 435 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 436 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 437 438 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 439 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 440 spinlock_t i_unsafe_lock; 441 442 union { 443 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 444 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ 445 }; 446 struct list_head i_snap_realm_item; 447 struct list_head i_snap_flush_item; 448 struct timespec64 i_btime; 449 struct timespec64 i_snap_btime; 450 451 struct work_struct i_work; 452 unsigned long i_work_mask; 453 }; 454 455 struct ceph_netfs_request_data { 456 int caps; 457 458 /* 459 * Maximum size of a file readahead request. 460 * The fadvise could update the bdi's default ra_pages. 461 */ 462 unsigned int file_ra_pages; 463 464 /* Set it if fadvise disables file readahead entirely */ 465 bool file_ra_disabled; 466 }; 467 468 static inline struct ceph_inode_info * 469 ceph_inode(const struct inode *inode) 470 { 471 return container_of(inode, struct ceph_inode_info, netfs.inode); 472 } 473 474 static inline struct ceph_fs_client * 475 ceph_inode_to_client(const struct inode *inode) 476 { 477 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 478 } 479 480 static inline struct ceph_fs_client * 481 ceph_sb_to_client(const struct super_block *sb) 482 { 483 return (struct ceph_fs_client *)sb->s_fs_info; 484 } 485 486 static inline struct ceph_mds_client * 487 ceph_sb_to_mdsc(const struct super_block *sb) 488 { 489 return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc; 490 } 491 492 static inline struct ceph_vino 493 ceph_vino(const struct inode *inode) 494 { 495 return ceph_inode(inode)->i_vino; 496 } 497 498 static inline u32 ceph_ino_to_ino32(u64 vino) 499 { 500 u32 ino = vino & 0xffffffff; 501 ino ^= vino >> 32; 502 if (!ino) 503 ino = 2; 504 return ino; 505 } 506 507 /* 508 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on 509 * some arches. We generally do not use this value inside the ceph driver, but 510 * we do want to set it to something, so that generic vfs code has an 511 * appropriate value for tracepoints and the like. 512 */ 513 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino) 514 { 515 if (sizeof(ino_t) == sizeof(u32)) 516 return ceph_ino_to_ino32(vino.ino); 517 return (ino_t)vino.ino; 518 } 519 520 /* for printf-style formatting */ 521 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 522 523 static inline u64 ceph_ino(struct inode *inode) 524 { 525 return ceph_inode(inode)->i_vino.ino; 526 } 527 528 static inline u64 ceph_snap(struct inode *inode) 529 { 530 return ceph_inode(inode)->i_vino.snap; 531 } 532 533 /** 534 * ceph_present_ino - format an inode number for presentation to userland 535 * @sb: superblock where the inode lives 536 * @ino: inode number to (possibly) convert 537 * 538 * If the user mounted with the ino32 option, then the 64-bit value needs 539 * to be converted to something that can fit inside 32 bits. Note that 540 * internal kernel code never uses this value, so this is entirely for 541 * userland consumption. 542 */ 543 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino) 544 { 545 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))) 546 return ceph_ino_to_ino32(ino); 547 return ino; 548 } 549 550 static inline u64 ceph_present_inode(struct inode *inode) 551 { 552 return ceph_present_ino(inode->i_sb, ceph_ino(inode)); 553 } 554 555 static inline int ceph_ino_compare(struct inode *inode, void *data) 556 { 557 struct ceph_vino *pvino = (struct ceph_vino *)data; 558 struct ceph_inode_info *ci = ceph_inode(inode); 559 return ci->i_vino.ino == pvino->ino && 560 ci->i_vino.snap == pvino->snap; 561 } 562 563 /* 564 * The MDS reserves a set of inodes for its own usage. These should never 565 * be accessible by clients, and so the MDS has no reason to ever hand these 566 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE. 567 * 568 * These come from src/mds/mdstypes.h in the ceph sources. 569 */ 570 #define CEPH_MAX_MDS 0x100 571 #define CEPH_NUM_STRAY 10 572 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS) 573 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS) 574 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY)) 575 576 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino) 577 { 578 if (vino.ino >= CEPH_INO_SYSTEM_BASE || 579 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET) 580 return false; 581 582 /* Don't warn on mdsdirs */ 583 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET, 584 "Attempt to access reserved inode number 0x%llx", 585 vino.ino); 586 return true; 587 } 588 589 static inline struct inode *ceph_find_inode(struct super_block *sb, 590 struct ceph_vino vino) 591 { 592 if (ceph_vino_is_reserved(vino)) 593 return NULL; 594 595 /* 596 * NB: The hashval will be run through the fs/inode.c hash function 597 * anyway, so there is no need to squash the inode number down to 598 * 32-bits first. Just use low-order bits on arches with 32-bit long. 599 */ 600 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino); 601 } 602 603 604 /* 605 * Ceph inode. 606 */ 607 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 608 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 609 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */ 610 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */ 611 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */ 612 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */ 613 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */ 614 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */ 615 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */ 616 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */ 617 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */ 618 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */ 619 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT) 620 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */ 621 #define CEPH_I_ASYNC_CHECK_CAPS (1 << 14) /* check caps immediately after async 622 creating finishes */ 623 624 /* 625 * Masks of ceph inode work. 626 */ 627 #define CEPH_I_WORK_WRITEBACK 0 628 #define CEPH_I_WORK_INVALIDATE_PAGES 1 629 #define CEPH_I_WORK_VMTRUNCATE 2 630 #define CEPH_I_WORK_CHECK_CAPS 3 631 #define CEPH_I_WORK_FLUSH_SNAPS 4 632 633 /* 634 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 635 * and then clear it when they start succeeding. Note that we do a lockless 636 * check first, and only take the lock if it looks like it needs to be changed. 637 * The write submission code just takes this as a hint, so we're not too 638 * worried if a few slip through in either direction. 639 */ 640 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 641 { 642 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { 643 spin_lock(&ci->i_ceph_lock); 644 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; 645 spin_unlock(&ci->i_ceph_lock); 646 } 647 } 648 649 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 650 { 651 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { 652 spin_lock(&ci->i_ceph_lock); 653 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; 654 spin_unlock(&ci->i_ceph_lock); 655 } 656 } 657 658 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 659 long long release_count, 660 long long ordered_count) 661 { 662 /* 663 * Makes sure operations that setup readdir cache (update page 664 * cache and i_size) are strongly ordered w.r.t. the following 665 * atomic64_set() operations. 666 */ 667 smp_mb(); 668 atomic64_set(&ci->i_complete_seq[0], release_count); 669 atomic64_set(&ci->i_complete_seq[1], ordered_count); 670 } 671 672 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 673 { 674 atomic64_inc(&ci->i_release_count); 675 } 676 677 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 678 { 679 atomic64_inc(&ci->i_ordered_count); 680 } 681 682 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 683 { 684 return atomic64_read(&ci->i_complete_seq[0]) == 685 atomic64_read(&ci->i_release_count); 686 } 687 688 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 689 { 690 return atomic64_read(&ci->i_complete_seq[0]) == 691 atomic64_read(&ci->i_release_count) && 692 atomic64_read(&ci->i_complete_seq[1]) == 693 atomic64_read(&ci->i_ordered_count); 694 } 695 696 static inline void ceph_dir_clear_complete(struct inode *inode) 697 { 698 __ceph_dir_clear_complete(ceph_inode(inode)); 699 } 700 701 static inline void ceph_dir_clear_ordered(struct inode *inode) 702 { 703 __ceph_dir_clear_ordered(ceph_inode(inode)); 704 } 705 706 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 707 { 708 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 709 smp_rmb(); 710 return ret; 711 } 712 713 /* find a specific frag @f */ 714 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 715 u32 f); 716 717 /* 718 * choose fragment for value @v. copy frag content to pfrag, if leaf 719 * exists 720 */ 721 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 722 struct ceph_inode_frag *pfrag, 723 int *found); 724 725 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) 726 { 727 return (struct ceph_dentry_info *)dentry->d_fsdata; 728 } 729 730 /* 731 * caps helpers 732 */ 733 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 734 { 735 return !RB_EMPTY_ROOT(&ci->i_caps); 736 } 737 738 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 739 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 740 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, 741 int t); 742 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 743 struct ceph_cap *cap); 744 745 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 746 { 747 int issued; 748 spin_lock(&ci->i_ceph_lock); 749 issued = __ceph_caps_issued(ci, NULL); 750 spin_unlock(&ci->i_ceph_lock); 751 return issued; 752 } 753 754 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, 755 int mask, int touch) 756 { 757 int r; 758 spin_lock(&ci->i_ceph_lock); 759 r = __ceph_caps_issued_mask_metric(ci, mask, touch); 760 spin_unlock(&ci->i_ceph_lock); 761 return r; 762 } 763 764 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 765 { 766 return ci->i_dirty_caps | ci->i_flushing_caps; 767 } 768 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 769 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 770 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 771 struct ceph_cap_flush **pcf); 772 773 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 774 struct ceph_cap *ocap, int mask); 775 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 776 extern int __ceph_caps_used(struct ceph_inode_info *ci); 777 778 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci) 779 { 780 return ci->i_nr_by_mode[0]; 781 } 782 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 783 extern int __ceph_caps_wanted(struct ceph_inode_info *ci); 784 785 /* what the mds thinks we want */ 786 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 787 788 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 789 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 790 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, 791 struct ceph_mount_options *fsopt); 792 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 793 struct ceph_cap_reservation *ctx, int need); 794 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, 795 struct ceph_cap_reservation *ctx); 796 extern void ceph_reservation_status(struct ceph_fs_client *client, 797 int *total, int *avail, int *used, 798 int *reserved, int *min); 799 extern void change_auth_cap_ses(struct ceph_inode_info *ci, 800 struct ceph_mds_session *session); 801 802 803 804 /* 805 * we keep buffered readdir results attached to file->private_data 806 */ 807 #define CEPH_F_SYNC 1 808 #define CEPH_F_ATEND 2 809 810 struct ceph_file_info { 811 short fmode; /* initialized on open */ 812 short flags; /* CEPH_F_* */ 813 814 spinlock_t rw_contexts_lock; 815 struct list_head rw_contexts; 816 817 u32 filp_gen; 818 }; 819 820 struct ceph_dir_file_info { 821 struct ceph_file_info file_info; 822 823 /* readdir: position within the dir */ 824 u32 frag; 825 struct ceph_mds_request *last_readdir; 826 827 /* readdir: position within a frag */ 828 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 829 char *last_name; /* last entry in previous chunk */ 830 long long dir_release_count; 831 long long dir_ordered_count; 832 int readdir_cache_idx; 833 834 /* used for -o dirstat read() on directory thing */ 835 char *dir_info; 836 int dir_info_len; 837 }; 838 839 struct ceph_rw_context { 840 struct list_head list; 841 struct task_struct *thread; 842 int caps; 843 }; 844 845 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ 846 struct ceph_rw_context _name = { \ 847 .thread = current, \ 848 .caps = _caps, \ 849 } 850 851 static inline void ceph_add_rw_context(struct ceph_file_info *cf, 852 struct ceph_rw_context *ctx) 853 { 854 spin_lock(&cf->rw_contexts_lock); 855 list_add(&ctx->list, &cf->rw_contexts); 856 spin_unlock(&cf->rw_contexts_lock); 857 } 858 859 static inline void ceph_del_rw_context(struct ceph_file_info *cf, 860 struct ceph_rw_context *ctx) 861 { 862 spin_lock(&cf->rw_contexts_lock); 863 list_del(&ctx->list); 864 spin_unlock(&cf->rw_contexts_lock); 865 } 866 867 static inline struct ceph_rw_context* 868 ceph_find_rw_context(struct ceph_file_info *cf) 869 { 870 struct ceph_rw_context *ctx, *found = NULL; 871 spin_lock(&cf->rw_contexts_lock); 872 list_for_each_entry(ctx, &cf->rw_contexts, list) { 873 if (ctx->thread == current) { 874 found = ctx; 875 break; 876 } 877 } 878 spin_unlock(&cf->rw_contexts_lock); 879 return found; 880 } 881 882 struct ceph_readdir_cache_control { 883 struct page *page; 884 struct dentry **dentries; 885 int index; 886 }; 887 888 /* 889 * A "snap realm" describes a subset of the file hierarchy sharing 890 * the same set of snapshots that apply to it. The realms themselves 891 * are organized into a hierarchy, such that children inherit (some of) 892 * the snapshots of their parents. 893 * 894 * All inodes within the realm that have capabilities are linked into a 895 * per-realm list. 896 */ 897 struct ceph_snap_realm { 898 u64 ino; 899 struct inode *inode; 900 atomic_t nref; 901 struct rb_node node; 902 903 u64 created, seq; 904 u64 parent_ino; 905 u64 parent_since; /* snapid when our current parent became so */ 906 907 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 908 u32 num_prior_parent_snaps; /* had prior to parent_since */ 909 u64 *snaps; /* snaps specific to this realm */ 910 u32 num_snaps; 911 912 struct ceph_snap_realm *parent; 913 struct list_head children; /* list of child realms */ 914 struct list_head child_item; 915 916 struct list_head empty_item; /* if i have ref==0 */ 917 918 struct list_head dirty_item; /* if realm needs new context */ 919 920 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */ 921 922 /* the current set of snaps for this realm */ 923 struct ceph_snap_context *cached_context; 924 925 struct list_head inodes_with_caps; 926 spinlock_t inodes_with_caps_lock; 927 }; 928 929 static inline int default_congestion_kb(void) 930 { 931 int congestion_kb; 932 933 /* 934 * Copied from NFS 935 * 936 * congestion size, scale with available memory. 937 * 938 * 64MB: 8192k 939 * 128MB: 11585k 940 * 256MB: 16384k 941 * 512MB: 23170k 942 * 1GB: 32768k 943 * 2GB: 46340k 944 * 4GB: 65536k 945 * 8GB: 92681k 946 * 16GB: 131072k 947 * 948 * This allows larger machines to have larger/more transfers. 949 * Limit the default to 256M 950 */ 951 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); 952 if (congestion_kb > 256*1024) 953 congestion_kb = 256*1024; 954 955 return congestion_kb; 956 } 957 958 959 /* super.c */ 960 extern int ceph_force_reconnect(struct super_block *sb); 961 /* snap.c */ 962 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 963 u64 ino); 964 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 965 struct ceph_snap_realm *realm); 966 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 967 struct ceph_snap_realm *realm); 968 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 969 void *p, void *e, bool deletion, 970 struct ceph_snap_realm **realm_ret); 971 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm); 972 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 973 struct ceph_mds_session *session, 974 struct ceph_msg *msg); 975 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 976 struct ceph_cap_snap *capsnap); 977 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc); 978 979 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, 980 u64 snap); 981 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, 982 struct ceph_snapid_map *sm); 983 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); 984 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); 985 void ceph_umount_begin(struct super_block *sb); 986 987 988 /* 989 * a cap_snap is "pending" if it is still awaiting an in-progress 990 * sync write (that may/may not still update size, mtime, etc.). 991 */ 992 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 993 { 994 return !list_empty(&ci->i_cap_snaps) && 995 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 996 ci_item)->writing; 997 } 998 999 /* inode.c */ 1000 struct ceph_mds_reply_info_in; 1001 struct ceph_mds_reply_dirfrag; 1002 1003 extern const struct inode_operations ceph_file_iops; 1004 1005 extern struct inode *ceph_alloc_inode(struct super_block *sb); 1006 extern void ceph_evict_inode(struct inode *inode); 1007 extern void ceph_free_inode(struct inode *inode); 1008 1009 extern struct inode *ceph_get_inode(struct super_block *sb, 1010 struct ceph_vino vino); 1011 extern struct inode *ceph_get_snapdir(struct inode *parent); 1012 extern int ceph_fill_file_size(struct inode *inode, int issued, 1013 u32 truncate_seq, u64 truncate_size, u64 size); 1014 extern void ceph_fill_file_time(struct inode *inode, int issued, 1015 u64 time_warp_seq, struct timespec64 *ctime, 1016 struct timespec64 *mtime, 1017 struct timespec64 *atime); 1018 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page, 1019 struct ceph_mds_reply_info_in *iinfo, 1020 struct ceph_mds_reply_dirfrag *dirinfo, 1021 struct ceph_mds_session *session, int cap_fmode, 1022 struct ceph_cap_reservation *caps_reservation); 1023 extern int ceph_fill_trace(struct super_block *sb, 1024 struct ceph_mds_request *req); 1025 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 1026 struct ceph_mds_session *session); 1027 1028 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 1029 1030 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 1031 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 1032 1033 void ceph_queue_inode_work(struct inode *inode, int work_bit); 1034 1035 static inline void ceph_queue_vmtruncate(struct inode *inode) 1036 { 1037 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE); 1038 } 1039 1040 static inline void ceph_queue_invalidate(struct inode *inode) 1041 { 1042 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES); 1043 } 1044 1045 static inline void ceph_queue_writeback(struct inode *inode) 1046 { 1047 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK); 1048 } 1049 1050 static inline void ceph_queue_check_caps(struct inode *inode) 1051 { 1052 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS); 1053 } 1054 1055 static inline void ceph_queue_flush_snaps(struct inode *inode) 1056 { 1057 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS); 1058 } 1059 1060 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask); 1061 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 1062 int mask, bool force); 1063 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 1064 { 1065 return __ceph_do_getattr(inode, NULL, mask, force); 1066 } 1067 extern int ceph_permission(struct mnt_idmap *idmap, 1068 struct inode *inode, int mask); 1069 extern int __ceph_setattr(struct inode *inode, struct iattr *attr); 1070 extern int ceph_setattr(struct mnt_idmap *idmap, 1071 struct dentry *dentry, struct iattr *attr); 1072 extern int ceph_getattr(struct mnt_idmap *idmap, 1073 const struct path *path, struct kstat *stat, 1074 u32 request_mask, unsigned int flags); 1075 void ceph_inode_shutdown(struct inode *inode); 1076 1077 static inline bool ceph_inode_is_shutdown(struct inode *inode) 1078 { 1079 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags); 1080 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1081 int state = READ_ONCE(fsc->mount_state); 1082 1083 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN; 1084 } 1085 1086 /* xattr.c */ 1087 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 1088 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size); 1089 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 1090 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 1091 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); 1092 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 1093 extern const struct xattr_handler *ceph_xattr_handlers[]; 1094 1095 struct ceph_acl_sec_ctx { 1096 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1097 void *default_acl; 1098 void *acl; 1099 #endif 1100 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1101 void *sec_ctx; 1102 u32 sec_ctxlen; 1103 #endif 1104 struct ceph_pagelist *pagelist; 1105 }; 1106 1107 #ifdef CONFIG_SECURITY 1108 extern bool ceph_security_xattr_deadlock(struct inode *in); 1109 extern bool ceph_security_xattr_wanted(struct inode *in); 1110 #else 1111 static inline bool ceph_security_xattr_deadlock(struct inode *in) 1112 { 1113 return false; 1114 } 1115 static inline bool ceph_security_xattr_wanted(struct inode *in) 1116 { 1117 return false; 1118 } 1119 #endif 1120 1121 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1122 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1123 struct ceph_acl_sec_ctx *ctx); 1124 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1125 { 1126 security_inode_invalidate_secctx(inode); 1127 } 1128 #else 1129 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1130 struct ceph_acl_sec_ctx *ctx) 1131 { 1132 return 0; 1133 } 1134 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1135 { 1136 } 1137 #endif 1138 1139 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); 1140 1141 /* acl.c */ 1142 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1143 1144 struct posix_acl *ceph_get_acl(struct inode *, int, bool); 1145 int ceph_set_acl(struct mnt_idmap *idmap, 1146 struct dentry *dentry, struct posix_acl *acl, int type); 1147 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1148 struct ceph_acl_sec_ctx *as_ctx); 1149 void ceph_init_inode_acls(struct inode *inode, 1150 struct ceph_acl_sec_ctx *as_ctx); 1151 1152 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1153 { 1154 forget_all_cached_acls(inode); 1155 } 1156 1157 #else 1158 1159 #define ceph_get_acl NULL 1160 #define ceph_set_acl NULL 1161 1162 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1163 struct ceph_acl_sec_ctx *as_ctx) 1164 { 1165 return 0; 1166 } 1167 static inline void ceph_init_inode_acls(struct inode *inode, 1168 struct ceph_acl_sec_ctx *as_ctx) 1169 { 1170 } 1171 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 1172 { 1173 return 0; 1174 } 1175 1176 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1177 { 1178 } 1179 1180 #endif 1181 1182 /* caps.c */ 1183 extern const char *ceph_cap_string(int c); 1184 extern void ceph_handle_caps(struct ceph_mds_session *session, 1185 struct ceph_msg *msg); 1186 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 1187 struct ceph_cap_reservation *ctx); 1188 extern void ceph_add_cap(struct inode *inode, 1189 struct ceph_mds_session *session, u64 cap_id, 1190 unsigned issued, unsigned wanted, 1191 unsigned cap, unsigned seq, u64 realmino, int flags, 1192 struct ceph_cap **new_cap); 1193 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1194 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1195 extern void __ceph_remove_caps(struct ceph_inode_info *ci); 1196 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 1197 struct ceph_cap *cap); 1198 extern int ceph_is_any_caps(struct inode *inode); 1199 1200 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 1201 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 1202 int datasync); 1203 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 1204 struct ceph_mds_session *session); 1205 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 1206 struct ceph_mds_session *session); 1207 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, 1208 struct ceph_inode_info *ci); 1209 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, 1210 int mds); 1211 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 1212 int mds); 1213 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, 1214 bool snap_rwsem_locked); 1215 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 1216 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 1217 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had); 1218 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, 1219 int had); 1220 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 1221 struct ceph_snap_context *snapc); 1222 extern void __ceph_remove_capsnap(struct inode *inode, 1223 struct ceph_cap_snap *capsnap, 1224 bool *wake_ci, bool *wake_mdsc); 1225 extern void ceph_remove_capsnap(struct inode *inode, 1226 struct ceph_cap_snap *capsnap, 1227 bool *wake_ci, bool *wake_mdsc); 1228 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 1229 struct ceph_mds_session **psession); 1230 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 1231 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags); 1232 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 1233 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 1234 extern int ceph_drop_caps_for_unlink(struct inode *inode); 1235 extern int ceph_encode_inode_release(void **p, struct inode *inode, 1236 int mds, int drop, int unless, int force); 1237 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 1238 struct inode *dir, 1239 int mds, int drop, int unless); 1240 1241 extern int ceph_get_caps(struct file *filp, int need, int want, 1242 loff_t endoff, int *got); 1243 extern int ceph_try_get_caps(struct inode *inode, 1244 int need, int want, bool nonblock, int *got); 1245 1246 /* for counting open files by mode */ 1247 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count); 1248 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count); 1249 extern void __ceph_touch_fmode(struct ceph_inode_info *ci, 1250 struct ceph_mds_client *mdsc, int fmode); 1251 1252 /* addr.c */ 1253 extern const struct address_space_operations ceph_aops; 1254 extern const struct netfs_request_ops ceph_netfs_ops; 1255 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 1256 extern int ceph_uninline_data(struct file *file); 1257 extern int ceph_pool_perm_check(struct inode *inode, int need); 1258 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 1259 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate); 1260 1261 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci) 1262 { 1263 if (ci->i_inline_version == CEPH_INLINE_NONE || 1264 ci->i_inline_version == 1) /* initial version, no data */ 1265 return false; 1266 return true; 1267 } 1268 1269 /* file.c */ 1270 extern const struct file_operations ceph_file_fops; 1271 1272 extern int ceph_renew_caps(struct inode *inode, int fmode); 1273 extern int ceph_open(struct inode *inode, struct file *file); 1274 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 1275 struct file *file, unsigned flags, umode_t mode); 1276 extern int ceph_release(struct inode *inode, struct file *filp); 1277 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 1278 char *data, size_t len); 1279 1280 /* dir.c */ 1281 extern const struct file_operations ceph_dir_fops; 1282 extern const struct file_operations ceph_snapdir_fops; 1283 extern const struct inode_operations ceph_dir_iops; 1284 extern const struct inode_operations ceph_snapdir_iops; 1285 extern const struct dentry_operations ceph_dentry_ops; 1286 1287 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 1288 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 1289 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req, 1290 struct dentry *dentry); 1291 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 1292 struct dentry *dentry, int err); 1293 1294 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); 1295 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); 1296 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 1297 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); 1298 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 1299 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 1300 1301 /* ioctl.c */ 1302 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1303 1304 /* export.c */ 1305 extern const struct export_operations ceph_export_ops; 1306 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); 1307 1308 /* locks.c */ 1309 extern __init void ceph_flock_init(void); 1310 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 1311 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 1312 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 1313 extern int ceph_encode_locks_to_buffer(struct inode *inode, 1314 struct ceph_filelock *flocks, 1315 int num_fcntl_locks, 1316 int num_flock_locks); 1317 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 1318 struct ceph_pagelist *pagelist, 1319 int num_fcntl_locks, int num_flock_locks); 1320 1321 /* debugfs.c */ 1322 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); 1323 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1324 1325 /* quota.c */ 1326 1327 enum quota_get_realm { 1328 QUOTA_GET_MAX_FILES, 1329 QUOTA_GET_MAX_BYTES, 1330 QUOTA_GET_ANY 1331 }; 1332 1333 static inline bool __ceph_has_quota(struct ceph_inode_info *ci, 1334 enum quota_get_realm which) 1335 { 1336 bool has_quota = false; 1337 1338 switch (which) { 1339 case QUOTA_GET_MAX_BYTES: 1340 has_quota = !!ci->i_max_bytes; 1341 break; 1342 case QUOTA_GET_MAX_FILES: 1343 has_quota = !!ci->i_max_files; 1344 break; 1345 default: 1346 has_quota = !!(ci->i_max_files || ci->i_max_bytes); 1347 } 1348 return has_quota; 1349 } 1350 1351 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); 1352 1353 static inline void __ceph_update_quota(struct ceph_inode_info *ci, 1354 u64 max_bytes, u64 max_files) 1355 { 1356 bool had_quota, has_quota; 1357 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1358 ci->i_max_bytes = max_bytes; 1359 ci->i_max_files = max_files; 1360 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1361 1362 if (had_quota != has_quota) 1363 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota); 1364 } 1365 1366 extern void ceph_handle_quota(struct ceph_mds_client *mdsc, 1367 struct ceph_mds_session *session, 1368 struct ceph_msg *msg); 1369 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); 1370 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); 1371 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, 1372 loff_t newlen); 1373 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, 1374 loff_t newlen); 1375 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, 1376 struct kstatfs *buf); 1377 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); 1378 1379 #endif /* _FS_CEPH_SUPER_H */ 1380