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 #include "crypto.h" 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 #define CEPH_MOUNT_OPT_SPARSEREAD (1<<17) /* always do sparse reads */ 47 48 #define CEPH_MOUNT_OPT_DEFAULT \ 49 (CEPH_MOUNT_OPT_DCACHE | \ 50 CEPH_MOUNT_OPT_NOCOPYFROM | \ 51 CEPH_MOUNT_OPT_ASYNC_DIROPS) 52 53 #define ceph_set_mount_opt(fsc, opt) \ 54 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt 55 #define ceph_clear_mount_opt(fsc, opt) \ 56 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt 57 #define ceph_test_mount_opt(fsc, opt) \ 58 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 59 60 /* max size of osd read request, limited by libceph */ 61 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN 62 /* osd has a configurable limitaion of max write size. 63 * CEPH_MSG_MAX_DATA_LEN should be small enough. */ 64 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN 65 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */ 66 #define CEPH_MAX_READDIR_DEFAULT 1024 67 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 68 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 69 70 /* 71 * Delay telling the MDS we no longer want caps, in case we reopen 72 * the file. Delay a minimum amount of time, even if we send a cap 73 * message for some other reason. Otherwise, take the oppotunity to 74 * update the mds to avoid sending another message later. 75 */ 76 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ 77 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ 78 79 struct ceph_mount_options { 80 unsigned int flags; 81 82 unsigned int wsize; /* max write size */ 83 unsigned int rsize; /* max read size */ 84 unsigned int rasize; /* max readahead */ 85 unsigned int congestion_kb; /* max writeback in flight */ 86 unsigned int caps_wanted_delay_min, caps_wanted_delay_max; 87 int caps_max; 88 unsigned int max_readdir; /* max readdir result (entries) */ 89 unsigned int max_readdir_bytes; /* max readdir result (bytes) */ 90 91 bool new_dev_syntax; 92 93 /* 94 * everything above this point can be memcmp'd; everything below 95 * is handled in compare_mount_options() 96 */ 97 98 char *snapdir_name; /* default ".snap" */ 99 char *mds_namespace; /* default NULL */ 100 char *server_path; /* default NULL (means "/") */ 101 char *fscache_uniq; /* default NULL */ 102 char *mon_addr; 103 struct fscrypt_dummy_policy dummy_enc_policy; 104 }; 105 106 /* mount state */ 107 enum { 108 CEPH_MOUNT_MOUNTING, 109 CEPH_MOUNT_MOUNTED, 110 CEPH_MOUNT_UNMOUNTING, 111 CEPH_MOUNT_UNMOUNTED, 112 CEPH_MOUNT_SHUTDOWN, 113 CEPH_MOUNT_RECOVER, 114 CEPH_MOUNT_FENCE_IO, 115 }; 116 117 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8 118 119 struct ceph_fs_client { 120 struct super_block *sb; 121 122 struct list_head metric_wakeup; 123 124 struct ceph_mount_options *mount_options; 125 struct ceph_client *client; 126 127 int mount_state; 128 129 bool blocklisted; 130 131 bool have_copy_from2; 132 133 u32 filp_gen; 134 loff_t max_file_size; 135 136 struct ceph_mds_client *mdsc; 137 138 atomic_long_t writeback_count; 139 bool write_congested; 140 141 struct workqueue_struct *inode_wq; 142 struct workqueue_struct *cap_wq; 143 144 DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS); 145 spinlock_t async_unlink_conflict_lock; 146 147 #ifdef CONFIG_DEBUG_FS 148 struct dentry *debugfs_dentry_lru, *debugfs_caps; 149 struct dentry *debugfs_congestion_kb; 150 struct dentry *debugfs_bdi; 151 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 152 struct dentry *debugfs_status; 153 struct dentry *debugfs_mds_sessions; 154 struct dentry *debugfs_metrics_dir; 155 #endif 156 157 #ifdef CONFIG_CEPH_FSCACHE 158 struct fscache_volume *fscache; 159 #endif 160 #ifdef CONFIG_FS_ENCRYPTION 161 struct fscrypt_dummy_policy fsc_dummy_enc_policy; 162 #endif 163 }; 164 165 /* 166 * File i/o capability. This tracks shared state with the metadata 167 * server that allows us to cache or writeback attributes or to read 168 * and write data. For any given inode, we should have one or more 169 * capabilities, one issued by each metadata server, and our 170 * cumulative access is the OR of all issued capabilities. 171 * 172 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 173 * session capability lists. 174 */ 175 struct ceph_cap { 176 struct ceph_inode_info *ci; 177 struct rb_node ci_node; /* per-ci cap tree */ 178 struct ceph_mds_session *session; 179 struct list_head session_caps; /* per-session caplist */ 180 u64 cap_id; /* unique cap id (mds provided) */ 181 union { 182 /* in-use caps */ 183 struct { 184 int issued; /* latest, from the mds */ 185 int implemented; /* implemented superset of 186 issued (for revocation) */ 187 int mds; /* mds index for this cap */ 188 int mds_wanted; /* caps wanted from this mds */ 189 }; 190 /* caps to release */ 191 struct { 192 u64 cap_ino; 193 int queue_release; 194 }; 195 }; 196 u32 seq, issue_seq, mseq; 197 u32 cap_gen; /* active/stale cycle */ 198 unsigned long last_used; 199 struct list_head caps_item; 200 }; 201 202 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */ 203 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */ 204 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */ 205 206 struct ceph_cap_flush { 207 u64 tid; 208 int caps; 209 bool wake; /* wake up flush waiters when finish ? */ 210 bool is_capsnap; /* true means capsnap */ 211 struct list_head g_list; // global 212 struct list_head i_list; // per inode 213 }; 214 215 /* 216 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 217 * we first complete any in-process sync writes and writeback any dirty 218 * data before flushing the snapped state (tracked here) back to the MDS. 219 */ 220 struct ceph_cap_snap { 221 refcount_t nref; 222 struct list_head ci_item; 223 224 struct ceph_cap_flush cap_flush; 225 226 u64 follows; 227 int issued, dirty; 228 struct ceph_snap_context *context; 229 230 umode_t mode; 231 kuid_t uid; 232 kgid_t gid; 233 234 struct ceph_buffer *xattr_blob; 235 u64 xattr_version; 236 237 u64 size; 238 u64 change_attr; 239 struct timespec64 mtime, atime, ctime, btime; 240 u64 time_warp_seq; 241 u64 truncate_size; 242 u32 truncate_seq; 243 int writing; /* a sync write is still in progress */ 244 int dirty_pages; /* dirty pages awaiting writeback */ 245 bool inline_data; 246 bool need_flush; 247 }; 248 249 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 250 { 251 if (refcount_dec_and_test(&capsnap->nref)) { 252 if (capsnap->xattr_blob) 253 ceph_buffer_put(capsnap->xattr_blob); 254 kmem_cache_free(ceph_cap_snap_cachep, capsnap); 255 } 256 } 257 258 /* 259 * The frag tree describes how a directory is fragmented, potentially across 260 * multiple metadata servers. It is also used to indicate points where 261 * metadata authority is delegated, and whether/where metadata is replicated. 262 * 263 * A _leaf_ frag will be present in the i_fragtree IFF there is 264 * delegation info. That is, if mds >= 0 || ndist > 0. 265 */ 266 #define CEPH_MAX_DIRFRAG_REP 4 267 268 struct ceph_inode_frag { 269 struct rb_node node; 270 271 /* fragtree state */ 272 u32 frag; 273 int split_by; /* i.e. 2^(split_by) children */ 274 275 /* delegation and replication info */ 276 int mds; /* -1 if same authority as parent */ 277 int ndist; /* >0 if replicated */ 278 int dist[CEPH_MAX_DIRFRAG_REP]; 279 }; 280 281 /* 282 * We cache inode xattrs as an encoded blob until they are first used, 283 * at which point we parse them into an rbtree. 284 */ 285 struct ceph_inode_xattr { 286 struct rb_node node; 287 288 const char *name; 289 int name_len; 290 const char *val; 291 int val_len; 292 int dirty; 293 294 int should_free_name; 295 int should_free_val; 296 }; 297 298 /* 299 * Ceph dentry state 300 */ 301 struct ceph_dentry_info { 302 struct dentry *dentry; 303 struct ceph_mds_session *lease_session; 304 struct list_head lease_list; 305 struct hlist_node hnode; 306 unsigned long flags; 307 int lease_shared_gen; 308 u32 lease_gen; 309 u32 lease_seq; 310 unsigned long lease_renew_after, lease_renew_from; 311 unsigned long time; 312 u64 offset; 313 }; 314 315 #define CEPH_DENTRY_REFERENCED (1 << 0) 316 #define CEPH_DENTRY_LEASE_LIST (1 << 1) 317 #define CEPH_DENTRY_SHRINK_LIST (1 << 2) 318 #define CEPH_DENTRY_PRIMARY_LINK (1 << 3) 319 #define CEPH_DENTRY_ASYNC_UNLINK_BIT (4) 320 #define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT) 321 #define CEPH_DENTRY_ASYNC_CREATE_BIT (5) 322 #define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT) 323 324 struct ceph_inode_xattrs_info { 325 /* 326 * (still encoded) xattr blob. we avoid the overhead of parsing 327 * this until someone actually calls getxattr, etc. 328 * 329 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 330 * NULL means we don't know. 331 */ 332 struct ceph_buffer *blob, *prealloc_blob; 333 334 struct rb_root index; 335 bool dirty; 336 int count; 337 int names_size; 338 int vals_size; 339 u64 version, index_version; 340 }; 341 342 /* 343 * Ceph inode. 344 */ 345 struct ceph_inode_info { 346 struct netfs_inode netfs; /* Netfslib context and vfs inode */ 347 struct ceph_vino i_vino; /* ceph ino + snap */ 348 349 spinlock_t i_ceph_lock; 350 351 u64 i_version; 352 u64 i_inline_version; 353 u32 i_time_warp_seq; 354 355 unsigned long i_ceph_flags; 356 atomic64_t i_release_count; 357 atomic64_t i_ordered_count; 358 atomic64_t i_complete_seq[2]; 359 360 struct ceph_dir_layout i_dir_layout; 361 struct ceph_file_layout i_layout; 362 struct ceph_file_layout i_cached_layout; // for async creates 363 char *i_symlink; 364 365 /* for dirs */ 366 struct timespec64 i_rctime; 367 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps; 368 u64 i_files, i_subdirs; 369 370 /* quotas */ 371 u64 i_max_bytes, i_max_files; 372 373 s32 i_dir_pin; 374 375 struct rb_root i_fragtree; 376 int i_fragtree_nsplits; 377 struct mutex i_fragtree_mutex; 378 379 struct ceph_inode_xattrs_info i_xattrs; 380 381 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 382 * s_mutex. */ 383 struct rb_root i_caps; /* cap list */ 384 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 385 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 386 387 /* 388 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty 389 * is protected by the mdsc->cap_dirty_lock, but each individual item 390 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty 391 * requires the mdsc->cap_dirty_lock. List presence for an item can 392 * be tested under the i_ceph_lock. Changing anything requires both. 393 */ 394 struct list_head i_dirty_item; 395 396 /* 397 * Link to session's s_cap_flushing list. Protected in a similar 398 * fashion to i_dirty_item, but also by the s_mutex for changes. The 399 * s_cap_flushing list can be walked while holding either the s_mutex 400 * or msdc->cap_dirty_lock. List presence can also be checked while 401 * holding the i_ceph_lock for this inode. 402 */ 403 struct list_head i_flushing_item; 404 405 /* we need to track cap writeback on a per-cap-bit basis, to allow 406 * overlapping, pipelined cap flushes to the mds. we can probably 407 * reduce the tid to 8 bits if we're concerned about inode size. */ 408 struct ceph_cap_flush *i_prealloc_cap_flush; 409 struct list_head i_cap_flush_list; 410 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 411 unsigned long i_hold_caps_max; /* jiffies */ 412 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 413 struct ceph_cap_reservation i_cap_migration_resv; 414 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 415 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 416 dirty|flushing caps */ 417 unsigned i_snap_caps; /* cap bits for snapped files */ 418 419 unsigned long i_last_rd; 420 unsigned long i_last_wr; 421 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ 422 423 struct mutex i_truncate_mutex; 424 u32 i_truncate_seq; /* last truncate to smaller size */ 425 u64 i_truncate_size; /* and the size we last truncated down to */ 426 int i_truncate_pending; /* still need to call vmtruncate */ 427 428 u64 i_max_size; /* max file size authorized by mds */ 429 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 430 u64 i_wanted_max_size; /* offset we'd like to write too */ 431 u64 i_requested_max_size; /* max_size we've requested */ 432 433 /* held references to caps */ 434 int i_pin_ref; 435 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; 436 int i_wrbuffer_ref, i_wrbuffer_ref_head; 437 atomic_t i_filelock_ref; 438 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ 439 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 440 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 441 442 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 443 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 444 spinlock_t i_unsafe_lock; 445 446 union { 447 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 448 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ 449 }; 450 struct list_head i_snap_realm_item; 451 struct list_head i_snap_flush_item; 452 struct timespec64 i_btime; 453 struct timespec64 i_snap_btime; 454 455 struct work_struct i_work; 456 unsigned long i_work_mask; 457 458 #ifdef CONFIG_FS_ENCRYPTION 459 u32 fscrypt_auth_len; 460 u32 fscrypt_file_len; 461 u8 *fscrypt_auth; 462 u8 *fscrypt_file; 463 #endif 464 }; 465 466 struct ceph_netfs_request_data { 467 int caps; 468 469 /* 470 * Maximum size of a file readahead request. 471 * The fadvise could update the bdi's default ra_pages. 472 */ 473 unsigned int file_ra_pages; 474 475 /* Set it if fadvise disables file readahead entirely */ 476 bool file_ra_disabled; 477 }; 478 479 static inline struct ceph_inode_info * 480 ceph_inode(const struct inode *inode) 481 { 482 return container_of(inode, struct ceph_inode_info, netfs.inode); 483 } 484 485 static inline struct ceph_fs_client * 486 ceph_inode_to_client(const struct inode *inode) 487 { 488 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 489 } 490 491 static inline struct ceph_fs_client * 492 ceph_sb_to_client(const struct super_block *sb) 493 { 494 return (struct ceph_fs_client *)sb->s_fs_info; 495 } 496 497 static inline struct ceph_mds_client * 498 ceph_sb_to_mdsc(const struct super_block *sb) 499 { 500 return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc; 501 } 502 503 static inline struct ceph_vino 504 ceph_vino(const struct inode *inode) 505 { 506 return ceph_inode(inode)->i_vino; 507 } 508 509 static inline u32 ceph_ino_to_ino32(u64 vino) 510 { 511 u32 ino = vino & 0xffffffff; 512 ino ^= vino >> 32; 513 if (!ino) 514 ino = 2; 515 return ino; 516 } 517 518 /* 519 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on 520 * some arches. We generally do not use this value inside the ceph driver, but 521 * we do want to set it to something, so that generic vfs code has an 522 * appropriate value for tracepoints and the like. 523 */ 524 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino) 525 { 526 if (sizeof(ino_t) == sizeof(u32)) 527 return ceph_ino_to_ino32(vino.ino); 528 return (ino_t)vino.ino; 529 } 530 531 /* for printf-style formatting */ 532 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 533 534 static inline u64 ceph_ino(struct inode *inode) 535 { 536 return ceph_inode(inode)->i_vino.ino; 537 } 538 539 static inline u64 ceph_snap(struct inode *inode) 540 { 541 return ceph_inode(inode)->i_vino.snap; 542 } 543 544 /** 545 * ceph_present_ino - format an inode number for presentation to userland 546 * @sb: superblock where the inode lives 547 * @ino: inode number to (possibly) convert 548 * 549 * If the user mounted with the ino32 option, then the 64-bit value needs 550 * to be converted to something that can fit inside 32 bits. Note that 551 * internal kernel code never uses this value, so this is entirely for 552 * userland consumption. 553 */ 554 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino) 555 { 556 if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))) 557 return ceph_ino_to_ino32(ino); 558 return ino; 559 } 560 561 static inline u64 ceph_present_inode(struct inode *inode) 562 { 563 return ceph_present_ino(inode->i_sb, ceph_ino(inode)); 564 } 565 566 static inline int ceph_ino_compare(struct inode *inode, void *data) 567 { 568 struct ceph_vino *pvino = (struct ceph_vino *)data; 569 struct ceph_inode_info *ci = ceph_inode(inode); 570 return ci->i_vino.ino == pvino->ino && 571 ci->i_vino.snap == pvino->snap; 572 } 573 574 /* 575 * The MDS reserves a set of inodes for its own usage. These should never 576 * be accessible by clients, and so the MDS has no reason to ever hand these 577 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE. 578 * 579 * These come from src/mds/mdstypes.h in the ceph sources. 580 */ 581 #define CEPH_MAX_MDS 0x100 582 #define CEPH_NUM_STRAY 10 583 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS) 584 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS) 585 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY)) 586 587 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino) 588 { 589 if (vino.ino >= CEPH_INO_SYSTEM_BASE || 590 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET) 591 return false; 592 593 /* Don't warn on mdsdirs */ 594 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET, 595 "Attempt to access reserved inode number 0x%llx", 596 vino.ino); 597 return true; 598 } 599 600 static inline struct inode *ceph_find_inode(struct super_block *sb, 601 struct ceph_vino vino) 602 { 603 if (ceph_vino_is_reserved(vino)) 604 return NULL; 605 606 /* 607 * NB: The hashval will be run through the fs/inode.c hash function 608 * anyway, so there is no need to squash the inode number down to 609 * 32-bits first. Just use low-order bits on arches with 32-bit long. 610 */ 611 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino); 612 } 613 614 615 /* 616 * Ceph inode. 617 */ 618 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 619 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 620 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */ 621 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */ 622 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */ 623 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */ 624 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */ 625 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */ 626 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */ 627 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */ 628 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */ 629 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */ 630 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT) 631 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */ 632 #define CEPH_I_ASYNC_CHECK_CAPS (1 << 14) /* check caps immediately after async 633 creating finishes */ 634 635 /* 636 * Masks of ceph inode work. 637 */ 638 #define CEPH_I_WORK_WRITEBACK 0 639 #define CEPH_I_WORK_INVALIDATE_PAGES 1 640 #define CEPH_I_WORK_VMTRUNCATE 2 641 #define CEPH_I_WORK_CHECK_CAPS 3 642 #define CEPH_I_WORK_FLUSH_SNAPS 4 643 644 /* 645 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 646 * and then clear it when they start succeeding. Note that we do a lockless 647 * check first, and only take the lock if it looks like it needs to be changed. 648 * The write submission code just takes this as a hint, so we're not too 649 * worried if a few slip through in either direction. 650 */ 651 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 652 { 653 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { 654 spin_lock(&ci->i_ceph_lock); 655 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; 656 spin_unlock(&ci->i_ceph_lock); 657 } 658 } 659 660 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 661 { 662 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { 663 spin_lock(&ci->i_ceph_lock); 664 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; 665 spin_unlock(&ci->i_ceph_lock); 666 } 667 } 668 669 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 670 long long release_count, 671 long long ordered_count) 672 { 673 /* 674 * Makes sure operations that setup readdir cache (update page 675 * cache and i_size) are strongly ordered w.r.t. the following 676 * atomic64_set() operations. 677 */ 678 smp_mb(); 679 atomic64_set(&ci->i_complete_seq[0], release_count); 680 atomic64_set(&ci->i_complete_seq[1], ordered_count); 681 } 682 683 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 684 { 685 atomic64_inc(&ci->i_release_count); 686 } 687 688 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 689 { 690 atomic64_inc(&ci->i_ordered_count); 691 } 692 693 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 694 { 695 return atomic64_read(&ci->i_complete_seq[0]) == 696 atomic64_read(&ci->i_release_count); 697 } 698 699 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 700 { 701 return atomic64_read(&ci->i_complete_seq[0]) == 702 atomic64_read(&ci->i_release_count) && 703 atomic64_read(&ci->i_complete_seq[1]) == 704 atomic64_read(&ci->i_ordered_count); 705 } 706 707 static inline void ceph_dir_clear_complete(struct inode *inode) 708 { 709 __ceph_dir_clear_complete(ceph_inode(inode)); 710 } 711 712 static inline void ceph_dir_clear_ordered(struct inode *inode) 713 { 714 __ceph_dir_clear_ordered(ceph_inode(inode)); 715 } 716 717 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 718 { 719 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 720 smp_rmb(); 721 return ret; 722 } 723 724 /* find a specific frag @f */ 725 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 726 u32 f); 727 728 /* 729 * choose fragment for value @v. copy frag content to pfrag, if leaf 730 * exists 731 */ 732 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 733 struct ceph_inode_frag *pfrag, 734 int *found); 735 736 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) 737 { 738 return (struct ceph_dentry_info *)dentry->d_fsdata; 739 } 740 741 /* 742 * caps helpers 743 */ 744 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 745 { 746 return !RB_EMPTY_ROOT(&ci->i_caps); 747 } 748 749 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 750 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 751 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, 752 int t); 753 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 754 struct ceph_cap *cap); 755 756 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 757 { 758 int issued; 759 spin_lock(&ci->i_ceph_lock); 760 issued = __ceph_caps_issued(ci, NULL); 761 spin_unlock(&ci->i_ceph_lock); 762 return issued; 763 } 764 765 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, 766 int mask, int touch) 767 { 768 int r; 769 spin_lock(&ci->i_ceph_lock); 770 r = __ceph_caps_issued_mask_metric(ci, mask, touch); 771 spin_unlock(&ci->i_ceph_lock); 772 return r; 773 } 774 775 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 776 { 777 return ci->i_dirty_caps | ci->i_flushing_caps; 778 } 779 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 780 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 781 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 782 struct ceph_cap_flush **pcf); 783 784 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 785 struct ceph_cap *ocap, int mask); 786 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 787 extern int __ceph_caps_used(struct ceph_inode_info *ci); 788 789 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci) 790 { 791 return ci->i_nr_by_mode[0]; 792 } 793 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 794 extern int __ceph_caps_wanted(struct ceph_inode_info *ci); 795 796 /* what the mds thinks we want */ 797 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 798 799 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 800 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 801 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, 802 struct ceph_mount_options *fsopt); 803 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 804 struct ceph_cap_reservation *ctx, int need); 805 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, 806 struct ceph_cap_reservation *ctx); 807 extern void ceph_reservation_status(struct ceph_fs_client *client, 808 int *total, int *avail, int *used, 809 int *reserved, int *min); 810 extern void change_auth_cap_ses(struct ceph_inode_info *ci, 811 struct ceph_mds_session *session); 812 813 814 815 /* 816 * we keep buffered readdir results attached to file->private_data 817 */ 818 #define CEPH_F_SYNC 1 819 #define CEPH_F_ATEND 2 820 821 struct ceph_file_info { 822 short fmode; /* initialized on open */ 823 short flags; /* CEPH_F_* */ 824 825 spinlock_t rw_contexts_lock; 826 struct list_head rw_contexts; 827 828 u32 filp_gen; 829 }; 830 831 struct ceph_dir_file_info { 832 struct ceph_file_info file_info; 833 834 /* readdir: position within the dir */ 835 u32 frag; 836 struct ceph_mds_request *last_readdir; 837 838 /* readdir: position within a frag */ 839 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 840 char *last_name; /* last entry in previous chunk */ 841 long long dir_release_count; 842 long long dir_ordered_count; 843 int readdir_cache_idx; 844 845 /* used for -o dirstat read() on directory thing */ 846 char *dir_info; 847 int dir_info_len; 848 }; 849 850 struct ceph_rw_context { 851 struct list_head list; 852 struct task_struct *thread; 853 int caps; 854 }; 855 856 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ 857 struct ceph_rw_context _name = { \ 858 .thread = current, \ 859 .caps = _caps, \ 860 } 861 862 static inline void ceph_add_rw_context(struct ceph_file_info *cf, 863 struct ceph_rw_context *ctx) 864 { 865 spin_lock(&cf->rw_contexts_lock); 866 list_add(&ctx->list, &cf->rw_contexts); 867 spin_unlock(&cf->rw_contexts_lock); 868 } 869 870 static inline void ceph_del_rw_context(struct ceph_file_info *cf, 871 struct ceph_rw_context *ctx) 872 { 873 spin_lock(&cf->rw_contexts_lock); 874 list_del(&ctx->list); 875 spin_unlock(&cf->rw_contexts_lock); 876 } 877 878 static inline struct ceph_rw_context* 879 ceph_find_rw_context(struct ceph_file_info *cf) 880 { 881 struct ceph_rw_context *ctx, *found = NULL; 882 spin_lock(&cf->rw_contexts_lock); 883 list_for_each_entry(ctx, &cf->rw_contexts, list) { 884 if (ctx->thread == current) { 885 found = ctx; 886 break; 887 } 888 } 889 spin_unlock(&cf->rw_contexts_lock); 890 return found; 891 } 892 893 struct ceph_readdir_cache_control { 894 struct page *page; 895 struct dentry **dentries; 896 int index; 897 }; 898 899 /* 900 * A "snap realm" describes a subset of the file hierarchy sharing 901 * the same set of snapshots that apply to it. The realms themselves 902 * are organized into a hierarchy, such that children inherit (some of) 903 * the snapshots of their parents. 904 * 905 * All inodes within the realm that have capabilities are linked into a 906 * per-realm list. 907 */ 908 struct ceph_snap_realm { 909 u64 ino; 910 struct inode *inode; 911 atomic_t nref; 912 struct rb_node node; 913 914 u64 created, seq; 915 u64 parent_ino; 916 u64 parent_since; /* snapid when our current parent became so */ 917 918 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 919 u32 num_prior_parent_snaps; /* had prior to parent_since */ 920 u64 *snaps; /* snaps specific to this realm */ 921 u32 num_snaps; 922 923 struct ceph_snap_realm *parent; 924 struct list_head children; /* list of child realms */ 925 struct list_head child_item; 926 927 struct list_head empty_item; /* if i have ref==0 */ 928 929 struct list_head dirty_item; /* if realm needs new context */ 930 931 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */ 932 933 /* the current set of snaps for this realm */ 934 struct ceph_snap_context *cached_context; 935 936 struct list_head inodes_with_caps; 937 spinlock_t inodes_with_caps_lock; 938 }; 939 940 static inline int default_congestion_kb(void) 941 { 942 int congestion_kb; 943 944 /* 945 * Copied from NFS 946 * 947 * congestion size, scale with available memory. 948 * 949 * 64MB: 8192k 950 * 128MB: 11585k 951 * 256MB: 16384k 952 * 512MB: 23170k 953 * 1GB: 32768k 954 * 2GB: 46340k 955 * 4GB: 65536k 956 * 8GB: 92681k 957 * 16GB: 131072k 958 * 959 * This allows larger machines to have larger/more transfers. 960 * Limit the default to 256M 961 */ 962 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); 963 if (congestion_kb > 256*1024) 964 congestion_kb = 256*1024; 965 966 return congestion_kb; 967 } 968 969 970 /* super.c */ 971 extern int ceph_force_reconnect(struct super_block *sb); 972 /* snap.c */ 973 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 974 u64 ino); 975 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 976 struct ceph_snap_realm *realm); 977 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 978 struct ceph_snap_realm *realm); 979 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 980 void *p, void *e, bool deletion, 981 struct ceph_snap_realm **realm_ret); 982 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm); 983 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 984 struct ceph_mds_session *session, 985 struct ceph_msg *msg); 986 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 987 struct ceph_cap_snap *capsnap); 988 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc); 989 990 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, 991 u64 snap); 992 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, 993 struct ceph_snapid_map *sm); 994 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); 995 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); 996 void ceph_umount_begin(struct super_block *sb); 997 998 999 /* 1000 * a cap_snap is "pending" if it is still awaiting an in-progress 1001 * sync write (that may/may not still update size, mtime, etc.). 1002 */ 1003 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 1004 { 1005 return !list_empty(&ci->i_cap_snaps) && 1006 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 1007 ci_item)->writing; 1008 } 1009 1010 /* inode.c */ 1011 struct ceph_mds_reply_info_in; 1012 struct ceph_mds_reply_dirfrag; 1013 struct ceph_acl_sec_ctx; 1014 1015 extern const struct inode_operations ceph_file_iops; 1016 1017 extern struct inode *ceph_alloc_inode(struct super_block *sb); 1018 extern void ceph_evict_inode(struct inode *inode); 1019 extern void ceph_free_inode(struct inode *inode); 1020 1021 struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry, 1022 umode_t *mode, struct ceph_acl_sec_ctx *as_ctx); 1023 void ceph_as_ctx_to_req(struct ceph_mds_request *req, 1024 struct ceph_acl_sec_ctx *as_ctx); 1025 1026 extern struct inode *ceph_get_inode(struct super_block *sb, 1027 struct ceph_vino vino, 1028 struct inode *newino); 1029 extern struct inode *ceph_get_snapdir(struct inode *parent); 1030 extern int ceph_fill_file_size(struct inode *inode, int issued, 1031 u32 truncate_seq, u64 truncate_size, u64 size); 1032 extern void ceph_fill_file_time(struct inode *inode, int issued, 1033 u64 time_warp_seq, struct timespec64 *ctime, 1034 struct timespec64 *mtime, 1035 struct timespec64 *atime); 1036 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page, 1037 struct ceph_mds_reply_info_in *iinfo, 1038 struct ceph_mds_reply_dirfrag *dirinfo, 1039 struct ceph_mds_session *session, int cap_fmode, 1040 struct ceph_cap_reservation *caps_reservation); 1041 extern int ceph_fill_trace(struct super_block *sb, 1042 struct ceph_mds_request *req); 1043 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 1044 struct ceph_mds_session *session); 1045 1046 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 1047 1048 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 1049 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 1050 1051 void ceph_queue_inode_work(struct inode *inode, int work_bit); 1052 1053 static inline void ceph_queue_vmtruncate(struct inode *inode) 1054 { 1055 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE); 1056 } 1057 1058 static inline void ceph_queue_invalidate(struct inode *inode) 1059 { 1060 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES); 1061 } 1062 1063 static inline void ceph_queue_writeback(struct inode *inode) 1064 { 1065 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK); 1066 } 1067 1068 static inline void ceph_queue_check_caps(struct inode *inode) 1069 { 1070 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS); 1071 } 1072 1073 static inline void ceph_queue_flush_snaps(struct inode *inode) 1074 { 1075 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS); 1076 } 1077 1078 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask); 1079 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 1080 int mask, bool force); 1081 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 1082 { 1083 return __ceph_do_getattr(inode, NULL, mask, force); 1084 } 1085 extern int ceph_permission(struct mnt_idmap *idmap, 1086 struct inode *inode, int mask); 1087 1088 struct ceph_iattr { 1089 struct ceph_fscrypt_auth *fscrypt_auth; 1090 }; 1091 1092 extern int __ceph_setattr(struct inode *inode, struct iattr *attr, 1093 struct ceph_iattr *cia); 1094 extern int ceph_setattr(struct mnt_idmap *idmap, 1095 struct dentry *dentry, struct iattr *attr); 1096 extern int ceph_getattr(struct mnt_idmap *idmap, 1097 const struct path *path, struct kstat *stat, 1098 u32 request_mask, unsigned int flags); 1099 void ceph_inode_shutdown(struct inode *inode); 1100 1101 static inline bool ceph_inode_is_shutdown(struct inode *inode) 1102 { 1103 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags); 1104 struct ceph_fs_client *fsc = ceph_inode_to_client(inode); 1105 int state = READ_ONCE(fsc->mount_state); 1106 1107 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN; 1108 } 1109 1110 /* xattr.c */ 1111 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 1112 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size); 1113 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 1114 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 1115 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); 1116 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 1117 extern const struct xattr_handler *ceph_xattr_handlers[]; 1118 1119 struct ceph_acl_sec_ctx { 1120 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1121 void *default_acl; 1122 void *acl; 1123 #endif 1124 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1125 void *sec_ctx; 1126 u32 sec_ctxlen; 1127 #endif 1128 #ifdef CONFIG_FS_ENCRYPTION 1129 struct ceph_fscrypt_auth *fscrypt_auth; 1130 #endif 1131 struct ceph_pagelist *pagelist; 1132 }; 1133 1134 #ifdef CONFIG_SECURITY 1135 extern bool ceph_security_xattr_deadlock(struct inode *in); 1136 extern bool ceph_security_xattr_wanted(struct inode *in); 1137 #else 1138 static inline bool ceph_security_xattr_deadlock(struct inode *in) 1139 { 1140 return false; 1141 } 1142 static inline bool ceph_security_xattr_wanted(struct inode *in) 1143 { 1144 return false; 1145 } 1146 #endif 1147 1148 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1149 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1150 struct ceph_acl_sec_ctx *ctx); 1151 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1152 { 1153 security_inode_invalidate_secctx(inode); 1154 } 1155 #else 1156 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1157 struct ceph_acl_sec_ctx *ctx) 1158 { 1159 return 0; 1160 } 1161 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1162 { 1163 } 1164 #endif 1165 1166 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); 1167 1168 /* acl.c */ 1169 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1170 1171 struct posix_acl *ceph_get_acl(struct inode *, int, bool); 1172 int ceph_set_acl(struct mnt_idmap *idmap, 1173 struct dentry *dentry, struct posix_acl *acl, int type); 1174 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1175 struct ceph_acl_sec_ctx *as_ctx); 1176 void ceph_init_inode_acls(struct inode *inode, 1177 struct ceph_acl_sec_ctx *as_ctx); 1178 1179 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1180 { 1181 forget_all_cached_acls(inode); 1182 } 1183 1184 #else 1185 1186 #define ceph_get_acl NULL 1187 #define ceph_set_acl NULL 1188 1189 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1190 struct ceph_acl_sec_ctx *as_ctx) 1191 { 1192 return 0; 1193 } 1194 static inline void ceph_init_inode_acls(struct inode *inode, 1195 struct ceph_acl_sec_ctx *as_ctx) 1196 { 1197 } 1198 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 1199 { 1200 return 0; 1201 } 1202 1203 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1204 { 1205 } 1206 1207 #endif 1208 1209 /* caps.c */ 1210 extern const char *ceph_cap_string(int c); 1211 extern void ceph_handle_caps(struct ceph_mds_session *session, 1212 struct ceph_msg *msg); 1213 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 1214 struct ceph_cap_reservation *ctx); 1215 extern void ceph_add_cap(struct inode *inode, 1216 struct ceph_mds_session *session, u64 cap_id, 1217 unsigned issued, unsigned wanted, 1218 unsigned cap, unsigned seq, u64 realmino, int flags, 1219 struct ceph_cap **new_cap); 1220 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1221 extern void ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1222 extern void __ceph_remove_caps(struct ceph_inode_info *ci); 1223 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 1224 struct ceph_cap *cap); 1225 extern int ceph_is_any_caps(struct inode *inode); 1226 1227 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 1228 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 1229 int datasync); 1230 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 1231 struct ceph_mds_session *session); 1232 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 1233 struct ceph_mds_session *session); 1234 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, 1235 struct ceph_inode_info *ci); 1236 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, 1237 int mds); 1238 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 1239 int mds); 1240 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, 1241 bool snap_rwsem_locked); 1242 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 1243 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 1244 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had); 1245 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, 1246 int had); 1247 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 1248 struct ceph_snap_context *snapc); 1249 extern void __ceph_remove_capsnap(struct inode *inode, 1250 struct ceph_cap_snap *capsnap, 1251 bool *wake_ci, bool *wake_mdsc); 1252 extern void ceph_remove_capsnap(struct inode *inode, 1253 struct ceph_cap_snap *capsnap, 1254 bool *wake_ci, bool *wake_mdsc); 1255 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 1256 struct ceph_mds_session **psession); 1257 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 1258 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags); 1259 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 1260 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 1261 extern int ceph_drop_caps_for_unlink(struct inode *inode); 1262 extern int ceph_encode_inode_release(void **p, struct inode *inode, 1263 int mds, int drop, int unless, int force); 1264 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 1265 struct inode *dir, 1266 int mds, int drop, int unless); 1267 1268 extern int ceph_get_caps(struct file *filp, int need, int want, 1269 loff_t endoff, int *got); 1270 extern int ceph_try_get_caps(struct inode *inode, 1271 int need, int want, bool nonblock, int *got); 1272 1273 /* for counting open files by mode */ 1274 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count); 1275 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count); 1276 extern void __ceph_touch_fmode(struct ceph_inode_info *ci, 1277 struct ceph_mds_client *mdsc, int fmode); 1278 1279 /* addr.c */ 1280 extern const struct address_space_operations ceph_aops; 1281 extern const struct netfs_request_ops ceph_netfs_ops; 1282 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 1283 extern int ceph_uninline_data(struct file *file); 1284 extern int ceph_pool_perm_check(struct inode *inode, int need); 1285 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 1286 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate); 1287 1288 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci) 1289 { 1290 if (ci->i_inline_version == CEPH_INLINE_NONE || 1291 ci->i_inline_version == 1) /* initial version, no data */ 1292 return false; 1293 return true; 1294 } 1295 1296 /* file.c */ 1297 extern const struct file_operations ceph_file_fops; 1298 1299 extern int ceph_renew_caps(struct inode *inode, int fmode); 1300 extern int ceph_open(struct inode *inode, struct file *file); 1301 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 1302 struct file *file, unsigned flags, umode_t mode); 1303 extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos, 1304 struct iov_iter *to, int *retry_op, 1305 u64 *last_objver); 1306 extern int ceph_release(struct inode *inode, struct file *filp); 1307 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 1308 char *data, size_t len); 1309 1310 /* dir.c */ 1311 extern const struct file_operations ceph_dir_fops; 1312 extern const struct file_operations ceph_snapdir_fops; 1313 extern const struct inode_operations ceph_dir_iops; 1314 extern const struct inode_operations ceph_snapdir_iops; 1315 extern const struct dentry_operations ceph_dentry_ops; 1316 1317 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 1318 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 1319 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req, 1320 struct dentry *dentry); 1321 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 1322 struct dentry *dentry, int err); 1323 1324 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); 1325 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); 1326 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 1327 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); 1328 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 1329 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 1330 1331 /* ioctl.c */ 1332 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1333 1334 /* export.c */ 1335 extern const struct export_operations ceph_export_ops; 1336 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); 1337 1338 /* locks.c */ 1339 extern __init void ceph_flock_init(void); 1340 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 1341 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 1342 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 1343 extern int ceph_encode_locks_to_buffer(struct inode *inode, 1344 struct ceph_filelock *flocks, 1345 int num_fcntl_locks, 1346 int num_flock_locks); 1347 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 1348 struct ceph_pagelist *pagelist, 1349 int num_fcntl_locks, int num_flock_locks); 1350 1351 /* debugfs.c */ 1352 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); 1353 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1354 1355 /* quota.c */ 1356 1357 enum quota_get_realm { 1358 QUOTA_GET_MAX_FILES, 1359 QUOTA_GET_MAX_BYTES, 1360 QUOTA_GET_ANY 1361 }; 1362 1363 static inline bool __ceph_has_quota(struct ceph_inode_info *ci, 1364 enum quota_get_realm which) 1365 { 1366 bool has_quota = false; 1367 1368 switch (which) { 1369 case QUOTA_GET_MAX_BYTES: 1370 has_quota = !!ci->i_max_bytes; 1371 break; 1372 case QUOTA_GET_MAX_FILES: 1373 has_quota = !!ci->i_max_files; 1374 break; 1375 default: 1376 has_quota = !!(ci->i_max_files || ci->i_max_bytes); 1377 } 1378 return has_quota; 1379 } 1380 1381 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); 1382 1383 static inline void __ceph_update_quota(struct ceph_inode_info *ci, 1384 u64 max_bytes, u64 max_files) 1385 { 1386 bool had_quota, has_quota; 1387 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1388 ci->i_max_bytes = max_bytes; 1389 ci->i_max_files = max_files; 1390 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1391 1392 if (had_quota != has_quota) 1393 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota); 1394 } 1395 1396 extern void ceph_handle_quota(struct ceph_mds_client *mdsc, 1397 struct ceph_mds_session *session, 1398 struct ceph_msg *msg); 1399 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); 1400 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); 1401 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, 1402 loff_t newlen); 1403 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, 1404 loff_t newlen); 1405 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, 1406 struct kstatfs *buf); 1407 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); 1408 1409 #endif /* _FS_CEPH_SUPER_H */ 1410