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 * For none fscrypt case it equals to i_truncate_size or it will 429 * equals to fscrypt_file_size 430 */ 431 u64 i_truncate_pagecache_size; 432 433 u64 i_max_size; /* max file size authorized by mds */ 434 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 435 u64 i_wanted_max_size; /* offset we'd like to write too */ 436 u64 i_requested_max_size; /* max_size we've requested */ 437 438 /* held references to caps */ 439 int i_pin_ref; 440 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; 441 int i_wrbuffer_ref, i_wrbuffer_ref_head; 442 atomic_t i_filelock_ref; 443 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ 444 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 445 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 446 447 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 448 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 449 spinlock_t i_unsafe_lock; 450 451 union { 452 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 453 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ 454 }; 455 struct list_head i_snap_realm_item; 456 struct list_head i_snap_flush_item; 457 struct timespec64 i_btime; 458 struct timespec64 i_snap_btime; 459 460 struct work_struct i_work; 461 unsigned long i_work_mask; 462 463 #ifdef CONFIG_FS_ENCRYPTION 464 u32 fscrypt_auth_len; 465 u32 fscrypt_file_len; 466 u8 *fscrypt_auth; 467 u8 *fscrypt_file; 468 #endif 469 }; 470 471 struct ceph_netfs_request_data { 472 int caps; 473 474 /* 475 * Maximum size of a file readahead request. 476 * The fadvise could update the bdi's default ra_pages. 477 */ 478 unsigned int file_ra_pages; 479 480 /* Set it if fadvise disables file readahead entirely */ 481 bool file_ra_disabled; 482 }; 483 484 static inline struct ceph_inode_info * 485 ceph_inode(const struct inode *inode) 486 { 487 return container_of(inode, struct ceph_inode_info, netfs.inode); 488 } 489 490 static inline struct ceph_fs_client * 491 ceph_inode_to_fs_client(const struct inode *inode) 492 { 493 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 494 } 495 496 static inline struct ceph_fs_client * 497 ceph_sb_to_fs_client(const struct super_block *sb) 498 { 499 return (struct ceph_fs_client *)sb->s_fs_info; 500 } 501 502 static inline struct ceph_mds_client * 503 ceph_sb_to_mdsc(const struct super_block *sb) 504 { 505 return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc; 506 } 507 508 static inline struct ceph_client * 509 ceph_inode_to_client(const struct inode *inode) 510 { 511 return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client; 512 } 513 514 static inline struct ceph_vino 515 ceph_vino(const struct inode *inode) 516 { 517 return ceph_inode(inode)->i_vino; 518 } 519 520 static inline u32 ceph_ino_to_ino32(u64 vino) 521 { 522 u32 ino = vino & 0xffffffff; 523 ino ^= vino >> 32; 524 if (!ino) 525 ino = 2; 526 return ino; 527 } 528 529 /* 530 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on 531 * some arches. We generally do not use this value inside the ceph driver, but 532 * we do want to set it to something, so that generic vfs code has an 533 * appropriate value for tracepoints and the like. 534 */ 535 static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino) 536 { 537 if (sizeof(ino_t) == sizeof(u32)) 538 return ceph_ino_to_ino32(vino.ino); 539 return (ino_t)vino.ino; 540 } 541 542 /* for printf-style formatting */ 543 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 544 545 static inline u64 ceph_ino(struct inode *inode) 546 { 547 return ceph_inode(inode)->i_vino.ino; 548 } 549 550 static inline u64 ceph_snap(struct inode *inode) 551 { 552 return ceph_inode(inode)->i_vino.snap; 553 } 554 555 /** 556 * ceph_present_ino - format an inode number for presentation to userland 557 * @sb: superblock where the inode lives 558 * @ino: inode number to (possibly) convert 559 * 560 * If the user mounted with the ino32 option, then the 64-bit value needs 561 * to be converted to something that can fit inside 32 bits. Note that 562 * internal kernel code never uses this value, so this is entirely for 563 * userland consumption. 564 */ 565 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino) 566 { 567 if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32))) 568 return ceph_ino_to_ino32(ino); 569 return ino; 570 } 571 572 static inline u64 ceph_present_inode(struct inode *inode) 573 { 574 return ceph_present_ino(inode->i_sb, ceph_ino(inode)); 575 } 576 577 static inline int ceph_ino_compare(struct inode *inode, void *data) 578 { 579 struct ceph_vino *pvino = (struct ceph_vino *)data; 580 struct ceph_inode_info *ci = ceph_inode(inode); 581 return ci->i_vino.ino == pvino->ino && 582 ci->i_vino.snap == pvino->snap; 583 } 584 585 /* 586 * The MDS reserves a set of inodes for its own usage. These should never 587 * be accessible by clients, and so the MDS has no reason to ever hand these 588 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE. 589 * 590 * These come from src/mds/mdstypes.h in the ceph sources. 591 */ 592 #define CEPH_MAX_MDS 0x100 593 #define CEPH_NUM_STRAY 10 594 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS) 595 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS) 596 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY)) 597 598 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino) 599 { 600 if (vino.ino >= CEPH_INO_SYSTEM_BASE || 601 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET) 602 return false; 603 604 /* Don't warn on mdsdirs */ 605 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET, 606 "Attempt to access reserved inode number 0x%llx", 607 vino.ino); 608 return true; 609 } 610 611 static inline struct inode *ceph_find_inode(struct super_block *sb, 612 struct ceph_vino vino) 613 { 614 if (ceph_vino_is_reserved(vino)) 615 return NULL; 616 617 /* 618 * NB: The hashval will be run through the fs/inode.c hash function 619 * anyway, so there is no need to squash the inode number down to 620 * 32-bits first. Just use low-order bits on arches with 32-bit long. 621 */ 622 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino); 623 } 624 625 626 /* 627 * Ceph inode. 628 */ 629 #define CEPH_I_DIR_ORDERED (1 << 0) /* dentries in dir are ordered */ 630 #define CEPH_I_FLUSH (1 << 2) /* do not delay flush of dirty metadata */ 631 #define CEPH_I_POOL_PERM (1 << 3) /* pool rd/wr bits are valid */ 632 #define CEPH_I_POOL_RD (1 << 4) /* can read from pool */ 633 #define CEPH_I_POOL_WR (1 << 5) /* can write to pool */ 634 #define CEPH_I_SEC_INITED (1 << 6) /* security initialized */ 635 #define CEPH_I_KICK_FLUSH (1 << 7) /* kick flushing caps */ 636 #define CEPH_I_FLUSH_SNAPS (1 << 8) /* need flush snapss */ 637 #define CEPH_I_ERROR_WRITE (1 << 9) /* have seen write errors */ 638 #define CEPH_I_ERROR_FILELOCK (1 << 10) /* have seen file lock errors */ 639 #define CEPH_I_ODIRECT (1 << 11) /* inode in direct I/O mode */ 640 #define CEPH_ASYNC_CREATE_BIT (12) /* async create in flight for this */ 641 #define CEPH_I_ASYNC_CREATE (1 << CEPH_ASYNC_CREATE_BIT) 642 #define CEPH_I_SHUTDOWN (1 << 13) /* inode is no longer usable */ 643 #define CEPH_I_ASYNC_CHECK_CAPS (1 << 14) /* check caps immediately after async 644 creating finishes */ 645 646 /* 647 * Masks of ceph inode work. 648 */ 649 #define CEPH_I_WORK_WRITEBACK 0 650 #define CEPH_I_WORK_INVALIDATE_PAGES 1 651 #define CEPH_I_WORK_VMTRUNCATE 2 652 #define CEPH_I_WORK_CHECK_CAPS 3 653 #define CEPH_I_WORK_FLUSH_SNAPS 4 654 655 /* 656 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 657 * and then clear it when they start succeeding. Note that we do a lockless 658 * check first, and only take the lock if it looks like it needs to be changed. 659 * The write submission code just takes this as a hint, so we're not too 660 * worried if a few slip through in either direction. 661 */ 662 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 663 { 664 if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) { 665 spin_lock(&ci->i_ceph_lock); 666 ci->i_ceph_flags |= CEPH_I_ERROR_WRITE; 667 spin_unlock(&ci->i_ceph_lock); 668 } 669 } 670 671 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 672 { 673 if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) { 674 spin_lock(&ci->i_ceph_lock); 675 ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE; 676 spin_unlock(&ci->i_ceph_lock); 677 } 678 } 679 680 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 681 long long release_count, 682 long long ordered_count) 683 { 684 /* 685 * Makes sure operations that setup readdir cache (update page 686 * cache and i_size) are strongly ordered w.r.t. the following 687 * atomic64_set() operations. 688 */ 689 smp_mb(); 690 atomic64_set(&ci->i_complete_seq[0], release_count); 691 atomic64_set(&ci->i_complete_seq[1], ordered_count); 692 } 693 694 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 695 { 696 atomic64_inc(&ci->i_release_count); 697 } 698 699 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 700 { 701 atomic64_inc(&ci->i_ordered_count); 702 } 703 704 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 705 { 706 return atomic64_read(&ci->i_complete_seq[0]) == 707 atomic64_read(&ci->i_release_count); 708 } 709 710 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 711 { 712 return atomic64_read(&ci->i_complete_seq[0]) == 713 atomic64_read(&ci->i_release_count) && 714 atomic64_read(&ci->i_complete_seq[1]) == 715 atomic64_read(&ci->i_ordered_count); 716 } 717 718 static inline void ceph_dir_clear_complete(struct inode *inode) 719 { 720 __ceph_dir_clear_complete(ceph_inode(inode)); 721 } 722 723 static inline void ceph_dir_clear_ordered(struct inode *inode) 724 { 725 __ceph_dir_clear_ordered(ceph_inode(inode)); 726 } 727 728 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 729 { 730 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 731 smp_rmb(); 732 return ret; 733 } 734 735 /* find a specific frag @f */ 736 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 737 u32 f); 738 739 /* 740 * choose fragment for value @v. copy frag content to pfrag, if leaf 741 * exists 742 */ 743 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 744 struct ceph_inode_frag *pfrag, 745 int *found); 746 747 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) 748 { 749 return (struct ceph_dentry_info *)dentry->d_fsdata; 750 } 751 752 /* 753 * caps helpers 754 */ 755 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 756 { 757 return !RB_EMPTY_ROOT(&ci->i_caps); 758 } 759 760 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 761 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 762 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, 763 int t); 764 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 765 struct ceph_cap *cap); 766 767 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 768 { 769 int issued; 770 spin_lock(&ci->i_ceph_lock); 771 issued = __ceph_caps_issued(ci, NULL); 772 spin_unlock(&ci->i_ceph_lock); 773 return issued; 774 } 775 776 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, 777 int mask, int touch) 778 { 779 int r; 780 spin_lock(&ci->i_ceph_lock); 781 r = __ceph_caps_issued_mask_metric(ci, mask, touch); 782 spin_unlock(&ci->i_ceph_lock); 783 return r; 784 } 785 786 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 787 { 788 return ci->i_dirty_caps | ci->i_flushing_caps; 789 } 790 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 791 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 792 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 793 struct ceph_cap_flush **pcf); 794 795 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 796 struct ceph_cap *ocap, int mask); 797 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 798 extern int __ceph_caps_used(struct ceph_inode_info *ci); 799 800 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci) 801 { 802 return ci->i_nr_by_mode[0]; 803 } 804 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 805 extern int __ceph_caps_wanted(struct ceph_inode_info *ci); 806 807 /* what the mds thinks we want */ 808 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 809 810 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 811 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 812 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, 813 struct ceph_mount_options *fsopt); 814 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 815 struct ceph_cap_reservation *ctx, int need); 816 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, 817 struct ceph_cap_reservation *ctx); 818 extern void ceph_reservation_status(struct ceph_fs_client *client, 819 int *total, int *avail, int *used, 820 int *reserved, int *min); 821 extern void change_auth_cap_ses(struct ceph_inode_info *ci, 822 struct ceph_mds_session *session); 823 824 825 826 /* 827 * we keep buffered readdir results attached to file->private_data 828 */ 829 #define CEPH_F_SYNC 1 830 #define CEPH_F_ATEND 2 831 832 struct ceph_file_info { 833 short fmode; /* initialized on open */ 834 short flags; /* CEPH_F_* */ 835 836 spinlock_t rw_contexts_lock; 837 struct list_head rw_contexts; 838 839 u32 filp_gen; 840 }; 841 842 struct ceph_dir_file_info { 843 struct ceph_file_info file_info; 844 845 /* readdir: position within the dir */ 846 u32 frag; 847 struct ceph_mds_request *last_readdir; 848 849 /* readdir: position within a frag */ 850 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 851 char *last_name; /* last entry in previous chunk */ 852 long long dir_release_count; 853 long long dir_ordered_count; 854 int readdir_cache_idx; 855 856 /* used for -o dirstat read() on directory thing */ 857 char *dir_info; 858 int dir_info_len; 859 }; 860 861 struct ceph_rw_context { 862 struct list_head list; 863 struct task_struct *thread; 864 int caps; 865 }; 866 867 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ 868 struct ceph_rw_context _name = { \ 869 .thread = current, \ 870 .caps = _caps, \ 871 } 872 873 static inline void ceph_add_rw_context(struct ceph_file_info *cf, 874 struct ceph_rw_context *ctx) 875 { 876 spin_lock(&cf->rw_contexts_lock); 877 list_add(&ctx->list, &cf->rw_contexts); 878 spin_unlock(&cf->rw_contexts_lock); 879 } 880 881 static inline void ceph_del_rw_context(struct ceph_file_info *cf, 882 struct ceph_rw_context *ctx) 883 { 884 spin_lock(&cf->rw_contexts_lock); 885 list_del(&ctx->list); 886 spin_unlock(&cf->rw_contexts_lock); 887 } 888 889 static inline struct ceph_rw_context* 890 ceph_find_rw_context(struct ceph_file_info *cf) 891 { 892 struct ceph_rw_context *ctx, *found = NULL; 893 spin_lock(&cf->rw_contexts_lock); 894 list_for_each_entry(ctx, &cf->rw_contexts, list) { 895 if (ctx->thread == current) { 896 found = ctx; 897 break; 898 } 899 } 900 spin_unlock(&cf->rw_contexts_lock); 901 return found; 902 } 903 904 struct ceph_readdir_cache_control { 905 struct page *page; 906 struct dentry **dentries; 907 int index; 908 }; 909 910 /* 911 * A "snap realm" describes a subset of the file hierarchy sharing 912 * the same set of snapshots that apply to it. The realms themselves 913 * are organized into a hierarchy, such that children inherit (some of) 914 * the snapshots of their parents. 915 * 916 * All inodes within the realm that have capabilities are linked into a 917 * per-realm list. 918 */ 919 struct ceph_snap_realm { 920 u64 ino; 921 struct inode *inode; 922 atomic_t nref; 923 struct rb_node node; 924 925 u64 created, seq; 926 u64 parent_ino; 927 u64 parent_since; /* snapid when our current parent became so */ 928 929 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 930 u32 num_prior_parent_snaps; /* had prior to parent_since */ 931 u64 *snaps; /* snaps specific to this realm */ 932 u32 num_snaps; 933 934 struct ceph_snap_realm *parent; 935 struct list_head children; /* list of child realms */ 936 struct list_head child_item; 937 938 struct list_head empty_item; /* if i have ref==0 */ 939 940 struct list_head dirty_item; /* if realm needs new context */ 941 942 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */ 943 944 /* the current set of snaps for this realm */ 945 struct ceph_snap_context *cached_context; 946 947 struct list_head inodes_with_caps; 948 spinlock_t inodes_with_caps_lock; 949 }; 950 951 static inline int default_congestion_kb(void) 952 { 953 int congestion_kb; 954 955 /* 956 * Copied from NFS 957 * 958 * congestion size, scale with available memory. 959 * 960 * 64MB: 8192k 961 * 128MB: 11585k 962 * 256MB: 16384k 963 * 512MB: 23170k 964 * 1GB: 32768k 965 * 2GB: 46340k 966 * 4GB: 65536k 967 * 8GB: 92681k 968 * 16GB: 131072k 969 * 970 * This allows larger machines to have larger/more transfers. 971 * Limit the default to 256M 972 */ 973 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); 974 if (congestion_kb > 256*1024) 975 congestion_kb = 256*1024; 976 977 return congestion_kb; 978 } 979 980 981 /* super.c */ 982 extern int ceph_force_reconnect(struct super_block *sb); 983 /* snap.c */ 984 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 985 u64 ino); 986 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 987 struct ceph_snap_realm *realm); 988 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 989 struct ceph_snap_realm *realm); 990 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 991 void *p, void *e, bool deletion, 992 struct ceph_snap_realm **realm_ret); 993 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm); 994 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 995 struct ceph_mds_session *session, 996 struct ceph_msg *msg); 997 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 998 struct ceph_cap_snap *capsnap); 999 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc); 1000 1001 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, 1002 u64 snap); 1003 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, 1004 struct ceph_snapid_map *sm); 1005 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); 1006 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); 1007 void ceph_umount_begin(struct super_block *sb); 1008 1009 1010 /* 1011 * a cap_snap is "pending" if it is still awaiting an in-progress 1012 * sync write (that may/may not still update size, mtime, etc.). 1013 */ 1014 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 1015 { 1016 return !list_empty(&ci->i_cap_snaps) && 1017 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 1018 ci_item)->writing; 1019 } 1020 1021 /* inode.c */ 1022 struct ceph_mds_reply_info_in; 1023 struct ceph_mds_reply_dirfrag; 1024 struct ceph_acl_sec_ctx; 1025 1026 extern const struct inode_operations ceph_file_iops; 1027 1028 extern struct inode *ceph_alloc_inode(struct super_block *sb); 1029 extern void ceph_evict_inode(struct inode *inode); 1030 extern void ceph_free_inode(struct inode *inode); 1031 1032 struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry, 1033 umode_t *mode, struct ceph_acl_sec_ctx *as_ctx); 1034 void ceph_as_ctx_to_req(struct ceph_mds_request *req, 1035 struct ceph_acl_sec_ctx *as_ctx); 1036 1037 extern struct inode *ceph_get_inode(struct super_block *sb, 1038 struct ceph_vino vino, 1039 struct inode *newino); 1040 extern struct inode *ceph_get_snapdir(struct inode *parent); 1041 extern int ceph_fill_file_size(struct inode *inode, int issued, 1042 u32 truncate_seq, u64 truncate_size, u64 size); 1043 extern void ceph_fill_file_time(struct inode *inode, int issued, 1044 u64 time_warp_seq, struct timespec64 *ctime, 1045 struct timespec64 *mtime, 1046 struct timespec64 *atime); 1047 extern int ceph_fill_inode(struct inode *inode, struct page *locked_page, 1048 struct ceph_mds_reply_info_in *iinfo, 1049 struct ceph_mds_reply_dirfrag *dirinfo, 1050 struct ceph_mds_session *session, int cap_fmode, 1051 struct ceph_cap_reservation *caps_reservation); 1052 extern int ceph_fill_trace(struct super_block *sb, 1053 struct ceph_mds_request *req); 1054 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 1055 struct ceph_mds_session *session); 1056 1057 extern int ceph_inode_holds_cap(struct inode *inode, int mask); 1058 1059 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 1060 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 1061 1062 void ceph_queue_inode_work(struct inode *inode, int work_bit); 1063 1064 static inline void ceph_queue_vmtruncate(struct inode *inode) 1065 { 1066 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE); 1067 } 1068 1069 static inline void ceph_queue_invalidate(struct inode *inode) 1070 { 1071 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES); 1072 } 1073 1074 static inline void ceph_queue_writeback(struct inode *inode) 1075 { 1076 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK); 1077 } 1078 1079 static inline void ceph_queue_check_caps(struct inode *inode) 1080 { 1081 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS); 1082 } 1083 1084 static inline void ceph_queue_flush_snaps(struct inode *inode) 1085 { 1086 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS); 1087 } 1088 1089 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask); 1090 extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 1091 int mask, bool force); 1092 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 1093 { 1094 return __ceph_do_getattr(inode, NULL, mask, force); 1095 } 1096 extern int ceph_permission(struct mnt_idmap *idmap, 1097 struct inode *inode, int mask); 1098 1099 struct ceph_iattr { 1100 struct ceph_fscrypt_auth *fscrypt_auth; 1101 }; 1102 1103 extern int __ceph_setattr(struct mnt_idmap *idmap, struct inode *inode, 1104 struct iattr *attr, struct ceph_iattr *cia); 1105 extern int ceph_setattr(struct mnt_idmap *idmap, 1106 struct dentry *dentry, struct iattr *attr); 1107 extern int ceph_getattr(struct mnt_idmap *idmap, 1108 const struct path *path, struct kstat *stat, 1109 u32 request_mask, unsigned int flags); 1110 void ceph_inode_shutdown(struct inode *inode); 1111 1112 static inline bool ceph_inode_is_shutdown(struct inode *inode) 1113 { 1114 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags); 1115 struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode); 1116 int state = READ_ONCE(fsc->mount_state); 1117 1118 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN; 1119 } 1120 1121 /* xattr.c */ 1122 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 1123 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size); 1124 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 1125 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 1126 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); 1127 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 1128 extern const struct xattr_handler * const ceph_xattr_handlers[]; 1129 1130 struct ceph_acl_sec_ctx { 1131 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1132 void *default_acl; 1133 void *acl; 1134 #endif 1135 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1136 void *sec_ctx; 1137 u32 sec_ctxlen; 1138 #endif 1139 #ifdef CONFIG_FS_ENCRYPTION 1140 struct ceph_fscrypt_auth *fscrypt_auth; 1141 #endif 1142 struct ceph_pagelist *pagelist; 1143 }; 1144 1145 #ifdef CONFIG_SECURITY 1146 extern bool ceph_security_xattr_deadlock(struct inode *in); 1147 extern bool ceph_security_xattr_wanted(struct inode *in); 1148 #else 1149 static inline bool ceph_security_xattr_deadlock(struct inode *in) 1150 { 1151 return false; 1152 } 1153 static inline bool ceph_security_xattr_wanted(struct inode *in) 1154 { 1155 return false; 1156 } 1157 #endif 1158 1159 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1160 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1161 struct ceph_acl_sec_ctx *ctx); 1162 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1163 { 1164 security_inode_invalidate_secctx(inode); 1165 } 1166 #else 1167 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1168 struct ceph_acl_sec_ctx *ctx) 1169 { 1170 return 0; 1171 } 1172 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1173 { 1174 } 1175 #endif 1176 1177 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); 1178 1179 /* acl.c */ 1180 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1181 1182 struct posix_acl *ceph_get_acl(struct inode *, int, bool); 1183 int ceph_set_acl(struct mnt_idmap *idmap, 1184 struct dentry *dentry, struct posix_acl *acl, int type); 1185 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1186 struct ceph_acl_sec_ctx *as_ctx); 1187 void ceph_init_inode_acls(struct inode *inode, 1188 struct ceph_acl_sec_ctx *as_ctx); 1189 1190 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1191 { 1192 forget_all_cached_acls(inode); 1193 } 1194 1195 #else 1196 1197 #define ceph_get_acl NULL 1198 #define ceph_set_acl NULL 1199 1200 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1201 struct ceph_acl_sec_ctx *as_ctx) 1202 { 1203 return 0; 1204 } 1205 static inline void ceph_init_inode_acls(struct inode *inode, 1206 struct ceph_acl_sec_ctx *as_ctx) 1207 { 1208 } 1209 static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 1210 { 1211 return 0; 1212 } 1213 1214 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1215 { 1216 } 1217 1218 #endif 1219 1220 /* caps.c */ 1221 extern const char *ceph_cap_string(int c); 1222 extern void ceph_handle_caps(struct ceph_mds_session *session, 1223 struct ceph_msg *msg); 1224 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 1225 struct ceph_cap_reservation *ctx); 1226 extern void ceph_add_cap(struct inode *inode, 1227 struct ceph_mds_session *session, u64 cap_id, 1228 unsigned issued, unsigned wanted, 1229 unsigned cap, unsigned seq, u64 realmino, int flags, 1230 struct ceph_cap **new_cap); 1231 extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 1232 extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap, 1233 bool queue_release); 1234 extern void __ceph_remove_caps(struct ceph_inode_info *ci); 1235 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 1236 struct ceph_cap *cap); 1237 extern int ceph_is_any_caps(struct inode *inode); 1238 1239 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 1240 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 1241 int datasync); 1242 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 1243 struct ceph_mds_session *session); 1244 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 1245 struct ceph_mds_session *session); 1246 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, 1247 struct ceph_inode_info *ci); 1248 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, 1249 int mds); 1250 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 1251 int mds); 1252 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, 1253 bool snap_rwsem_locked); 1254 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 1255 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 1256 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had); 1257 extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci, 1258 int had); 1259 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 1260 struct ceph_snap_context *snapc); 1261 extern void __ceph_remove_capsnap(struct inode *inode, 1262 struct ceph_cap_snap *capsnap, 1263 bool *wake_ci, bool *wake_mdsc); 1264 extern void ceph_remove_capsnap(struct inode *inode, 1265 struct ceph_cap_snap *capsnap, 1266 bool *wake_ci, bool *wake_mdsc); 1267 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 1268 struct ceph_mds_session **psession); 1269 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 1270 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags); 1271 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 1272 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 1273 extern int ceph_drop_caps_for_unlink(struct inode *inode); 1274 extern int ceph_encode_inode_release(void **p, struct inode *inode, 1275 int mds, int drop, int unless, int force); 1276 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 1277 struct inode *dir, 1278 int mds, int drop, int unless); 1279 1280 extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi, 1281 int need, int want, loff_t endoff, int *got); 1282 extern int ceph_get_caps(struct file *filp, int need, int want, 1283 loff_t endoff, int *got); 1284 extern int ceph_try_get_caps(struct inode *inode, 1285 int need, int want, bool nonblock, int *got); 1286 1287 /* for counting open files by mode */ 1288 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count); 1289 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count); 1290 extern void __ceph_touch_fmode(struct ceph_inode_info *ci, 1291 struct ceph_mds_client *mdsc, int fmode); 1292 1293 /* addr.c */ 1294 extern const struct address_space_operations ceph_aops; 1295 extern const struct netfs_request_ops ceph_netfs_ops; 1296 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 1297 extern int ceph_uninline_data(struct file *file); 1298 extern int ceph_pool_perm_check(struct inode *inode, int need); 1299 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 1300 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate); 1301 1302 static inline bool ceph_has_inline_data(struct ceph_inode_info *ci) 1303 { 1304 if (ci->i_inline_version == CEPH_INLINE_NONE || 1305 ci->i_inline_version == 1) /* initial version, no data */ 1306 return false; 1307 return true; 1308 } 1309 1310 /* file.c */ 1311 extern const struct file_operations ceph_file_fops; 1312 1313 extern int ceph_renew_caps(struct inode *inode, int fmode); 1314 extern int ceph_open(struct inode *inode, struct file *file); 1315 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 1316 struct file *file, unsigned flags, umode_t mode); 1317 extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos, 1318 struct iov_iter *to, int *retry_op, 1319 u64 *last_objver); 1320 extern int ceph_release(struct inode *inode, struct file *filp); 1321 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 1322 char *data, size_t len); 1323 1324 /* dir.c */ 1325 extern const struct file_operations ceph_dir_fops; 1326 extern const struct file_operations ceph_snapdir_fops; 1327 extern const struct inode_operations ceph_dir_iops; 1328 extern const struct inode_operations ceph_snapdir_iops; 1329 extern const struct dentry_operations ceph_dentry_ops; 1330 1331 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 1332 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 1333 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req, 1334 struct dentry *dentry); 1335 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 1336 struct dentry *dentry, int err); 1337 1338 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); 1339 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); 1340 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 1341 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); 1342 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 1343 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 1344 1345 /* ioctl.c */ 1346 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1347 1348 /* export.c */ 1349 extern const struct export_operations ceph_export_ops; 1350 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); 1351 1352 /* locks.c */ 1353 extern __init void ceph_flock_init(void); 1354 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 1355 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 1356 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 1357 extern int ceph_encode_locks_to_buffer(struct inode *inode, 1358 struct ceph_filelock *flocks, 1359 int num_fcntl_locks, 1360 int num_flock_locks); 1361 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 1362 struct ceph_pagelist *pagelist, 1363 int num_fcntl_locks, int num_flock_locks); 1364 1365 /* debugfs.c */ 1366 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); 1367 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1368 1369 /* quota.c */ 1370 1371 enum quota_get_realm { 1372 QUOTA_GET_MAX_FILES, 1373 QUOTA_GET_MAX_BYTES, 1374 QUOTA_GET_ANY 1375 }; 1376 1377 static inline bool __ceph_has_quota(struct ceph_inode_info *ci, 1378 enum quota_get_realm which) 1379 { 1380 bool has_quota = false; 1381 1382 switch (which) { 1383 case QUOTA_GET_MAX_BYTES: 1384 has_quota = !!ci->i_max_bytes; 1385 break; 1386 case QUOTA_GET_MAX_FILES: 1387 has_quota = !!ci->i_max_files; 1388 break; 1389 default: 1390 has_quota = !!(ci->i_max_files || ci->i_max_bytes); 1391 } 1392 return has_quota; 1393 } 1394 1395 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); 1396 1397 static inline void __ceph_update_quota(struct ceph_inode_info *ci, 1398 u64 max_bytes, u64 max_files) 1399 { 1400 bool had_quota, has_quota; 1401 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1402 ci->i_max_bytes = max_bytes; 1403 ci->i_max_files = max_files; 1404 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1405 1406 if (had_quota != has_quota) 1407 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota); 1408 } 1409 1410 extern void ceph_handle_quota(struct ceph_mds_client *mdsc, 1411 struct ceph_mds_session *session, 1412 struct ceph_msg *msg); 1413 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); 1414 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); 1415 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, 1416 loff_t newlen); 1417 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, 1418 loff_t newlen); 1419 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, 1420 struct kstatfs *buf); 1421 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); 1422 1423 bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc, 1424 struct ceph_mds_session *session); 1425 void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc); 1426 bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc); 1427 void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc); 1428 #endif /* _FS_CEPH_SUPER_H */ 1429