1#ifndef _FS_CEPH_SUPER_H 2#define _FS_CEPH_SUPER_H 3 4#include <linux/ceph/ceph_debug.h> 5 6#include <asm/unaligned.h> 7#include <linux/backing-dev.h> 8#include <linux/completion.h> 9#include <linux/exportfs.h> 10#include <linux/fs.h> 11#include <linux/mempool.h> 12#include <linux/pagemap.h> 13#include <linux/wait.h> 14#include <linux/writeback.h> 15#include <linux/slab.h> 16#include <linux/posix_acl.h> 17 18#include <linux/ceph/libceph.h> 19 20#ifdef CONFIG_CEPH_FSCACHE 21#include <linux/fscache.h> 22#endif 23 24/* f_type in struct statfs */ 25#define CEPH_SUPER_MAGIC 0x00c36400 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 32#define CEPH_MOUNT_OPT_DIRSTAT (1<<4) /* `cat dirname` for stats */ 33#define CEPH_MOUNT_OPT_RBYTES (1<<5) /* dir st_bytes = rbytes */ 34#define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ 35#define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ 36#define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ 37#define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ 38 39#define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES | \ 40 CEPH_MOUNT_OPT_DCACHE) 41 42#define ceph_set_mount_opt(fsc, opt) \ 43 (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt; 44#define ceph_test_mount_opt(fsc, opt) \ 45 (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt)) 46 47#define CEPH_RSIZE_DEFAULT 0 /* max read size */ 48#define CEPH_RASIZE_DEFAULT (8192*1024) /* readahead */ 49#define CEPH_MAX_READDIR_DEFAULT 1024 50#define CEPH_MAX_READDIR_BYTES_DEFAULT (512*1024) 51#define CEPH_SNAPDIRNAME_DEFAULT ".snap" 52 53struct ceph_mount_options { 54 int flags; 55 int sb_flags; 56 57 int wsize; /* max write size */ 58 int rsize; /* max read size */ 59 int rasize; /* max readahead */ 60 int congestion_kb; /* max writeback in flight */ 61 int caps_wanted_delay_min, caps_wanted_delay_max; 62 int cap_release_safety; 63 int max_readdir; /* max readdir result (entires) */ 64 int max_readdir_bytes; /* max readdir result (bytes) */ 65 66 /* 67 * everything above this point can be memcmp'd; everything below 68 * is handled in compare_mount_options() 69 */ 70 71 char *snapdir_name; /* default ".snap" */ 72}; 73 74struct ceph_fs_client { 75 struct super_block *sb; 76 77 struct ceph_mount_options *mount_options; 78 struct ceph_client *client; 79 80 unsigned long mount_state; 81 int min_caps; /* min caps i added */ 82 83 struct ceph_mds_client *mdsc; 84 85 /* writeback */ 86 mempool_t *wb_pagevec_pool; 87 struct workqueue_struct *wb_wq; 88 struct workqueue_struct *pg_inv_wq; 89 struct workqueue_struct *trunc_wq; 90 atomic_long_t writeback_count; 91 92 struct backing_dev_info backing_dev_info; 93 94#ifdef CONFIG_DEBUG_FS 95 struct dentry *debugfs_dentry_lru, *debugfs_caps; 96 struct dentry *debugfs_congestion_kb; 97 struct dentry *debugfs_bdi; 98 struct dentry *debugfs_mdsc, *debugfs_mdsmap; 99 struct dentry *debugfs_mds_sessions; 100#endif 101 102#ifdef CONFIG_CEPH_FSCACHE 103 struct fscache_cookie *fscache; 104 struct workqueue_struct *revalidate_wq; 105#endif 106}; 107 108 109/* 110 * File i/o capability. This tracks shared state with the metadata 111 * server that allows us to cache or writeback attributes or to read 112 * and write data. For any given inode, we should have one or more 113 * capabilities, one issued by each metadata server, and our 114 * cumulative access is the OR of all issued capabilities. 115 * 116 * Each cap is referenced by the inode's i_caps rbtree and by per-mds 117 * session capability lists. 118 */ 119struct ceph_cap { 120 struct ceph_inode_info *ci; 121 struct rb_node ci_node; /* per-ci cap tree */ 122 struct ceph_mds_session *session; 123 struct list_head session_caps; /* per-session caplist */ 124 int mds; 125 u64 cap_id; /* unique cap id (mds provided) */ 126 int issued; /* latest, from the mds */ 127 int implemented; /* implemented superset of issued (for revocation) */ 128 int mds_wanted; 129 u32 seq, issue_seq, mseq; 130 u32 cap_gen; /* active/stale cycle */ 131 unsigned long last_used; 132 struct list_head caps_item; 133}; 134 135#define CHECK_CAPS_NODELAY 1 /* do not delay any further */ 136#define CHECK_CAPS_AUTHONLY 2 /* only check auth cap */ 137#define CHECK_CAPS_FLUSH 4 /* flush any dirty caps */ 138 139/* 140 * Snapped cap state that is pending flush to mds. When a snapshot occurs, 141 * we first complete any in-process sync writes and writeback any dirty 142 * data before flushing the snapped state (tracked here) back to the MDS. 143 */ 144struct ceph_cap_snap { 145 atomic_t nref; 146 struct ceph_inode_info *ci; 147 struct list_head ci_item, flushing_item; 148 149 u64 follows, flush_tid; 150 int issued, dirty; 151 struct ceph_snap_context *context; 152 153 umode_t mode; 154 kuid_t uid; 155 kgid_t gid; 156 157 struct ceph_buffer *xattr_blob; 158 u64 xattr_version; 159 160 u64 size; 161 struct timespec mtime, atime, ctime; 162 u64 time_warp_seq; 163 int writing; /* a sync write is still in progress */ 164 int dirty_pages; /* dirty pages awaiting writeback */ 165 bool inline_data; 166}; 167 168static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap) 169{ 170 if (atomic_dec_and_test(&capsnap->nref)) { 171 if (capsnap->xattr_blob) 172 ceph_buffer_put(capsnap->xattr_blob); 173 kfree(capsnap); 174 } 175} 176 177/* 178 * The frag tree describes how a directory is fragmented, potentially across 179 * multiple metadata servers. It is also used to indicate points where 180 * metadata authority is delegated, and whether/where metadata is replicated. 181 * 182 * A _leaf_ frag will be present in the i_fragtree IFF there is 183 * delegation info. That is, if mds >= 0 || ndist > 0. 184 */ 185#define CEPH_MAX_DIRFRAG_REP 4 186 187struct ceph_inode_frag { 188 struct rb_node node; 189 190 /* fragtree state */ 191 u32 frag; 192 int split_by; /* i.e. 2^(split_by) children */ 193 194 /* delegation and replication info */ 195 int mds; /* -1 if same authority as parent */ 196 int ndist; /* >0 if replicated */ 197 int dist[CEPH_MAX_DIRFRAG_REP]; 198}; 199 200/* 201 * We cache inode xattrs as an encoded blob until they are first used, 202 * at which point we parse them into an rbtree. 203 */ 204struct ceph_inode_xattr { 205 struct rb_node node; 206 207 const char *name; 208 int name_len; 209 const char *val; 210 int val_len; 211 int dirty; 212 213 int should_free_name; 214 int should_free_val; 215}; 216 217/* 218 * Ceph dentry state 219 */ 220struct ceph_dentry_info { 221 struct ceph_mds_session *lease_session; 222 u32 lease_gen, lease_shared_gen; 223 u32 lease_seq; 224 unsigned long lease_renew_after, lease_renew_from; 225 struct list_head lru; 226 struct dentry *dentry; 227 u64 time; 228 u64 offset; 229}; 230 231struct ceph_inode_xattrs_info { 232 /* 233 * (still encoded) xattr blob. we avoid the overhead of parsing 234 * this until someone actually calls getxattr, etc. 235 * 236 * blob->vec.iov_len == 4 implies there are no xattrs; blob == 237 * NULL means we don't know. 238 */ 239 struct ceph_buffer *blob, *prealloc_blob; 240 241 struct rb_root index; 242 bool dirty; 243 int count; 244 int names_size; 245 int vals_size; 246 u64 version, index_version; 247}; 248 249/* 250 * Ceph inode. 251 */ 252struct ceph_inode_info { 253 struct ceph_vino i_vino; /* ceph ino + snap */ 254 255 spinlock_t i_ceph_lock; 256 257 u64 i_version; 258 u64 i_inline_version; 259 u32 i_time_warp_seq; 260 261 unsigned i_ceph_flags; 262 int i_ordered_count; 263 atomic_t i_release_count; 264 atomic_t i_complete_count; 265 266 struct ceph_dir_layout i_dir_layout; 267 struct ceph_file_layout i_layout; 268 char *i_symlink; 269 270 /* for dirs */ 271 struct timespec i_rctime; 272 u64 i_rbytes, i_rfiles, i_rsubdirs; 273 u64 i_files, i_subdirs; 274 275 struct rb_root i_fragtree; 276 struct mutex i_fragtree_mutex; 277 278 struct ceph_inode_xattrs_info i_xattrs; 279 280 /* capabilities. protected _both_ by i_ceph_lock and cap->session's 281 * s_mutex. */ 282 struct rb_root i_caps; /* cap list */ 283 struct ceph_cap *i_auth_cap; /* authoritative cap, if any */ 284 unsigned i_dirty_caps, i_flushing_caps; /* mask of dirtied fields */ 285 struct list_head i_dirty_item, i_flushing_item; 286 u64 i_cap_flush_seq; 287 /* we need to track cap writeback on a per-cap-bit basis, to allow 288 * overlapping, pipelined cap flushes to the mds. we can probably 289 * reduce the tid to 8 bits if we're concerned about inode size. */ 290 u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS]; 291 wait_queue_head_t i_cap_wq; /* threads waiting on a capability */ 292 unsigned long i_hold_caps_min; /* jiffies */ 293 unsigned long i_hold_caps_max; /* jiffies */ 294 struct list_head i_cap_delay_list; /* for delayed cap release to mds */ 295 struct ceph_cap_reservation i_cap_migration_resv; 296 struct list_head i_cap_snaps; /* snapped state pending flush to mds */ 297 struct ceph_snap_context *i_head_snapc; /* set if wr_buffer_head > 0 or 298 dirty|flushing caps */ 299 unsigned i_snap_caps; /* cap bits for snapped files */ 300 301 int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ 302 303 struct mutex i_truncate_mutex; 304 u32 i_truncate_seq; /* last truncate to smaller size */ 305 u64 i_truncate_size; /* and the size we last truncated down to */ 306 int i_truncate_pending; /* still need to call vmtruncate */ 307 308 u64 i_max_size; /* max file size authorized by mds */ 309 u64 i_reported_size; /* (max_)size reported to or requested of mds */ 310 u64 i_wanted_max_size; /* offset we'd like to write too */ 311 u64 i_requested_max_size; /* max_size we've requested */ 312 313 /* held references to caps */ 314 int i_pin_ref; 315 int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref; 316 int i_wrbuffer_ref, i_wrbuffer_ref_head; 317 u32 i_shared_gen; /* increment each time we get FILE_SHARED */ 318 u32 i_rdcache_gen; /* incremented each time we get FILE_CACHE. */ 319 u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */ 320 321 struct list_head i_unsafe_writes; /* uncommitted sync writes */ 322 struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */ 323 spinlock_t i_unsafe_lock; 324 325 struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */ 326 int i_snap_realm_counter; /* snap realm (if caps) */ 327 struct list_head i_snap_realm_item; 328 struct list_head i_snap_flush_item; 329 330 struct work_struct i_wb_work; /* writeback work */ 331 struct work_struct i_pg_inv_work; /* page invalidation work */ 332 333 struct work_struct i_vmtruncate_work; 334 335#ifdef CONFIG_CEPH_FSCACHE 336 struct fscache_cookie *fscache; 337 u32 i_fscache_gen; /* sequence, for delayed fscache validate */ 338 struct work_struct i_revalidate_work; 339#endif 340 struct inode vfs_inode; /* at end */ 341}; 342 343static inline struct ceph_inode_info *ceph_inode(struct inode *inode) 344{ 345 return container_of(inode, struct ceph_inode_info, vfs_inode); 346} 347 348static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode) 349{ 350 return (struct ceph_fs_client *)inode->i_sb->s_fs_info; 351} 352 353static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb) 354{ 355 return (struct ceph_fs_client *)sb->s_fs_info; 356} 357 358static inline struct ceph_vino ceph_vino(struct inode *inode) 359{ 360 return ceph_inode(inode)->i_vino; 361} 362 363/* 364 * ino_t is <64 bits on many architectures, blech. 365 * 366 * i_ino (kernel inode) st_ino (userspace) 367 * i386 32 32 368 * x86_64+ino32 64 32 369 * x86_64 64 64 370 */ 371static inline u32 ceph_ino_to_ino32(__u64 vino) 372{ 373 u32 ino = vino & 0xffffffff; 374 ino ^= vino >> 32; 375 if (!ino) 376 ino = 2; 377 return ino; 378} 379 380/* 381 * kernel i_ino value 382 */ 383static inline ino_t ceph_vino_to_ino(struct ceph_vino vino) 384{ 385#if BITS_PER_LONG == 32 386 return ceph_ino_to_ino32(vino.ino); 387#else 388 return (ino_t)vino.ino; 389#endif 390} 391 392/* 393 * user-visible ino (stat, filldir) 394 */ 395#if BITS_PER_LONG == 32 396static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 397{ 398 return ino; 399} 400#else 401static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino) 402{ 403 if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)) 404 ino = ceph_ino_to_ino32(ino); 405 return ino; 406} 407#endif 408 409 410/* for printf-style formatting */ 411#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap 412 413static inline u64 ceph_ino(struct inode *inode) 414{ 415 return ceph_inode(inode)->i_vino.ino; 416} 417static inline u64 ceph_snap(struct inode *inode) 418{ 419 return ceph_inode(inode)->i_vino.snap; 420} 421 422static inline int ceph_ino_compare(struct inode *inode, void *data) 423{ 424 struct ceph_vino *pvino = (struct ceph_vino *)data; 425 struct ceph_inode_info *ci = ceph_inode(inode); 426 return ci->i_vino.ino == pvino->ino && 427 ci->i_vino.snap == pvino->snap; 428} 429 430static inline struct inode *ceph_find_inode(struct super_block *sb, 431 struct ceph_vino vino) 432{ 433 ino_t t = ceph_vino_to_ino(vino); 434 return ilookup5(sb, t, ceph_ino_compare, &vino); 435} 436 437 438/* 439 * Ceph inode. 440 */ 441#define CEPH_I_DIR_ORDERED 1 /* dentries in dir are ordered */ 442#define CEPH_I_NODELAY 4 /* do not delay cap release */ 443#define CEPH_I_FLUSH 8 /* do not delay flush of dirty metadata */ 444#define CEPH_I_NOFLUSH 16 /* do not flush dirty caps */ 445 446static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci, 447 int release_count, int ordered_count) 448{ 449 atomic_set(&ci->i_complete_count, release_count); 450 if (ci->i_ordered_count == ordered_count) 451 ci->i_ceph_flags |= CEPH_I_DIR_ORDERED; 452 else 453 ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED; 454} 455 456static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci) 457{ 458 atomic_inc(&ci->i_release_count); 459} 460 461static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci) 462{ 463 return atomic_read(&ci->i_complete_count) == 464 atomic_read(&ci->i_release_count); 465} 466 467static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci) 468{ 469 return __ceph_dir_is_complete(ci) && 470 (ci->i_ceph_flags & CEPH_I_DIR_ORDERED); 471} 472 473static inline void ceph_dir_clear_complete(struct inode *inode) 474{ 475 __ceph_dir_clear_complete(ceph_inode(inode)); 476} 477 478static inline void ceph_dir_clear_ordered(struct inode *inode) 479{ 480 struct ceph_inode_info *ci = ceph_inode(inode); 481 spin_lock(&ci->i_ceph_lock); 482 ci->i_ordered_count++; 483 ci->i_ceph_flags &= ~CEPH_I_DIR_ORDERED; 484 spin_unlock(&ci->i_ceph_lock); 485} 486 487static inline bool ceph_dir_is_complete_ordered(struct inode *inode) 488{ 489 struct ceph_inode_info *ci = ceph_inode(inode); 490 bool ret; 491 spin_lock(&ci->i_ceph_lock); 492 ret = __ceph_dir_is_complete_ordered(ci); 493 spin_unlock(&ci->i_ceph_lock); 494 return ret; 495} 496 497/* find a specific frag @f */ 498extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci, 499 u32 f); 500 501/* 502 * choose fragment for value @v. copy frag content to pfrag, if leaf 503 * exists 504 */ 505extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v, 506 struct ceph_inode_frag *pfrag, 507 int *found); 508 509static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry) 510{ 511 return (struct ceph_dentry_info *)dentry->d_fsdata; 512} 513 514static inline loff_t ceph_make_fpos(unsigned frag, unsigned off) 515{ 516 return ((loff_t)frag << 32) | (loff_t)off; 517} 518 519/* 520 * caps helpers 521 */ 522static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci) 523{ 524 return !RB_EMPTY_ROOT(&ci->i_caps); 525} 526 527extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented); 528extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t); 529extern int __ceph_caps_issued_other(struct ceph_inode_info *ci, 530 struct ceph_cap *cap); 531 532static inline int ceph_caps_issued(struct ceph_inode_info *ci) 533{ 534 int issued; 535 spin_lock(&ci->i_ceph_lock); 536 issued = __ceph_caps_issued(ci, NULL); 537 spin_unlock(&ci->i_ceph_lock); 538 return issued; 539} 540 541static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, 542 int touch) 543{ 544 int r; 545 spin_lock(&ci->i_ceph_lock); 546 r = __ceph_caps_issued_mask(ci, mask, touch); 547 spin_unlock(&ci->i_ceph_lock); 548 return r; 549} 550 551static inline int __ceph_caps_dirty(struct ceph_inode_info *ci) 552{ 553 return ci->i_dirty_caps | ci->i_flushing_caps; 554} 555extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask); 556 557extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci, 558 struct ceph_cap *ocap, int mask); 559extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask); 560extern int __ceph_caps_used(struct ceph_inode_info *ci); 561 562extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci); 563 564/* 565 * wanted, by virtue of open file modes AND cap refs (buffered/cached data) 566 */ 567static inline int __ceph_caps_wanted(struct ceph_inode_info *ci) 568{ 569 int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci); 570 if (w & CEPH_CAP_FILE_BUFFER) 571 w |= CEPH_CAP_FILE_EXCL; /* we want EXCL if dirty data */ 572 return w; 573} 574 575/* what the mds thinks we want */ 576extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci); 577 578extern void ceph_caps_init(struct ceph_mds_client *mdsc); 579extern void ceph_caps_finalize(struct ceph_mds_client *mdsc); 580extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta); 581extern void ceph_reserve_caps(struct ceph_mds_client *mdsc, 582 struct ceph_cap_reservation *ctx, int need); 583extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc, 584 struct ceph_cap_reservation *ctx); 585extern void ceph_reservation_status(struct ceph_fs_client *client, 586 int *total, int *avail, int *used, 587 int *reserved, int *min); 588 589 590 591/* 592 * we keep buffered readdir results attached to file->private_data 593 */ 594#define CEPH_F_SYNC 1 595#define CEPH_F_ATEND 2 596 597struct ceph_file_info { 598 short fmode; /* initialized on open */ 599 short flags; /* CEPH_F_* */ 600 601 /* readdir: position within the dir */ 602 u32 frag; 603 struct ceph_mds_request *last_readdir; 604 605 /* readdir: position within a frag */ 606 unsigned offset; /* offset of last chunk, adjusted for . and .. */ 607 unsigned next_offset; /* offset of next chunk (last_name's + 1) */ 608 char *last_name; /* last entry in previous chunk */ 609 struct dentry *dentry; /* next dentry (for dcache readdir) */ 610 int dir_release_count; 611 int dir_ordered_count; 612 613 /* used for -o dirstat read() on directory thing */ 614 char *dir_info; 615 int dir_info_len; 616}; 617 618 619 620/* 621 * A "snap realm" describes a subset of the file hierarchy sharing 622 * the same set of snapshots that apply to it. The realms themselves 623 * are organized into a hierarchy, such that children inherit (some of) 624 * the snapshots of their parents. 625 * 626 * All inodes within the realm that have capabilities are linked into a 627 * per-realm list. 628 */ 629struct ceph_snap_realm { 630 u64 ino; 631 atomic_t nref; 632 struct rb_node node; 633 634 u64 created, seq; 635 u64 parent_ino; 636 u64 parent_since; /* snapid when our current parent became so */ 637 638 u64 *prior_parent_snaps; /* snaps inherited from any parents we */ 639 u32 num_prior_parent_snaps; /* had prior to parent_since */ 640 u64 *snaps; /* snaps specific to this realm */ 641 u32 num_snaps; 642 643 struct ceph_snap_realm *parent; 644 struct list_head children; /* list of child realms */ 645 struct list_head child_item; 646 647 struct list_head empty_item; /* if i have ref==0 */ 648 649 struct list_head dirty_item; /* if realm needs new context */ 650 651 /* the current set of snaps for this realm */ 652 struct ceph_snap_context *cached_context; 653 654 struct list_head inodes_with_caps; 655 spinlock_t inodes_with_caps_lock; 656}; 657 658static inline int default_congestion_kb(void) 659{ 660 int congestion_kb; 661 662 /* 663 * Copied from NFS 664 * 665 * congestion size, scale with available memory. 666 * 667 * 64MB: 8192k 668 * 128MB: 11585k 669 * 256MB: 16384k 670 * 512MB: 23170k 671 * 1GB: 32768k 672 * 2GB: 46340k 673 * 4GB: 65536k 674 * 8GB: 92681k 675 * 16GB: 131072k 676 * 677 * This allows larger machines to have larger/more transfers. 678 * Limit the default to 256M 679 */ 680 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10); 681 if (congestion_kb > 256*1024) 682 congestion_kb = 256*1024; 683 684 return congestion_kb; 685} 686 687 688 689/* snap.c */ 690struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc, 691 u64 ino); 692extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc, 693 struct ceph_snap_realm *realm); 694extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc, 695 struct ceph_snap_realm *realm); 696extern int ceph_update_snap_trace(struct ceph_mds_client *m, 697 void *p, void *e, bool deletion, 698 struct ceph_snap_realm **realm_ret); 699extern void ceph_handle_snap(struct ceph_mds_client *mdsc, 700 struct ceph_mds_session *session, 701 struct ceph_msg *msg); 702extern void ceph_queue_cap_snap(struct ceph_inode_info *ci); 703extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci, 704 struct ceph_cap_snap *capsnap); 705extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc); 706extern int ceph_snap_init(void); 707extern void ceph_snap_exit(void); 708 709/* 710 * a cap_snap is "pending" if it is still awaiting an in-progress 711 * sync write (that may/may not still update size, mtime, etc.). 712 */ 713static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci) 714{ 715 return !list_empty(&ci->i_cap_snaps) && 716 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap, 717 ci_item)->writing; 718} 719 720/* inode.c */ 721extern const struct inode_operations ceph_file_iops; 722 723extern struct inode *ceph_alloc_inode(struct super_block *sb); 724extern void ceph_destroy_inode(struct inode *inode); 725extern int ceph_drop_inode(struct inode *inode); 726 727extern struct inode *ceph_get_inode(struct super_block *sb, 728 struct ceph_vino vino); 729extern struct inode *ceph_get_snapdir(struct inode *parent); 730extern int ceph_fill_file_size(struct inode *inode, int issued, 731 u32 truncate_seq, u64 truncate_size, u64 size); 732extern void ceph_fill_file_time(struct inode *inode, int issued, 733 u64 time_warp_seq, struct timespec *ctime, 734 struct timespec *mtime, struct timespec *atime); 735extern int ceph_fill_trace(struct super_block *sb, 736 struct ceph_mds_request *req, 737 struct ceph_mds_session *session); 738extern int ceph_readdir_prepopulate(struct ceph_mds_request *req, 739 struct ceph_mds_session *session); 740 741extern int ceph_inode_holds_cap(struct inode *inode, int mask); 742 743extern int ceph_inode_set_size(struct inode *inode, loff_t size); 744extern void __ceph_do_pending_vmtruncate(struct inode *inode); 745extern void ceph_queue_vmtruncate(struct inode *inode); 746 747extern void ceph_queue_invalidate(struct inode *inode); 748extern void ceph_queue_writeback(struct inode *inode); 749 750extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page, 751 int mask, bool force); 752static inline int ceph_do_getattr(struct inode *inode, int mask, bool force) 753{ 754 return __ceph_do_getattr(inode, NULL, mask, force); 755} 756extern int ceph_permission(struct inode *inode, int mask); 757extern int ceph_setattr(struct dentry *dentry, struct iattr *attr); 758extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry, 759 struct kstat *stat); 760 761/* xattr.c */ 762extern int ceph_setxattr(struct dentry *, const char *, const void *, 763 size_t, int); 764int __ceph_setxattr(struct dentry *, const char *, const void *, size_t, int); 765ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t); 766int __ceph_removexattr(struct dentry *, const char *); 767extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t); 768extern ssize_t ceph_listxattr(struct dentry *, char *, size_t); 769extern int ceph_removexattr(struct dentry *, const char *); 770extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci); 771extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci); 772extern void __init ceph_xattr_init(void); 773extern void ceph_xattr_exit(void); 774extern const struct xattr_handler *ceph_xattr_handlers[]; 775 776/* acl.c */ 777struct ceph_acls_info { 778 void *default_acl; 779 void *acl; 780 struct ceph_pagelist *pagelist; 781}; 782 783#ifdef CONFIG_CEPH_FS_POSIX_ACL 784 785struct posix_acl *ceph_get_acl(struct inode *, int); 786int ceph_set_acl(struct inode *inode, struct posix_acl *acl, int type); 787int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 788 struct ceph_acls_info *info); 789void ceph_init_inode_acls(struct inode *inode, struct ceph_acls_info *info); 790void ceph_release_acls_info(struct ceph_acls_info *info); 791 792static inline void ceph_forget_all_cached_acls(struct inode *inode) 793{ 794 forget_all_cached_acls(inode); 795} 796 797#else 798 799#define ceph_get_acl NULL 800#define ceph_set_acl NULL 801 802static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode, 803 struct ceph_acls_info *info) 804{ 805 return 0; 806} 807static inline void ceph_init_inode_acls(struct inode *inode, 808 struct ceph_acls_info *info) 809{ 810} 811static inline void ceph_release_acls_info(struct ceph_acls_info *info) 812{ 813} 814static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode) 815{ 816 return 0; 817} 818 819static inline void ceph_forget_all_cached_acls(struct inode *inode) 820{ 821} 822 823#endif 824 825/* caps.c */ 826extern const char *ceph_cap_string(int c); 827extern void ceph_handle_caps(struct ceph_mds_session *session, 828 struct ceph_msg *msg); 829extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc, 830 struct ceph_cap_reservation *ctx); 831extern void ceph_add_cap(struct inode *inode, 832 struct ceph_mds_session *session, u64 cap_id, 833 int fmode, unsigned issued, unsigned wanted, 834 unsigned cap, unsigned seq, u64 realmino, int flags, 835 struct ceph_cap **new_cap); 836extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release); 837extern void ceph_put_cap(struct ceph_mds_client *mdsc, 838 struct ceph_cap *cap); 839extern int ceph_is_any_caps(struct inode *inode); 840 841extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino, 842 u64 cap_id, u32 migrate_seq, u32 issue_seq); 843extern void ceph_queue_caps_release(struct inode *inode); 844extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc); 845extern int ceph_fsync(struct file *file, loff_t start, loff_t end, 846 int datasync); 847extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc, 848 struct ceph_mds_session *session); 849extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci, 850 int mds); 851extern int ceph_get_cap_mds(struct inode *inode); 852extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps); 853extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had); 854extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, 855 struct ceph_snap_context *snapc); 856extern void __ceph_flush_snaps(struct ceph_inode_info *ci, 857 struct ceph_mds_session **psession, 858 int again); 859extern void ceph_check_caps(struct ceph_inode_info *ci, int flags, 860 struct ceph_mds_session *session); 861extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc); 862extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc); 863 864extern int ceph_encode_inode_release(void **p, struct inode *inode, 865 int mds, int drop, int unless, int force); 866extern int ceph_encode_dentry_release(void **p, struct dentry *dn, 867 int mds, int drop, int unless); 868 869extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want, 870 loff_t endoff, int *got, struct page **pinned_page); 871 872/* for counting open files by mode */ 873static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode) 874{ 875 ci->i_nr_by_mode[mode]++; 876} 877extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode); 878 879/* addr.c */ 880extern const struct address_space_operations ceph_aops; 881extern int ceph_mmap(struct file *file, struct vm_area_struct *vma); 882 883/* file.c */ 884extern const struct file_operations ceph_file_fops; 885 886extern int ceph_open(struct inode *inode, struct file *file); 887extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry, 888 struct file *file, unsigned flags, umode_t mode, 889 int *opened); 890extern int ceph_release(struct inode *inode, struct file *filp); 891extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page, 892 char *data, size_t len); 893int ceph_uninline_data(struct file *filp, struct page *locked_page); 894/* dir.c */ 895extern const struct file_operations ceph_dir_fops; 896extern const struct file_operations ceph_snapdir_fops; 897extern const struct inode_operations ceph_dir_iops; 898extern const struct inode_operations ceph_snapdir_iops; 899extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops, 900 ceph_snapdir_dentry_ops; 901 902extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry); 903extern int ceph_handle_snapdir(struct ceph_mds_request *req, 904 struct dentry *dentry, int err); 905extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req, 906 struct dentry *dentry, int err); 907 908extern void ceph_dentry_lru_add(struct dentry *dn); 909extern void ceph_dentry_lru_touch(struct dentry *dn); 910extern void ceph_dentry_lru_del(struct dentry *dn); 911extern void ceph_invalidate_dentry_lease(struct dentry *dentry); 912extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn); 913extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry); 914 915/* 916 * our d_ops vary depending on whether the inode is live, 917 * snapshotted (read-only), or a virtual ".snap" directory. 918 */ 919int ceph_init_dentry(struct dentry *dentry); 920 921 922/* ioctl.c */ 923extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg); 924 925/* export.c */ 926extern const struct export_operations ceph_export_ops; 927 928/* locks.c */ 929extern __init void ceph_flock_init(void); 930extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl); 931extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl); 932extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num); 933extern int ceph_encode_locks_to_buffer(struct inode *inode, 934 struct ceph_filelock *flocks, 935 int num_fcntl_locks, 936 int num_flock_locks); 937extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks, 938 struct ceph_pagelist *pagelist, 939 int num_fcntl_locks, int num_flock_locks); 940extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c); 941 942/* debugfs.c */ 943extern int ceph_fs_debugfs_init(struct ceph_fs_client *client); 944extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client); 945 946#endif /* _FS_CEPH_SUPER_H */ 947