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 #include <linux/ceph/osd_client.h> 7 8 #include <linux/unaligned.h> 9 #include <linux/backing-dev.h> 10 #include <linux/completion.h> 11 #include <linux/exportfs.h> 12 #include <linux/fs.h> 13 #include <linux/mempool.h> 14 #include <linux/pagemap.h> 15 #include <linux/wait.h> 16 #include <linux/writeback.h> 17 #include <linux/slab.h> 18 #include <linux/posix_acl.h> 19 #include <linux/refcount.h> 20 #include <linux/security.h> 21 #include <linux/netfs.h> 22 #include <linux/fscache.h> 23 #include <linux/hashtable.h> 24 25 #include <linux/ceph/libceph.h> 26 #include "crypto.h" 27 28 /* large granularity for statfs utilization stats to facilitate 29 * large volume sizes on 32-bit machines. */ 30 #define CEPH_BLOCK_SHIFT 22 /* 4 MB */ 31 #define CEPH_BLOCK (1 << CEPH_BLOCK_SHIFT) 32 #define CEPH_4K_BLOCK_SHIFT 12 /* 4 KB */ 33 34 #define CEPH_MOUNT_OPT_CLEANRECOVER (1<<1) /* auto reonnect (clean mode) after blocklisted */ 35 #define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ 36 #define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ 37 #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ 38 #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ 39 #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ 40 #define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ 41 #define CEPH_MOUNT_OPT_NOPOOLPERM (1<<11) /* no pool permission check */ 42 #define CEPH_MOUNT_OPT_MOUNTWAIT (1<<12) /* mount waits if no mds is up */ 43 #define CEPH_MOUNT_OPT_NOQUOTADF (1<<13) /* no root dir quota in statfs */ 44 #define CEPH_MOUNT_OPT_NOCOPYFROM (1<<14) /* don't use RADOS 'copy-from' op */ 45 #define CEPH_MOUNT_OPT_ASYNC_DIROPS (1<<15) /* allow async directory ops */ 46 #define CEPH_MOUNT_OPT_NOPAGECACHE (1<<16) /* bypass pagecache altogether */ 47 #define CEPH_MOUNT_OPT_SPARSEREAD (1<<17) /* always do sparse reads */ 48 #define CEPH_MOUNT_OPT_NEARFULL_SYNC (1<<18) /* sync writes when nearfull */ 49 #define CEPH_MOUNT_OPT_LAZYIO (1<<19) /* force lazyio for all file opens */ 50 51 #define CEPH_MOUNT_OPT_DEFAULT \ 52 (CEPH_MOUNT_OPT_DCACHE | \ 53 CEPH_MOUNT_OPT_NOCOPYFROM | \ 54 CEPH_MOUNT_OPT_ASYNC_DIROPS) 55 56 #define ceph_set_mount_opt(fsc, opt) \ 57 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt 58 #define ceph_clear_mount_opt(fsc, opt) \ 59 (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt 60 #define ceph_test_mount_opt(fsc, opt) \ 61 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 62 63 /* max size of osd read request, limited by libceph */ 64 #define CEPH_MAX_READ_SIZE CEPH_MSG_MAX_DATA_LEN 65 /* osd has a configurable limitation of max write size. 66 * CEPH_MSG_MAX_DATA_LEN should be small enough. */ 67 #define CEPH_MAX_WRITE_SIZE CEPH_MSG_MAX_DATA_LEN 68 #define CEPH_RASIZE_DEFAULT (8192*1024) /* max readahead */ 69 #define CEPH_MAX_READDIR_DEFAULT 1024 70 #define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 71 #define CEPH_SNAPDIRNAME_DEFAULT ".snap" 72 73 /* 74 * Delay telling the MDS we no longer want caps, in case we reopen 75 * the file. Delay a minimum amount of time, even if we send a cap 76 * message for some other reason. Otherwise, take the oppotunity to 77 * update the mds to avoid sending another message later. 78 */ 79 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT 5 /* cap release delay */ 80 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT 60 /* cap release delay */ 81 82 struct ceph_mount_options { 83 unsigned int flags; 84 85 unsigned int wsize; /* max write size */ 86 unsigned int rsize; /* max read size */ 87 unsigned int rasize; /* max readahead */ 88 unsigned int congestion_kb; /* max writeback in flight */ 89 unsigned int caps_wanted_delay_min, caps_wanted_delay_max; 90 int caps_max; 91 unsigned int max_readdir; /* max readdir result (entries) */ 92 unsigned int max_readdir_bytes; /* max readdir result (bytes) */ 93 94 bool new_dev_syntax; 95 96 /* 97 * everything above this point can be memcmp'd; everything below 98 * is handled in compare_mount_options() 99 */ 100 101 char *snapdir_name; /* default ".snap" */ 102 char *mds_namespace; /* default NULL */ 103 char *server_path; /* default NULL (means "/") */ 104 char *fscache_uniq; /* default NULL */ 105 char *mon_addr; 106 struct fscrypt_dummy_policy dummy_enc_policy; 107 }; 108 109 #define CEPH_NAMESPACE_WILDCARD "*" 110 111 static inline bool ceph_namespace_match(const char *pattern, 112 const char *target) 113 { 114 if (!pattern || !pattern[0] || 115 !strcmp(pattern, CEPH_NAMESPACE_WILDCARD)) 116 return true; 117 118 return !strcmp(pattern, target); 119 } 120 121 /* 122 * Check if the mds namespace in ceph_mount_options matches 123 * the passed in namespace string. First time match (when 124 * ->mds_namespace is NULL) is treated specially, since 125 * ->mds_namespace needs to be initialized by the caller. 126 */ 127 static inline bool namespace_equals(struct ceph_mount_options *fsopt, 128 const char *namespace, size_t len) 129 { 130 return !(fsopt->mds_namespace && 131 (strlen(fsopt->mds_namespace) != len || 132 strncmp(fsopt->mds_namespace, namespace, len))); 133 } 134 135 /* mount state */ 136 enum { 137 CEPH_MOUNT_MOUNTING, 138 CEPH_MOUNT_MOUNTED, 139 CEPH_MOUNT_UNMOUNTING, 140 CEPH_MOUNT_UNMOUNTED, 141 CEPH_MOUNT_SHUTDOWN, 142 CEPH_MOUNT_RECOVER, 143 CEPH_MOUNT_FENCE_IO, 144 }; 145 146 #define CEPH_ASYNC_CREATE_CONFLICT_BITS 8 147 148 struct ceph_fs_client { 149 struct super_block *sb; 150 151 struct list_head metric_wakeup; 152 153 struct ceph_mount_options *mount_options; 154 struct ceph_client *client; 155 156 int mount_state; 157 158 bool blocklisted; 159 160 bool have_copy_from2; 161 162 u32 filp_gen; 163 loff_t max_file_size; 164 165 struct ceph_mds_client *mdsc; 166 167 atomic_long_t writeback_count; 168 bool write_congested; 169 170 struct workqueue_struct *inode_wq; 171 struct workqueue_struct *cap_wq; 172 173 DECLARE_HASHTABLE(async_unlink_conflict, CEPH_ASYNC_CREATE_CONFLICT_BITS); 174 spinlock_t async_unlink_conflict_lock; 175 176 #ifdef CONFIG_DEBUG_FS 177 struct dentry *debugfs_dentry_lru, *debugfs_caps; 178 struct dentry *debugfs_congestion_kb; 179 struct dentry *debugfs_bdi; 180 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 181 struct dentry *debugfs_status; 182 struct dentry *debugfs_mds_sessions; 183 struct dentry *debugfs_metrics_dir; 184 struct dentry *debugfs_reset_dir; 185 struct dentry *debugfs_subvolume_metrics; 186 #endif 187 188 #ifdef CONFIG_CEPH_FSCACHE 189 struct fscache_volume *fscache; 190 #endif 191 #ifdef CONFIG_FS_ENCRYPTION 192 struct fscrypt_dummy_policy fsc_dummy_enc_policy; 193 #endif 194 }; 195 196 /* 197 * File i/o capability. This tracks shared state with the metadata 198 * server that allows us to cache or writeback attributes or to read 199 * and write data. For any given inode, we should have one or more 200 * capabilities, one issued by each metadata server, and our 201 * cumulative access is the OR of all issued capabilities. 202 * 203 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 204 * session capability lists. 205 */ 206 struct ceph_cap { 207 struct ceph_inode_info *ci; 208 209 /** 210 * Per-ci cap tree. Protected with 211 * `ceph_inode_info.i_ceph_lock`. 212 * 213 * Clearing this field with RB_CLEAR_NODE() requires holding 214 * both `ceph_inode_info.i_ceph_lock` and 215 * `ceph_mds_session->s_cap_lock`. Calling RB_EMPTY_NODE() 216 * (via ceph_cap_is_removed()) requires holding at least one 217 * of these. 218 */ 219 struct rb_node ci_node; 220 221 struct ceph_mds_session *session; 222 struct list_head session_caps; /* per-session caplist */ 223 u64 cap_id; /* unique cap id (mds provided) */ 224 union { 225 /* in-use caps */ 226 struct { 227 int issued; /* latest, from the mds */ 228 int implemented; /* implemented superset of 229 issued (for revocation) */ 230 int mds; /* mds index for this cap */ 231 int mds_wanted; /* caps wanted from this mds */ 232 }; 233 /* caps to release */ 234 struct { 235 u64 cap_ino; 236 int queue_release; 237 }; 238 }; 239 u32 seq, issue_seq, mseq; 240 u32 cap_gen; /* active/stale cycle */ 241 unsigned long last_used; 242 struct list_head caps_item; 243 }; 244 245 #define CHECK_CAPS_AUTHONLY 1 /* only check auth cap */ 246 #define CHECK_CAPS_FLUSH 2 /* flush any dirty caps */ 247 #define CHECK_CAPS_NOINVAL 4 /* don't invalidate pagecache */ 248 #define CHECK_CAPS_FLUSH_FORCE 8 /* force flush any caps */ 249 250 struct ceph_cap_flush { 251 u64 tid; 252 int caps; 253 bool wake; /* wake up flush waiters when finish ? */ 254 bool is_capsnap; /* true means capsnap */ 255 struct list_head g_list; // global 256 struct list_head i_list; // per inode 257 struct ceph_inode_info *ci; 258 }; 259 260 /* 261 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 262 * we first complete any in-process sync writes and writeback any dirty 263 * data before flushing the snapped state (tracked here) back to the MDS. 264 */ 265 struct ceph_cap_snap { 266 refcount_t nref; 267 struct list_head ci_item; 268 269 struct ceph_cap_flush cap_flush; 270 271 u64 follows; 272 int issued, dirty; 273 struct ceph_snap_context *context; 274 275 umode_t mode; 276 kuid_t uid; 277 kgid_t gid; 278 279 struct ceph_buffer *xattr_blob; 280 u64 xattr_version; 281 282 u64 size; 283 u64 change_attr; 284 struct timespec64 mtime, atime, ctime, btime; 285 u32 time_warp_seq; 286 u64 truncate_size; 287 u32 truncate_seq; 288 bool writing; /* a sync write is still in progress */ 289 int dirty_pages; /* dirty pages awaiting writeback */ 290 bool inline_data; 291 bool need_flush; 292 }; 293 294 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 295 { 296 if (refcount_dec_and_test(&capsnap->nref)) { 297 if (capsnap->xattr_blob) 298 ceph_buffer_put(capsnap->xattr_blob); 299 kmem_cache_free(ceph_cap_snap_cachep, capsnap); 300 } 301 } 302 303 /* 304 * The frag tree describes how a directory is fragmented, potentially across 305 * multiple metadata servers. It is also used to indicate points where 306 * metadata authority is delegated, and whether/where metadata is replicated. 307 * 308 * A _leaf_ frag will be present in the i_fragtree IFF there is 309 * delegation info. That is, if mds >= 0 || ndist > 0. 310 */ 311 #define CEPH_MAX_DIRFRAG_REP 4 312 313 struct ceph_inode_frag { 314 struct rb_node node; 315 316 /* fragtree state */ 317 u32 frag; 318 int split_by; /* i.e. 2^(split_by) children */ 319 320 /* delegation and replication info */ 321 int mds; /* -1 if same authority as parent */ 322 int ndist; /* >0 if replicated */ 323 int dist[CEPH_MAX_DIRFRAG_REP]; 324 }; 325 326 /* 327 * We cache inode xattrs as an encoded blob until they are first used, 328 * at which point we parse them into an rbtree. 329 */ 330 struct ceph_inode_xattr { 331 struct rb_node node; 332 333 const char *name; 334 int name_len; 335 const char *val; 336 int val_len; 337 bool dirty; 338 339 bool should_free_name; 340 bool should_free_val; 341 }; 342 343 /* 344 * Ceph dentry state 345 */ 346 struct ceph_dentry_info { 347 struct dentry *dentry; 348 struct ceph_mds_session *lease_session; 349 struct list_head lease_list; 350 struct hlist_node hnode; 351 unsigned long flags; 352 int lease_shared_gen; 353 u32 lease_gen; 354 u32 lease_seq; 355 unsigned long lease_renew_after, lease_renew_from; 356 unsigned long time; 357 u64 offset; 358 }; 359 360 #define CEPH_DENTRY_REFERENCED (1 << 0) 361 #define CEPH_DENTRY_LEASE_LIST (1 << 1) 362 #define CEPH_DENTRY_SHRINK_LIST (1 << 2) 363 #define CEPH_DENTRY_PRIMARY_LINK (1 << 3) 364 #define CEPH_DENTRY_ASYNC_UNLINK_BIT (4) 365 #define CEPH_DENTRY_ASYNC_UNLINK (1 << CEPH_DENTRY_ASYNC_UNLINK_BIT) 366 #define CEPH_DENTRY_ASYNC_CREATE_BIT (5) 367 #define CEPH_DENTRY_ASYNC_CREATE (1 << CEPH_DENTRY_ASYNC_CREATE_BIT) 368 369 struct ceph_inode_xattrs_info { 370 /* 371 * (still encoded) xattr blob. we avoid the overhead of parsing 372 * this until someone actually calls getxattr, etc. 373 * 374 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 375 * NULL means we don't know. 376 */ 377 struct ceph_buffer *blob, *prealloc_blob; 378 379 struct rb_root index; 380 bool dirty; 381 int count; 382 int names_size; 383 int vals_size; 384 u64 version, index_version; 385 }; 386 387 /* 388 * Ceph inode. 389 */ 390 struct ceph_inode_info { 391 struct netfs_inode netfs; /* Netfslib context and vfs inode */ 392 393 #if BITS_PER_LONG >= 64 394 /* on 64 bit systems, the full Ceph inode number is stored in 395 * netfs.inode.i_ino 396 */ 397 u64 i_snap; 398 #else 399 struct ceph_vino i_vino; /* ceph ino + snap */ 400 #endif 401 402 spinlock_t i_ceph_lock; 403 404 u64 i_version; 405 #ifdef CONFIG_CEPH_FS_INLINE_DATA 406 u64 i_inline_version; 407 #endif 408 u32 i_time_warp_seq; 409 410 unsigned long i_ceph_flags; 411 atomic64_t i_release_count; 412 atomic64_t i_ordered_count; 413 atomic64_t i_complete_seq[2]; 414 415 struct ceph_dir_layout i_dir_layout; 416 struct ceph_file_layout i_layout; 417 struct ceph_file_layout i_cached_layout; // for async creates 418 char *i_symlink; 419 420 /* for dirs */ 421 struct timespec64 i_rctime; 422 u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps; 423 u64 i_files, i_subdirs; 424 425 /* quotas */ 426 u64 i_max_bytes, i_max_files; 427 428 /* 429 * Subvolume ID this inode belongs to. CEPH_SUBVOLUME_ID_NONE (0) 430 * means unknown/unset, matching the FUSE client convention. 431 * Once set to a valid (non-zero) value, it should not change 432 * during the inode's lifetime. 433 */ 434 #define CEPH_SUBVOLUME_ID_NONE 0 435 u64 i_subvolume_id; 436 437 s32 i_dir_pin; 438 439 struct rb_root i_fragtree; 440 int i_fragtree_nsplits; 441 struct mutex i_fragtree_mutex; 442 443 struct ceph_inode_xattrs_info i_xattrs; 444 445 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 446 * s_mutex. */ 447 struct rb_root i_caps; /* cap list */ 448 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 449 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 450 451 /* 452 * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty 453 * is protected by the mdsc->cap_dirty_lock, but each individual item 454 * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty 455 * requires the mdsc->cap_dirty_lock. List presence for an item can 456 * be tested under the i_ceph_lock. Changing anything requires both. 457 */ 458 struct list_head i_dirty_item; 459 460 /* 461 * Link to session's s_cap_flushing list. Protected in a similar 462 * fashion to i_dirty_item, but also by the s_mutex for changes. The 463 * s_cap_flushing list can be walked while holding either the s_mutex 464 * or msdc->cap_dirty_lock. List presence can also be checked while 465 * holding the i_ceph_lock for this inode. 466 */ 467 struct list_head i_flushing_item; 468 469 /* we need to track cap writeback on a per-cap-bit basis, to allow 470 * overlapping, pipelined cap flushes to the mds. we can probably 471 * reduce the tid to 8 bits if we're concerned about inode size. */ 472 struct ceph_cap_flush *i_prealloc_cap_flush; 473 struct list_head i_cap_flush_list; 474 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 475 unsigned long i_hold_caps_max; /* jiffies */ 476 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 477 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 478 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 479 dirty|flushing caps */ 480 unsigned i_snap_caps; /* cap bits for snapped files */ 481 /* 482 * Written under i_ceph_lock, read via READ_ONCE() 483 * from diagnostic paths. 484 */ 485 u64 i_last_cap_flush_ack; 486 487 unsigned long i_last_rd; 488 unsigned long i_last_wr; 489 int i_nr_by_mode[CEPH_FILE_MODE_BITS]; /* open file counts */ 490 491 struct mutex i_truncate_mutex; 492 u32 i_truncate_seq; /* last truncate to smaller size */ 493 u64 i_truncate_size; /* and the size we last truncated down to */ 494 bool i_truncate_pending; /* still need to call vmtruncate */ 495 bool i_truncate_shrank; /* Truncate shrank the file */ 496 /* 497 * For none fscrypt case it equals to i_truncate_size or it will 498 * equals to fscrypt_file_size 499 */ 500 u64 i_truncate_pagecache_size; 501 502 u64 i_max_size; /* max file size authorized by mds */ 503 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 504 u64 i_wanted_max_size; /* offset we'd like to write too */ 505 u64 i_requested_max_size; /* max_size we've requested */ 506 507 /* held references to caps */ 508 int i_pin_ref; 509 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref; 510 int i_wrbuffer_ref, i_wrbuffer_ref_head; 511 atomic_t i_filelock_ref; 512 atomic_t i_shared_gen; /* increment each time we get FILE_SHARED */ 513 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 514 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 515 516 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 517 struct list_head i_unsafe_iops; /* uncommitted mds inode ops */ 518 spinlock_t i_unsafe_lock; 519 520 union { 521 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 522 struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */ 523 }; 524 struct list_head i_snap_realm_item; 525 struct list_head i_snap_flush_item; 526 struct timespec64 i_btime; 527 struct timespec64 i_snap_btime; 528 529 struct work_struct i_work; 530 unsigned long i_work_mask; 531 532 #ifdef CONFIG_FS_ENCRYPTION 533 struct fscrypt_inode_info *i_crypt_info; 534 u32 fscrypt_auth_len; 535 u32 fscrypt_file_len; 536 u8 *fscrypt_auth; 537 u8 *fscrypt_file; 538 #endif 539 }; 540 541 struct ceph_netfs_request_data { 542 int caps; 543 544 /* 545 * Maximum size of a file readahead request. 546 * The fadvise could update the bdi's default ra_pages. 547 */ 548 unsigned int file_ra_pages; 549 550 /* Set it if fadvise disables file readahead entirely */ 551 bool file_ra_disabled; 552 }; 553 554 static inline struct ceph_inode_info * 555 ceph_inode(const struct inode *inode) 556 { 557 return container_of(inode, struct ceph_inode_info, netfs.inode); 558 } 559 560 static inline struct ceph_fs_client * 561 ceph_inode_to_fs_client(const struct inode *inode) 562 { 563 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 564 } 565 566 static inline struct ceph_fs_client * 567 ceph_sb_to_fs_client(const struct super_block *sb) 568 { 569 return (struct ceph_fs_client *)sb->s_fs_info; 570 } 571 572 static inline struct ceph_mds_client * 573 ceph_sb_to_mdsc(const struct super_block *sb) 574 { 575 return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc; 576 } 577 578 static inline struct ceph_client * 579 ceph_inode_to_client(const struct inode *inode) 580 { 581 return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client; 582 } 583 584 static inline struct ceph_vino 585 ceph_vino(const struct inode *inode) 586 { 587 #if BITS_PER_LONG >= 64 588 return (struct ceph_vino){ 589 .ino = inode->i_ino, 590 .snap = ceph_inode(inode)->i_snap, 591 }; 592 #else 593 return ceph_inode(inode)->i_vino; 594 #endif 595 } 596 597 static inline u32 ceph_ino_to_ino32(u64 vino) 598 { 599 u32 ino = vino & 0xffffffff; 600 ino ^= vino >> 32; 601 if (!ino) 602 ino = 2; 603 return ino; 604 } 605 606 /* 607 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on 608 * some arches. We generally do not use this value inside the ceph driver, but 609 * we do want to set it to something, so that generic vfs code has an 610 * appropriate value for tracepoints and the like. 611 */ 612 static inline ino_t ceph_vino_to_ino_t(const struct ceph_vino vino) 613 { 614 if (sizeof(ino_t) == sizeof(u32)) 615 return ceph_ino_to_ino32(vino.ino); 616 return (ino_t)vino.ino; 617 } 618 619 /* for printf-style formatting */ 620 #define ceph_vinop(i) ceph_ino(i), ceph_snap(i) 621 622 static inline u64 ceph_ino(const struct inode *inode) 623 { 624 #if BITS_PER_LONG >= 64 625 return inode->i_ino; 626 #else 627 return ceph_inode(inode)->i_vino.ino; 628 #endif 629 } 630 631 static inline u64 ceph_snap(const struct inode *inode) 632 { 633 #if BITS_PER_LONG >= 64 634 return ceph_inode(inode)->i_snap; 635 #else 636 return ceph_inode(inode)->i_vino.snap; 637 #endif 638 } 639 640 /** 641 * Is this inode in a ".snap" directory? 642 */ 643 static inline bool ceph_in_snap(const struct inode *inode) 644 { 645 return ceph_snap(inode) != CEPH_NOSNAP; 646 } 647 648 /** 649 * ceph_present_ino - format an inode number for presentation to userland 650 * @sb: superblock where the inode lives 651 * @ino: inode number to (possibly) convert 652 * 653 * If the user mounted with the ino32 option, then the 64-bit value needs 654 * to be converted to something that can fit inside 32 bits. Note that 655 * internal kernel code never uses this value, so this is entirely for 656 * userland consumption. 657 */ 658 static inline u64 ceph_present_ino(struct super_block *sb, u64 ino) 659 { 660 if (unlikely(ceph_test_mount_opt(ceph_sb_to_fs_client(sb), INO32))) 661 return ceph_ino_to_ino32(ino); 662 return ino; 663 } 664 665 static inline u64 ceph_present_inode(struct inode *inode) 666 { 667 return ceph_present_ino(inode->i_sb, ceph_ino(inode)); 668 } 669 670 static inline int ceph_ino_compare(struct inode *inode, void *data) 671 { 672 struct ceph_vino *pvino = (struct ceph_vino *)data; 673 return ceph_ino(inode) == pvino->ino && 674 ceph_snap(inode) == pvino->snap; 675 } 676 677 /* 678 * The MDS reserves a set of inodes for its own usage. These should never 679 * be accessible by clients, and so the MDS has no reason to ever hand these 680 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE. 681 * 682 * These come from src/mds/mdstypes.h in the ceph sources. 683 */ 684 #define CEPH_MAX_MDS 0x100 685 #define CEPH_NUM_STRAY 10 686 #define CEPH_MDS_INO_MDSDIR_OFFSET (1 * CEPH_MAX_MDS) 687 #define CEPH_MDS_INO_LOG_OFFSET (2 * CEPH_MAX_MDS) 688 #define CEPH_INO_SYSTEM_BASE ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY)) 689 690 /* 691 * Upper bound on the number of delegated inodes a single MDS session may 692 * hold. The MDS normally hands out a small preallocation window (the 693 * userspace mds_client_prealloc_inos option defaults to 1000) and refills 694 * it as the client consumes entries. This leaves generous headroom while 695 * bounding the CPU and memory a malformed delegation interval can consume. 696 */ 697 #define CEPH_MAX_DELEG_INOS 8192 698 699 static inline bool ceph_vino_is_reserved(const struct ceph_vino vino) 700 { 701 if (vino.ino >= CEPH_INO_SYSTEM_BASE || 702 vino.ino < CEPH_MDS_INO_MDSDIR_OFFSET) 703 return false; 704 705 /* Don't warn on mdsdirs */ 706 WARN_RATELIMIT(vino.ino >= CEPH_MDS_INO_LOG_OFFSET, 707 "Attempt to access reserved inode number 0x%llx", 708 vino.ino); 709 return true; 710 } 711 712 static inline struct inode *ceph_find_inode(struct super_block *sb, 713 struct ceph_vino vino) 714 { 715 if (ceph_vino_is_reserved(vino)) 716 return NULL; 717 718 /* 719 * NB: The hashval will be run through the fs/inode.c hash function 720 * anyway, so there is no need to squash the inode number down to 721 * 32-bits first. Just use low-order bits on arches with 32-bit long. 722 */ 723 return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino); 724 } 725 726 727 /* 728 * Ceph inode. 729 */ 730 #define CEPH_I_DIR_ORDERED_BIT (0) /* dentries in dir are ordered */ 731 /* bit 1 historically unused */ 732 #define CEPH_I_FLUSH_BIT (2) /* do not delay flush of dirty metadata */ 733 #define CEPH_I_POOL_PERM_BIT (3) /* pool rd/wr bits are valid */ 734 #define CEPH_I_POOL_RD_BIT (4) /* can read from pool */ 735 #define CEPH_I_POOL_WR_BIT (5) /* can write to pool */ 736 #define CEPH_I_SEC_INITED_BIT (6) /* security initialized */ 737 #define CEPH_I_KICK_FLUSH_BIT (7) /* kick flushing caps */ 738 #define CEPH_I_FLUSH_SNAPS_BIT (8) /* need flush snaps */ 739 #define CEPH_I_ERROR_WRITE_BIT (9) /* have seen write errors */ 740 #define CEPH_I_ERROR_FILELOCK_BIT (10) /* have seen file lock errors */ 741 #define CEPH_I_ODIRECT_BIT (11) /* inode in direct I/O mode */ 742 #define CEPH_I_ASYNC_CREATE_BIT (12) /* async create in flight for this */ 743 #define CEPH_I_SHUTDOWN_BIT (13) /* inode is no longer usable */ 744 #define CEPH_I_ASYNC_CHECK_CAPS_BIT (14) /* check caps after async creating finishes */ 745 #define CEPH_I_FLUSH_FORCE_BIT (15) /* a revoke's response was deferred; 746 * force a cap message to the MDS once 747 * the deferred work completes 748 */ 749 #define CEPH_I_EVICT_ON_FINAL_IPUT_BIT (16) /* evict at the final iput() */ 750 751 #define CEPH_I_DIR_ORDERED (1 << CEPH_I_DIR_ORDERED_BIT) 752 #define CEPH_I_FLUSH (1 << CEPH_I_FLUSH_BIT) 753 #define CEPH_I_POOL_PERM (1 << CEPH_I_POOL_PERM_BIT) 754 #define CEPH_I_POOL_RD (1 << CEPH_I_POOL_RD_BIT) 755 #define CEPH_I_POOL_WR (1 << CEPH_I_POOL_WR_BIT) 756 #define CEPH_I_SEC_INITED (1 << CEPH_I_SEC_INITED_BIT) 757 #define CEPH_I_KICK_FLUSH (1 << CEPH_I_KICK_FLUSH_BIT) 758 #define CEPH_I_FLUSH_SNAPS (1 << CEPH_I_FLUSH_SNAPS_BIT) 759 #define CEPH_I_ODIRECT (1 << CEPH_I_ODIRECT_BIT) 760 #define CEPH_I_ASYNC_CREATE (1 << CEPH_I_ASYNC_CREATE_BIT) 761 #define CEPH_I_SHUTDOWN (1 << CEPH_I_SHUTDOWN_BIT) 762 #define CEPH_I_FLUSH_FORCE (1 << CEPH_I_FLUSH_FORCE_BIT) 763 764 /* 765 * Masks of ceph inode work. 766 */ 767 #define CEPH_I_WORK_WRITEBACK 0 768 #define CEPH_I_WORK_INVALIDATE_PAGES 1 769 #define CEPH_I_WORK_VMTRUNCATE 2 770 #define CEPH_I_WORK_CHECK_CAPS 3 771 #define CEPH_I_WORK_FLUSH_SNAPS 4 772 773 /* 774 * We set the ERROR_WRITE bit when we start seeing write errors on an inode 775 * and then clear it when they start succeeding. The write submission code 776 * just takes this as a hint, so we're not too worried if a few slip through 777 * in either direction. 778 */ 779 static inline void ceph_set_error_write(struct ceph_inode_info *ci) 780 { 781 set_bit(CEPH_I_ERROR_WRITE_BIT, &ci->i_ceph_flags); 782 } 783 784 static inline void ceph_clear_error_write(struct ceph_inode_info *ci) 785 { 786 clear_bit(CEPH_I_ERROR_WRITE_BIT, &ci->i_ceph_flags); 787 } 788 789 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 790 long long release_count, 791 long long ordered_count) 792 { 793 /* 794 * Makes sure operations that setup readdir cache (update page 795 * cache and i_size) are strongly ordered w.r.t. the following 796 * atomic64_set() operations. 797 */ 798 smp_mb(); 799 atomic64_set(&ci->i_complete_seq[0], release_count); 800 atomic64_set(&ci->i_complete_seq[1], ordered_count); 801 } 802 803 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 804 { 805 atomic64_inc(&ci->i_release_count); 806 } 807 808 static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci) 809 { 810 atomic64_inc(&ci->i_ordered_count); 811 } 812 813 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 814 { 815 return atomic64_read(&ci->i_complete_seq[0]) == 816 atomic64_read(&ci->i_release_count); 817 } 818 819 static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 820 { 821 return atomic64_read(&ci->i_complete_seq[0]) == 822 atomic64_read(&ci->i_release_count) && 823 atomic64_read(&ci->i_complete_seq[1]) == 824 atomic64_read(&ci->i_ordered_count); 825 } 826 827 static inline void ceph_dir_clear_complete(struct inode *inode) 828 { 829 __ceph_dir_clear_complete(ceph_inode(inode)); 830 } 831 832 static inline void ceph_dir_clear_ordered(struct inode *inode) 833 { 834 __ceph_dir_clear_ordered(ceph_inode(inode)); 835 } 836 837 static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 838 { 839 bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode)); 840 smp_rmb(); 841 return ret; 842 } 843 844 /* find a specific frag @f */ 845 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 846 u32 f); 847 848 /* 849 * choose fragment for value @v. copy frag content to pfrag, if leaf 850 * exists 851 */ 852 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 853 struct ceph_inode_frag *pfrag, 854 int *found); 855 856 static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry) 857 { 858 return (struct ceph_dentry_info *)dentry->d_fsdata; 859 } 860 861 /* 862 * caps helpers 863 */ 864 static inline bool __ceph_is_any_real_caps(const struct ceph_inode_info *ci) 865 { 866 return !RB_EMPTY_ROOT(&ci->i_caps); 867 } 868 869 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 870 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 871 extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask, 872 int t); 873 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 874 struct ceph_cap *cap); 875 876 static inline int ceph_caps_issued(struct ceph_inode_info *ci) 877 { 878 int issued; 879 spin_lock(&ci->i_ceph_lock); 880 issued = __ceph_caps_issued(ci, NULL); 881 spin_unlock(&ci->i_ceph_lock); 882 return issued; 883 } 884 885 static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, 886 int mask, int touch) 887 { 888 int r; 889 spin_lock(&ci->i_ceph_lock); 890 r = __ceph_caps_issued_mask_metric(ci, mask, touch); 891 spin_unlock(&ci->i_ceph_lock); 892 return r; 893 } 894 895 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 896 { 897 return ci->i_dirty_caps | ci->i_flushing_caps; 898 } 899 extern struct ceph_cap_flush *ceph_alloc_cap_flush(void); 900 extern void ceph_free_cap_flush(struct ceph_cap_flush *cf); 901 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask, 902 struct ceph_cap_flush **pcf); 903 904 extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 905 struct ceph_cap *ocap, int mask); 906 extern int __ceph_caps_used(struct ceph_inode_info *ci); 907 908 static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci) 909 { 910 return ci->i_nr_by_mode[0]; 911 } 912 913 /* 914 * True if the inode is open and every opener is a lazy one. Bit 0 of 915 * i_nr_by_mode[] is set for every open regardless of mode (see 916 * ceph_get_fmode()), so it doubles as the total open count. 917 * 918 * Callers must hold i_ceph_lock. 919 */ 920 static inline bool __ceph_all_opens_lazy(struct ceph_inode_info *ci) 921 { 922 int nr_open = ci->i_nr_by_mode[ffs(CEPH_FILE_MODE_PIN)]; 923 int nr_lazy = ci->i_nr_by_mode[ffs(CEPH_FILE_MODE_LAZY)]; 924 925 return nr_open > 0 && nr_lazy >= nr_open; 926 } 927 928 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 929 extern int __ceph_caps_wanted(struct ceph_inode_info *ci); 930 931 /* what the mds thinks we want */ 932 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check); 933 934 extern void ceph_caps_init(struct ceph_mds_client *mdsc); 935 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 936 extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc, 937 struct ceph_mount_options *fsopt); 938 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc, 939 struct ceph_cap_reservation *ctx, int need); 940 extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc, 941 struct ceph_cap_reservation *ctx); 942 extern void ceph_reservation_status(struct ceph_fs_client *client, 943 int *total, int *avail, int *used, 944 int *reserved, int *min); 945 extern void change_auth_cap_ses(struct ceph_inode_info *ci, 946 struct ceph_mds_session *session); 947 948 949 950 /* 951 * we keep buffered readdir results attached to file->private_data 952 */ 953 #define CEPH_F_SYNC 1 954 #define CEPH_F_ATEND 2 955 956 struct ceph_file_info { 957 short fmode; /* initialized on open */ 958 short flags; /* CEPH_F_* */ 959 960 spinlock_t rw_contexts_lock; 961 struct list_head rw_contexts; 962 963 u32 filp_gen; 964 }; 965 966 struct ceph_dir_file_info { 967 struct ceph_file_info file_info; 968 969 /* readdir: position within the dir */ 970 u32 frag; 971 struct ceph_mds_request *last_readdir; 972 973 /* readdir: position within a frag */ 974 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 975 char *last_name; /* last entry in previous chunk */ 976 long long dir_release_count; 977 long long dir_ordered_count; 978 int readdir_cache_idx; 979 980 /* used for -o dirstat read() on directory thing */ 981 char *dir_info; 982 int dir_info_len; 983 }; 984 985 struct ceph_rw_context { 986 struct list_head list; 987 struct task_struct *thread; 988 int caps; 989 }; 990 991 #define CEPH_DEFINE_RW_CONTEXT(_name, _caps) \ 992 struct ceph_rw_context _name = { \ 993 .thread = current, \ 994 .caps = _caps, \ 995 } 996 997 static inline void ceph_add_rw_context(struct ceph_file_info *cf, 998 struct ceph_rw_context *ctx) 999 { 1000 spin_lock(&cf->rw_contexts_lock); 1001 list_add(&ctx->list, &cf->rw_contexts); 1002 spin_unlock(&cf->rw_contexts_lock); 1003 } 1004 1005 static inline void ceph_del_rw_context(struct ceph_file_info *cf, 1006 struct ceph_rw_context *ctx) 1007 { 1008 spin_lock(&cf->rw_contexts_lock); 1009 list_del(&ctx->list); 1010 spin_unlock(&cf->rw_contexts_lock); 1011 } 1012 1013 static inline struct ceph_rw_context* 1014 ceph_find_rw_context(struct ceph_file_info *cf) 1015 { 1016 struct ceph_rw_context *ctx, *found = NULL; 1017 spin_lock(&cf->rw_contexts_lock); 1018 list_for_each_entry(ctx, &cf->rw_contexts, list) { 1019 if (ctx->thread == current) { 1020 found = ctx; 1021 break; 1022 } 1023 } 1024 spin_unlock(&cf->rw_contexts_lock); 1025 return found; 1026 } 1027 1028 struct ceph_readdir_cache_control { 1029 struct folio *folio; 1030 struct dentry **dentries; 1031 int index; 1032 }; 1033 1034 /* 1035 * A "snap realm" describes a subset of the file hierarchy sharing 1036 * the same set of snapshots that apply to it. The realms themselves 1037 * are organized into a hierarchy, such that children inherit (some of) 1038 * the snapshots of their parents. 1039 * 1040 * All inodes within the realm that have capabilities are linked into a 1041 * per-realm list. 1042 */ 1043 struct ceph_snap_realm { 1044 u64 ino; 1045 struct inode *inode; 1046 atomic_t nref; 1047 struct rb_node node; 1048 1049 u64 created, seq; 1050 u64 parent_ino; 1051 u64 parent_since; /* snapid when our current parent became so */ 1052 1053 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 1054 u32 num_prior_parent_snaps; /* had prior to parent_since */ 1055 u64 *snaps; /* snaps specific to this realm */ 1056 u32 num_snaps; 1057 1058 struct ceph_snap_realm *parent; 1059 struct list_head children; /* list of child realms */ 1060 struct list_head child_item; 1061 1062 struct list_head empty_item; /* if i have ref==0 */ 1063 1064 struct list_head dirty_item; /* if realm needs new context */ 1065 1066 struct list_head rebuild_item; /* rebuild snap realms _downward_ in hierarchy */ 1067 1068 /* the current set of snaps for this realm */ 1069 struct ceph_snap_context *cached_context; 1070 1071 struct list_head inodes_with_caps; 1072 spinlock_t inodes_with_caps_lock; 1073 }; 1074 1075 static inline int default_congestion_kb(void) 1076 { 1077 int congestion_kb; 1078 1079 /* 1080 * Copied from NFS 1081 * 1082 * congestion size, scale with available memory. 1083 * 1084 * 64MB: 8192k 1085 * 128MB: 11585k 1086 * 256MB: 16384k 1087 * 512MB: 23170k 1088 * 1GB: 32768k 1089 * 2GB: 46340k 1090 * 4GB: 65536k 1091 * 8GB: 92681k 1092 * 16GB: 131072k 1093 * 1094 * This allows larger machines to have larger/more transfers. 1095 * Limit the default to 256M 1096 */ 1097 congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10); 1098 if (congestion_kb > 256*1024) 1099 congestion_kb = 256*1024; 1100 1101 return congestion_kb; 1102 } 1103 1104 1105 /* super.c */ 1106 extern int ceph_force_reconnect(struct super_block *sb); 1107 /* snap.c */ 1108 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 1109 u64 ino); 1110 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 1111 struct ceph_snap_realm *realm); 1112 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 1113 struct ceph_snap_realm *realm); 1114 extern int ceph_update_snap_trace(struct ceph_mds_client *m, 1115 void *p, void *e, bool deletion, 1116 struct ceph_snap_realm **realm_ret); 1117 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm); 1118 extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 1119 struct ceph_mds_session *session, 1120 struct ceph_msg *msg); 1121 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 1122 struct ceph_cap_snap *capsnap); 1123 extern void ceph_cleanup_global_and_empty_realms(struct ceph_mds_client *mdsc); 1124 1125 extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc, 1126 u64 snap); 1127 extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc, 1128 struct ceph_snapid_map *sm); 1129 extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc); 1130 extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc); 1131 void ceph_umount_begin(struct super_block *sb); 1132 1133 1134 /* 1135 * a cap_snap is "pending" if it is still awaiting an in-progress 1136 * sync write (that may/may not still update size, mtime, etc.). 1137 */ 1138 static inline bool __ceph_have_pending_cap_snap(const struct ceph_inode_info *ci) 1139 { 1140 return !list_empty(&ci->i_cap_snaps) && 1141 list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap, 1142 ci_item)->writing; 1143 } 1144 1145 /* inode.c */ 1146 struct ceph_mds_reply_info_in; 1147 struct ceph_mds_reply_dirfrag; 1148 struct ceph_acl_sec_ctx; 1149 1150 extern const struct inode_operations ceph_file_iops; 1151 1152 extern struct inode *ceph_alloc_inode(struct super_block *sb); 1153 extern int ceph_drop_inode(struct inode *inode); 1154 extern void ceph_evict_inode(struct inode *inode); 1155 extern void ceph_free_inode(struct inode *inode); 1156 1157 struct inode *ceph_new_inode(struct inode *dir, struct dentry *dentry, 1158 umode_t *mode, struct ceph_acl_sec_ctx *as_ctx); 1159 void ceph_as_ctx_to_req(struct ceph_mds_request *req, 1160 struct ceph_acl_sec_ctx *as_ctx); 1161 1162 extern struct inode *ceph_get_inode(struct super_block *sb, 1163 struct ceph_vino vino, 1164 struct inode *newino); 1165 extern struct inode *ceph_get_snapdir(struct inode *parent); 1166 extern int ceph_fill_file_size(struct inode *inode, int issued, 1167 u32 truncate_seq, u64 truncate_size, u64 size); 1168 extern void ceph_inode_set_subvolume(struct inode *inode, u64 subvolume_id); 1169 extern void ceph_fill_file_time(struct inode *inode, int issued, 1170 u32 time_warp_seq, struct timespec64 *ctime, 1171 struct timespec64 *mtime, 1172 struct timespec64 *atime); 1173 int ceph_fill_inode(struct inode *inode, struct folio *locked_folio, 1174 struct ceph_mds_reply_info_in *iinfo, 1175 struct ceph_mds_reply_dirfrag *dirinfo, 1176 struct ceph_mds_session *session, int cap_fmode, 1177 struct ceph_cap_reservation *caps_reservation); 1178 extern int ceph_fill_trace(struct super_block *sb, 1179 struct ceph_mds_request *req); 1180 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 1181 struct ceph_mds_session *session); 1182 1183 extern bool ceph_inode_set_size(struct inode *inode, loff_t size); 1184 extern void __ceph_do_pending_vmtruncate(struct inode *inode); 1185 1186 void ceph_queue_inode_work(struct inode *inode, int work_bit); 1187 1188 static inline void ceph_queue_vmtruncate(struct inode *inode) 1189 { 1190 ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE); 1191 } 1192 1193 static inline void ceph_queue_invalidate(struct inode *inode) 1194 { 1195 ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES); 1196 } 1197 1198 static inline void ceph_queue_writeback(struct inode *inode) 1199 { 1200 ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK); 1201 } 1202 1203 static inline void ceph_queue_check_caps(struct inode *inode) 1204 { 1205 ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS); 1206 } 1207 1208 static inline void ceph_queue_flush_snaps(struct inode *inode) 1209 { 1210 ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS); 1211 } 1212 1213 extern int ceph_try_to_choose_auth_mds(struct inode *inode, int mask); 1214 int __ceph_do_getattr(struct inode *inode, struct folio *locked_folio, 1215 int mask, bool force); 1216 static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 1217 { 1218 return __ceph_do_getattr(inode, NULL, mask, force); 1219 } 1220 extern int ceph_permission(const struct mnt_idmap *idmap, 1221 struct inode *inode, int mask); 1222 1223 struct ceph_iattr { 1224 struct ceph_fscrypt_auth *fscrypt_auth; 1225 }; 1226 1227 extern int __ceph_setattr(const struct mnt_idmap *idmap, struct inode *inode, 1228 struct iattr *attr, struct ceph_iattr *cia); 1229 extern int ceph_setattr(const struct mnt_idmap *idmap, 1230 struct dentry *dentry, struct iattr *attr); 1231 extern int ceph_getattr(const struct mnt_idmap *idmap, 1232 const struct path *path, struct kstat *stat, 1233 u32 request_mask, unsigned int flags); 1234 void ceph_inode_shutdown(struct inode *inode); 1235 1236 static inline bool ceph_inode_is_shutdown(const struct inode *inode) 1237 { 1238 unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags); 1239 struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode); 1240 int state = READ_ONCE(fsc->mount_state); 1241 1242 return (flags & CEPH_I_SHUTDOWN) || state >= CEPH_MOUNT_SHUTDOWN; 1243 } 1244 1245 /* xattr.c */ 1246 int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int); 1247 int ceph_do_getvxattr(struct inode *inode, const char *name, void *value, size_t size); 1248 ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 1249 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 1250 extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci); 1251 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 1252 extern const struct xattr_handler * const ceph_xattr_handlers[]; 1253 1254 struct ceph_acl_sec_ctx { 1255 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1256 void *default_acl; 1257 void *acl; 1258 #endif 1259 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1260 struct lsm_context lsmctx; 1261 #endif 1262 #ifdef CONFIG_FS_ENCRYPTION 1263 struct ceph_fscrypt_auth *fscrypt_auth; 1264 #endif 1265 struct ceph_pagelist *pagelist; 1266 }; 1267 1268 #ifdef CONFIG_SECURITY 1269 extern bool ceph_security_xattr_deadlock(struct inode *in); 1270 extern bool ceph_security_xattr_wanted(struct inode *in); 1271 #else 1272 static inline bool ceph_security_xattr_deadlock(struct inode *in) 1273 { 1274 return false; 1275 } 1276 static inline bool ceph_security_xattr_wanted(struct inode *in) 1277 { 1278 return false; 1279 } 1280 #endif 1281 1282 #ifdef CONFIG_CEPH_FS_SECURITY_LABEL 1283 extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1284 struct ceph_acl_sec_ctx *ctx); 1285 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1286 { 1287 security_inode_invalidate_secctx(inode); 1288 } 1289 #else 1290 static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode, 1291 struct ceph_acl_sec_ctx *ctx) 1292 { 1293 return 0; 1294 } 1295 static inline void ceph_security_invalidate_secctx(struct inode *inode) 1296 { 1297 } 1298 #endif 1299 1300 void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx); 1301 1302 /* acl.c */ 1303 #ifdef CONFIG_CEPH_FS_POSIX_ACL 1304 1305 struct posix_acl *ceph_get_acl(struct inode *, int, bool); 1306 int ceph_set_acl(const struct mnt_idmap *idmap, 1307 struct dentry *dentry, struct posix_acl *acl, int type); 1308 int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1309 struct ceph_acl_sec_ctx *as_ctx); 1310 void ceph_init_inode_acls(struct inode *inode, 1311 struct ceph_acl_sec_ctx *as_ctx); 1312 1313 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1314 { 1315 forget_all_cached_acls(inode); 1316 } 1317 1318 #else 1319 1320 #define ceph_get_acl NULL 1321 #define ceph_set_acl NULL 1322 1323 static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 1324 struct ceph_acl_sec_ctx *as_ctx) 1325 { 1326 return 0; 1327 } 1328 static inline void ceph_init_inode_acls(struct inode *inode, 1329 struct ceph_acl_sec_ctx *as_ctx) 1330 { 1331 } 1332 1333 static inline void ceph_forget_all_cached_acls(struct inode *inode) 1334 { 1335 } 1336 1337 #endif 1338 1339 /* caps.c */ 1340 extern const char *ceph_cap_string(int c); 1341 extern void ceph_handle_caps(struct ceph_mds_session *session, 1342 struct ceph_msg *msg); 1343 extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 1344 struct ceph_cap_reservation *ctx); 1345 extern void ceph_add_cap(struct inode *inode, 1346 struct ceph_mds_session *session, u64 cap_id, 1347 unsigned issued, unsigned wanted, 1348 unsigned cap, unsigned seq, u64 realmino, int flags, 1349 struct ceph_cap **new_cap); 1350 1351 /** 1352 * Determine whether __ceph_remove_cap() has been called on this #cap 1353 * (but the object has not yet been freed because it is protected by 1354 * `ceph_mds_session.s_cap_iterator`). 1355 * 1356 * Caller must lock either `ceph_inode_info.i_ceph_lock` or 1357 * `ceph_mds_session.s_cap_lock`. 1358 */ 1359 static inline bool ceph_cap_is_removed(const struct ceph_cap *cap) 1360 { 1361 return RB_EMPTY_NODE(&cap->ci_node); 1362 } 1363 1364 extern void ceph_remove_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap, 1365 struct ceph_inode_info *ci, 1366 bool queue_release); 1367 extern void __ceph_remove_caps(struct ceph_inode_info *ci); 1368 extern void ceph_put_cap(struct ceph_mds_client *mdsc, 1369 struct ceph_cap *cap); 1370 extern int ceph_is_any_caps(struct inode *inode); 1371 1372 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 1373 extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 1374 int datasync); 1375 extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc, 1376 struct ceph_mds_session *session); 1377 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 1378 struct ceph_mds_session *session); 1379 void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session, 1380 struct ceph_inode_info *ci); 1381 extern struct ceph_cap *__get_cap_for_mds(struct ceph_inode_info *ci, 1382 int mds); 1383 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 1384 int mds); 1385 extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps, 1386 bool snap_rwsem_locked); 1387 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 1388 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 1389 extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had); 1390 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 1391 struct ceph_snap_context *snapc); 1392 extern void __ceph_remove_capsnap(struct inode *inode, 1393 struct ceph_cap_snap *capsnap, 1394 bool *wake_ci, bool *wake_mdsc); 1395 extern void ceph_remove_capsnap(struct inode *inode, 1396 struct ceph_cap_snap *capsnap, 1397 bool *wake_ci, bool *wake_mdsc); 1398 extern void ceph_flush_snaps(struct ceph_inode_info *ci, 1399 struct ceph_mds_session **psession); 1400 extern bool __ceph_should_report_size(struct ceph_inode_info *ci); 1401 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags); 1402 extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 1403 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 1404 extern void ceph_flush_cap_releases(struct ceph_mds_client *mdsc); 1405 extern int ceph_drop_caps_for_unlink(struct inode *inode); 1406 extern int ceph_encode_inode_release(void **p, struct inode *inode, 1407 int mds, int drop, int unless, int force); 1408 extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 1409 struct inode *dir, 1410 int mds, int drop, int unless); 1411 1412 extern int __ceph_get_caps(struct inode *inode, struct ceph_file_info *fi, 1413 int need, int want, loff_t endoff, int *got); 1414 extern int ceph_get_caps(struct file *filp, int need, int want, 1415 loff_t endoff, int *got); 1416 extern int ceph_try_get_caps(struct inode *inode, 1417 int need, int want, bool nonblock, int *got); 1418 1419 /* for counting open files by mode */ 1420 extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count); 1421 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count); 1422 extern void __ceph_touch_fmode(struct ceph_inode_info *ci, 1423 struct ceph_mds_client *mdsc, int fmode); 1424 1425 /* addr.c */ 1426 extern const struct address_space_operations ceph_aops; 1427 extern const struct netfs_request_ops ceph_netfs_ops; 1428 int ceph_mmap_prepare(struct vm_area_desc *desc); 1429 extern int ceph_pool_perm_check(struct inode *inode, int need); 1430 extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc); 1431 int ceph_purge_inode_cap(struct inode *inode, struct ceph_cap *cap, bool *invalidate); 1432 1433 #ifdef CONFIG_CEPH_FS_INLINE_DATA 1434 extern int ceph_uninline_data(struct file *file); 1435 void ceph_fill_inline_data(struct inode *inode, struct folio *locked_folio, 1436 char *data, size_t len); 1437 1438 static inline u64 ceph_inline_version(const struct ceph_inode_info *ci) 1439 { 1440 return ci->i_inline_version; 1441 } 1442 1443 static inline void ceph_set_inline_version(struct ceph_inode_info *ci, 1444 u64 version) 1445 { 1446 ci->i_inline_version = version; 1447 } 1448 #else 1449 /* 1450 * Only ever called under ceph_has_inline_data(), which is a constant 1451 * false here, so this is unreachable; 0 keeps the error paths above 1452 * simple without implying anything was done. 1453 */ 1454 static inline int ceph_uninline_data(struct file *file) 1455 { 1456 return 0; 1457 } 1458 1459 static inline void ceph_fill_inline_data(struct inode *inode, 1460 struct folio *locked_folio, 1461 char *data, size_t len) 1462 { 1463 } 1464 1465 /* 1466 * Without inline data support an inode never has any, so report the 1467 * version the MDS uses for "no inline data" and ignore any update. 1468 * This constant-folds all the inline data handling away. 1469 */ 1470 static inline u64 ceph_inline_version(const struct ceph_inode_info *ci) 1471 { 1472 return CEPH_INLINE_NONE; 1473 } 1474 1475 static inline void ceph_set_inline_version(struct ceph_inode_info *ci, 1476 u64 version) 1477 { 1478 } 1479 #endif /* CONFIG_CEPH_FS_INLINE_DATA */ 1480 1481 static inline bool ceph_has_inline_data(const struct ceph_inode_info *ci) 1482 { 1483 u64 version = ceph_inline_version(ci); 1484 1485 if (version == CEPH_INLINE_NONE || 1486 version == 1) /* initial version, no data */ 1487 return false; 1488 return true; 1489 } 1490 1491 /* file.c */ 1492 extern const struct file_operations ceph_file_fops; 1493 1494 extern int ceph_renew_caps(struct inode *inode, int fmode); 1495 extern int ceph_open(struct inode *inode, struct file *file); 1496 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 1497 struct file *file, unsigned flags, umode_t mode); 1498 extern ssize_t __ceph_sync_read(struct inode *inode, loff_t *ki_pos, 1499 struct iov_iter *to, int *retry_op, 1500 u64 *last_objver); 1501 extern int ceph_release(struct inode *inode, struct file *filp); 1502 1503 /* dir.c */ 1504 extern const struct file_operations ceph_dir_fops; 1505 extern const struct file_operations ceph_snapdir_fops; 1506 extern const struct inode_operations ceph_dir_iops; 1507 extern const struct inode_operations ceph_snapdir_iops; 1508 extern const struct dentry_operations ceph_dentry_ops; 1509 1510 extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order); 1511 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 1512 extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req, 1513 struct dentry *dentry); 1514 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 1515 struct dentry *dentry, int err); 1516 1517 extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di); 1518 extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di); 1519 extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 1520 extern int ceph_trim_dentries(struct ceph_mds_client *mdsc); 1521 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 1522 extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl); 1523 1524 /* ioctl.c */ 1525 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 1526 1527 /* export.c */ 1528 extern const struct export_operations ceph_export_ops; 1529 struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino); 1530 1531 /* locks.c */ 1532 extern __init void ceph_flock_init(void); 1533 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 1534 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 1535 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 1536 extern int ceph_encode_locks_to_buffer(struct inode *inode, 1537 struct ceph_filelock *flocks, 1538 int num_fcntl_locks, 1539 int num_flock_locks); 1540 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 1541 struct ceph_pagelist *pagelist, 1542 int num_fcntl_locks, int num_flock_locks); 1543 1544 /* debugfs.c */ 1545 extern void ceph_fs_debugfs_init(struct ceph_fs_client *client); 1546 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 1547 1548 /* quota.c */ 1549 1550 enum quota_get_realm { 1551 QUOTA_GET_MAX_FILES, 1552 QUOTA_GET_MAX_BYTES, 1553 QUOTA_GET_ANY 1554 }; 1555 1556 static inline bool __ceph_has_quota(const struct ceph_inode_info *ci, 1557 enum quota_get_realm which) 1558 { 1559 bool has_quota = false; 1560 1561 switch (which) { 1562 case QUOTA_GET_MAX_BYTES: 1563 has_quota = !!ci->i_max_bytes; 1564 break; 1565 case QUOTA_GET_MAX_FILES: 1566 has_quota = !!ci->i_max_files; 1567 break; 1568 default: 1569 has_quota = !!(ci->i_max_files || ci->i_max_bytes); 1570 } 1571 return has_quota; 1572 } 1573 1574 extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc); 1575 1576 static inline void __ceph_update_quota(struct ceph_inode_info *ci, 1577 u64 max_bytes, u64 max_files) 1578 { 1579 bool had_quota, has_quota; 1580 had_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1581 ci->i_max_bytes = max_bytes; 1582 ci->i_max_files = max_files; 1583 has_quota = __ceph_has_quota(ci, QUOTA_GET_ANY); 1584 1585 if (had_quota != has_quota) 1586 ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota); 1587 } 1588 1589 static inline int __ceph_sparse_read_ext_count(const struct inode *inode, u64 len) 1590 { 1591 int cnt = 0; 1592 1593 if (IS_ENCRYPTED(inode)) { 1594 cnt = len >> CEPH_FSCRYPT_BLOCK_SHIFT; 1595 if (cnt > CEPH_SPARSE_EXT_ARRAY_INITIAL) 1596 cnt = 0; 1597 } 1598 1599 return cnt; 1600 } 1601 1602 extern void ceph_handle_quota(struct ceph_mds_client *mdsc, 1603 struct ceph_mds_session *session, 1604 struct ceph_msg *msg); 1605 extern bool ceph_quota_is_max_files_exceeded(struct inode *inode); 1606 extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new); 1607 extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode, 1608 loff_t newlen); 1609 extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode, 1610 loff_t newlen); 1611 extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc, 1612 struct kstatfs *buf); 1613 extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc); 1614 1615 bool ceph_inc_mds_stopping_blocker(struct ceph_mds_client *mdsc, 1616 struct ceph_mds_session *session); 1617 void ceph_dec_mds_stopping_blocker(struct ceph_mds_client *mdsc); 1618 bool ceph_inc_osd_stopping_blocker(struct ceph_mds_client *mdsc); 1619 void ceph_dec_osd_stopping_blocker(struct ceph_mds_client *mdsc); 1620 #endif /* _FS_CEPH_SUPER_H */ 1621