fs/ceph/super.h

   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
  17 #include <linux/ceph/libceph.h>
  18
  19 /* f_type in struct statfs */
  20 #define CEPH_SUPER_MAGIC 0x00c36400
  21
  22 /* large granularity for statfs utilization stats to facilitate
  23  * large volume sizes on 32-bit machines. */
  24 #define CEPH_BLOCK_SHIFT   20  /* 1 MB */
  25 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
  26
  27 #define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
  28 #define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
  29 #define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
  30
  31 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES)
  32
  33 #define ceph_set_mount_opt(fsc, opt) \
  34         (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
  35 #define ceph_test_mount_opt(fsc, opt) \
  36         (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
  37
  38 #define CEPH_MAX_READDIR_DEFAULT        1024
  39 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
  40 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
  41
  42 struct ceph_mount_options {
  43         int flags;
  44         int sb_flags;
  45
  46         int wsize;
  47         int rsize;            /* max readahead */
  48         int congestion_kb;    /* max writeback in flight */
  49         int caps_wanted_delay_min, caps_wanted_delay_max;
  50         int cap_release_safety;
  51         int max_readdir;       /* max readdir result (entires) */
  52         int max_readdir_bytes; /* max readdir result (bytes) */
  53
  54         /*
  55          * everything above this point can be memcmp'd; everything below
  56          * is handled in compare_mount_options()
  57          */
  58
  59         char *snapdir_name;   /* default ".snap" */
  60 };
  61
  62 struct ceph_fs_client {
  63         struct super_block *sb;
  64
  65         struct ceph_mount_options *mount_options;
  66         struct ceph_client *client;
  67
  68         unsigned long mount_state;
  69         int min_caps;                  /* min caps i added */
  70
  71         struct ceph_mds_client *mdsc;
  72
  73         /* writeback */
  74         mempool_t *wb_pagevec_pool;
  75         struct workqueue_struct *wb_wq;
  76         struct workqueue_struct *pg_inv_wq;
  77         struct workqueue_struct *trunc_wq;
  78         atomic_long_t writeback_count;
  79
  80         struct backing_dev_info backing_dev_info;
  81
  82 #ifdef CONFIG_DEBUG_FS
  83         struct dentry *debugfs_dentry_lru, *debugfs_caps;
  84         struct dentry *debugfs_congestion_kb;
  85         struct dentry *debugfs_bdi;
  86         struct dentry *debugfs_mdsc, *debugfs_mdsmap;
  87 #endif
  88 };
  89
  90
  91 /*
  92  * File i/o capability.  This tracks shared state with the metadata
  93  * server that allows us to cache or writeback attributes or to read
  94  * and write data.  For any given inode, we should have one or more
  95  * capabilities, one issued by each metadata server, and our
  96  * cumulative access is the OR of all issued capabilities.
  97  *
  98  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
  99  * session capability lists.
 100  */
 101 struct ceph_cap {
 102         struct ceph_inode_info *ci;
 103         struct rb_node ci_node;          /* per-ci cap tree */
 104         struct ceph_mds_session *session;
 105         struct list_head session_caps;   /* per-session caplist */
 106         int mds;
 107         u64 cap_id;       /* unique cap id (mds provided) */
 108         int issued;       /* latest, from the mds */
 109         int implemented;  /* implemented superset of issued (for revocation) */
 110         int mds_wanted;
 111         u32 seq, issue_seq, mseq;
 112         u32 cap_gen;      /* active/stale cycle */
 113         unsigned long last_used;
 114         struct list_head caps_item;
 115 };
 116
 117 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
 118 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
 119 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
 120
 121 /*
 122  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 123  * we first complete any in-process sync writes and writeback any dirty
 124  * data before flushing the snapped state (tracked here) back to the MDS.
 125  */
 126 struct ceph_cap_snap {
 127         atomic_t nref;
 128         struct ceph_inode_info *ci;
 129         struct list_head ci_item, flushing_item;
 130
 131         u64 follows, flush_tid;
 132         int issued, dirty;
 133         struct ceph_snap_context *context;
 134
 135         mode_t mode;
 136         uid_t uid;
 137         gid_t gid;
 138
 139         struct ceph_buffer *xattr_blob;
 140         u64 xattr_version;
 141
 142         u64 size;
 143         struct timespec mtime, atime, ctime;
 144         u64 time_warp_seq;
 145         int writing;   /* a sync write is still in progress */
 146         int dirty_pages;     /* dirty pages awaiting writeback */
 147 };
 148
 149 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 150 {
 151         if (atomic_dec_and_test(&capsnap->nref)) {
 152                 if (capsnap->xattr_blob)
 153                         ceph_buffer_put(capsnap->xattr_blob);
 154                 kfree(capsnap);
 155         }
 156 }
 157
 158 /*
 159  * The frag tree describes how a directory is fragmented, potentially across
 160  * multiple metadata servers.  It is also used to indicate points where
 161  * metadata authority is delegated, and whether/where metadata is replicated.
 162  *
 163  * A _leaf_ frag will be present in the i_fragtree IFF there is
 164  * delegation info.  That is, if mds >= 0 || ndist > 0.
 165  */
 166 #define CEPH_MAX_DIRFRAG_REP 4
 167
 168 struct ceph_inode_frag {
 169         struct rb_node node;
 170
 171         /* fragtree state */
 172         u32 frag;
 173         int split_by;         /* i.e. 2^(split_by) children */
 174
 175         /* delegation and replication info */
 176         int mds;              /* -1 if same authority as parent */
 177         int ndist;            /* >0 if replicated */
 178         int dist[CEPH_MAX_DIRFRAG_REP];
 179 };
 180
 181 /*
 182  * We cache inode xattrs as an encoded blob until they are first used,
 183  * at which point we parse them into an rbtree.
 184  */
 185 struct ceph_inode_xattr {
 186         struct rb_node node;
 187
 188         const char *name;
 189         int name_len;
 190         const char *val;
 191         int val_len;
 192         int dirty;
 193
 194         int should_free_name;
 195         int should_free_val;
 196 };
 197
 198 /*
 199  * Ceph dentry state
 200  */
 201 struct ceph_dentry_info {
 202         struct ceph_mds_session *lease_session;
 203         u32 lease_gen, lease_shared_gen;
 204         u32 lease_seq;
 205         unsigned long lease_renew_after, lease_renew_from;
 206         struct list_head lru;
 207         struct dentry *dentry;
 208         u64 time;
 209         u64 offset;
 210 };
 211
 212 struct ceph_inode_xattrs_info {
 213         /*
 214          * (still encoded) xattr blob. we avoid the overhead of parsing
 215          * this until someone actually calls getxattr, etc.
 216          *
 217          * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 218          * NULL means we don't know.
 219         */
 220         struct ceph_buffer *blob, *prealloc_blob;
 221
 222         struct rb_root index;
 223         bool dirty;
 224         int count;
 225         int names_size;
 226         int vals_size;
 227         u64 version, index_version;
 228 };
 229
 230 /*
 231  * Ceph inode.
 232  */
 233 struct ceph_inode_info {
 234         struct ceph_vino i_vino;   /* ceph ino + snap */
 235
 236         u64 i_version;
 237         u32 i_time_warp_seq;
 238
 239         unsigned i_ceph_flags;
 240         unsigned long i_release_count;
 241
 242         struct ceph_dir_layout i_dir_layout;
 243         struct ceph_file_layout i_layout;
 244         char *i_symlink;
 245
 246         /* for dirs */
 247         struct timespec i_rctime;
 248         u64 i_rbytes, i_rfiles, i_rsubdirs;
 249         u64 i_files, i_subdirs;
 250         u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
 251
 252         struct rb_root i_fragtree;
 253         struct mutex i_fragtree_mutex;
 254
 255         struct ceph_inode_xattrs_info i_xattrs;
 256
 257         /* capabilities.  protected _both_ by i_lock and cap->session's
 258          * s_mutex. */
 259         struct rb_root i_caps;           /* cap list */
 260         struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 261         unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 262         struct list_head i_dirty_item, i_flushing_item;
 263         u64 i_cap_flush_seq;
 264         /* we need to track cap writeback on a per-cap-bit basis, to allow
 265          * overlapping, pipelined cap flushes to the mds.  we can probably
 266          * reduce the tid to 8 bits if we're concerned about inode size. */
 267         u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
 268         wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 269         unsigned long i_hold_caps_min; /* jiffies */
 270         unsigned long i_hold_caps_max; /* jiffies */
 271         struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 272         int i_cap_exporting_mds;         /* to handle cap migration between */
 273         unsigned i_cap_exporting_mseq;   /*  mds's. */
 274         unsigned i_cap_exporting_issued;
 275         struct ceph_cap_reservation i_cap_migration_resv;
 276         struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 277         struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 278                                                     dirty|flushing caps */
 279         unsigned i_snap_caps;           /* cap bits for snapped files */
 280
 281         int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 282
 283         u32 i_truncate_seq;        /* last truncate to smaller size */
 284         u64 i_truncate_size;       /*  and the size we last truncated down to */
 285         int i_truncate_pending;    /*  still need to call vmtruncate */
 286
 287         u64 i_max_size;            /* max file size authorized by mds */
 288         u64 i_reported_size; /* (max_)size reported to or requested of mds */
 289         u64 i_wanted_max_size;     /* offset we'd like to write too */
 290         u64 i_requested_max_size;  /* max_size we've requested */
 291
 292         /* held references to caps */
 293         int i_pin_ref;
 294         int i_rd_ref, i_rdcache_ref, i_wr_ref;
 295         int i_wrbuffer_ref, i_wrbuffer_ref_head;
 296         u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 297         u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 298         u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 299
 300         struct list_head i_unsafe_writes; /* uncommitted sync writes */
 301         struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 302         spinlock_t i_unsafe_lock;
 303
 304         struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 305         int i_snap_realm_counter; /* snap realm (if caps) */
 306         struct list_head i_snap_realm_item;
 307         struct list_head i_snap_flush_item;
 308
 309         struct work_struct i_wb_work;  /* writeback work */
 310         struct work_struct i_pg_inv_work;  /* page invalidation work */
 311
 312         struct work_struct i_vmtruncate_work;
 313
 314         struct inode vfs_inode; /* at end */
 315 };
 316
 317 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 318 {
 319         return container_of(inode, struct ceph_inode_info, vfs_inode);
 320 }
 321
 322 static inline struct ceph_vino ceph_vino(struct inode *inode)
 323 {
 324         return ceph_inode(inode)->i_vino;
 325 }
 326
 327 /*
 328  * ino_t is <64 bits on many architectures, blech.
 329  *
 330  * don't include snap in ino hash, at least for now.
 331  */
 332 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 333 {
 334         ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
 335 #if BITS_PER_LONG == 32
 336         ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
 337         if (!ino)
 338                 ino = 1;
 339 #endif
 340         return ino;
 341 }
 342
 343 /* for printf-style formatting */
 344 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 345
 346 static inline u64 ceph_ino(struct inode *inode)
 347 {
 348         return ceph_inode(inode)->i_vino.ino;
 349 }
 350 static inline u64 ceph_snap(struct inode *inode)
 351 {
 352         return ceph_inode(inode)->i_vino.snap;
 353 }
 354
 355 static inline int ceph_ino_compare(struct inode *inode, void *data)
 356 {
 357         struct ceph_vino *pvino = (struct ceph_vino *)data;
 358         struct ceph_inode_info *ci = ceph_inode(inode);
 359         return ci->i_vino.ino == pvino->ino &&
 360                 ci->i_vino.snap == pvino->snap;
 361 }
 362
 363 static inline struct inode *ceph_find_inode(struct super_block *sb,
 364                                             struct ceph_vino vino)
 365 {
 366         ino_t t = ceph_vino_to_ino(vino);
 367         return ilookup5(sb, t, ceph_ino_compare, &vino);
 368 }
 369
 370
 371 /*
 372  * Ceph inode.
 373  */
 374 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
 375 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 376 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 377 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
 378
 379 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 380 {
 381         struct ceph_inode_info *ci = ceph_inode(inode);
 382
 383         spin_lock(&inode->i_lock);
 384         ci->i_ceph_flags &= ~mask;
 385         spin_unlock(&inode->i_lock);
 386 }
 387
 388 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 389 {
 390         struct ceph_inode_info *ci = ceph_inode(inode);
 391
 392         spin_lock(&inode->i_lock);
 393         ci->i_ceph_flags |= mask;
 394         spin_unlock(&inode->i_lock);
 395 }
 396
 397 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
 398 {
 399         struct ceph_inode_info *ci = ceph_inode(inode);
 400         bool r;
 401
 402         spin_lock(&inode->i_lock);
 403         r = (ci->i_ceph_flags & mask) == mask;
 404         spin_unlock(&inode->i_lock);
 405         return r;
 406 }
 407
 408
 409 /* find a specific frag @f */
 410 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 411                                                 u32 f);
 412
 413 /*
 414  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 415  * exists
 416  */
 417 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 418                             struct ceph_inode_frag *pfrag,
 419                             int *found);
 420
 421 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 422 {
 423         return (struct ceph_dentry_info *)dentry->d_fsdata;
 424 }
 425
 426 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 427 {
 428         return ((loff_t)frag << 32) | (loff_t)off;
 429 }
 430
 431 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
 432 {
 433         ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
 434         inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
 435         return 0;
 436 }
 437
 438 /*
 439  * caps helpers
 440  */
 441 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 442 {
 443         return !RB_EMPTY_ROOT(&ci->i_caps);
 444 }
 445
 446 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 447 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 448 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 449                                     struct ceph_cap *cap);
 450
 451 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 452 {
 453         int issued;
 454         spin_lock(&ci->vfs_inode.i_lock);
 455         issued = __ceph_caps_issued(ci, NULL);
 456         spin_unlock(&ci->vfs_inode.i_lock);
 457         return issued;
 458 }
 459
 460 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 461                                         int touch)
 462 {
 463         int r;
 464         spin_lock(&ci->vfs_inode.i_lock);
 465         r = __ceph_caps_issued_mask(ci, mask, touch);
 466         spin_unlock(&ci->vfs_inode.i_lock);
 467         return r;
 468 }
 469
 470 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 471 {
 472         return ci->i_dirty_caps | ci->i_flushing_caps;
 473 }
 474 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 475
 476 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 477 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 478
 479 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 480
 481 /*
 482  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 483  */
 484 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 485 {
 486         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 487         if (w & CEPH_CAP_FILE_BUFFER)
 488                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 489         return w;
 490 }
 491
 492 /* what the mds thinks we want */
 493 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 494
 495 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 496 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 497 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
 498 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 499                              struct ceph_cap_reservation *ctx, int need);
 500 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 501                                struct ceph_cap_reservation *ctx);
 502 extern void ceph_reservation_status(struct ceph_fs_client *client,
 503                                     int *total, int *avail, int *used,
 504                                     int *reserved, int *min);
 505
 506 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
 507 {
 508         return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 509 }
 510
 511 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
 512 {
 513         return (struct ceph_fs_client *)sb->s_fs_info;
 514 }
 515
 516
 517 /*
 518  * we keep buffered readdir results attached to file->private_data
 519  */
 520 struct ceph_file_info {
 521         int fmode;     /* initialized on open */
 522
 523         /* readdir: position within the dir */
 524         u32 frag;
 525         struct ceph_mds_request *last_readdir;
 526         int at_end;
 527
 528         /* readdir: position within a frag */
 529         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 530         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 531         char *last_name;       /* last entry in previous chunk */
 532         struct dentry *dentry; /* next dentry (for dcache readdir) */
 533         unsigned long dir_release_count;
 534
 535         /* used for -o dirstat read() on directory thing */
 536         char *dir_info;
 537         int dir_info_len;
 538 };
 539
 540
 541
 542 /*
 543  * A "snap realm" describes a subset of the file hierarchy sharing
 544  * the same set of snapshots that apply to it.  The realms themselves
 545  * are organized into a hierarchy, such that children inherit (some of)
 546  * the snapshots of their parents.
 547  *
 548  * All inodes within the realm that have capabilities are linked into a
 549  * per-realm list.
 550  */
 551 struct ceph_snap_realm {
 552         u64 ino;
 553         atomic_t nref;
 554         struct rb_node node;
 555
 556         u64 created, seq;
 557         u64 parent_ino;
 558         u64 parent_since;   /* snapid when our current parent became so */
 559
 560         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 561         int num_prior_parent_snaps;   /*  had prior to parent_since */
 562         u64 *snaps;                   /* snaps specific to this realm */
 563         int num_snaps;
 564
 565         struct ceph_snap_realm *parent;
 566         struct list_head children;       /* list of child realms */
 567         struct list_head child_item;
 568
 569         struct list_head empty_item;     /* if i have ref==0 */
 570
 571         struct list_head dirty_item;     /* if realm needs new context */
 572
 573         /* the current set of snaps for this realm */
 574         struct ceph_snap_context *cached_context;
 575
 576         struct list_head inodes_with_caps;
 577         spinlock_t inodes_with_caps_lock;
 578 };
 579
 580 static inline int default_congestion_kb(void)
 581 {
 582         int congestion_kb;
 583
 584         /*
 585          * Copied from NFS
 586          *
 587          * congestion size, scale with available memory.
 588          *
 589          *  64MB:    8192k
 590          * 128MB:   11585k
 591          * 256MB:   16384k
 592          * 512MB:   23170k
 593          *   1GB:   32768k
 594          *   2GB:   46340k
 595          *   4GB:   65536k
 596          *   8GB:   92681k
 597          *  16GB:  131072k
 598          *
 599          * This allows larger machines to have larger/more transfers.
 600          * Limit the default to 256M
 601          */
 602         congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
 603         if (congestion_kb > 256*1024)
 604                 congestion_kb = 256*1024;
 605
 606         return congestion_kb;
 607 }
 608
 609
 610
 611 /* snap.c */
 612 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 613                                                u64 ino);
 614 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 615                                 struct ceph_snap_realm *realm);
 616 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 617                                 struct ceph_snap_realm *realm);
 618 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 619                                   void *p, void *e, bool deletion);
 620 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 621                              struct ceph_mds_session *session,
 622                              struct ceph_msg *msg);
 623 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 624 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 625                                   struct ceph_cap_snap *capsnap);
 626 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 627
 628 /*
 629  * a cap_snap is "pending" if it is still awaiting an in-progress
 630  * sync write (that may/may not still update size, mtime, etc.).
 631  */
 632 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 633 {
 634         return !list_empty(&ci->i_cap_snaps) &&
 635                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 636                            ci_item)->writing;
 637 }
 638
 639 /* inode.c */
 640 extern const struct inode_operations ceph_file_iops;
 641
 642 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 643 extern void ceph_destroy_inode(struct inode *inode);
 644
 645 extern struct inode *ceph_get_inode(struct super_block *sb,
 646                                     struct ceph_vino vino);
 647 extern struct inode *ceph_get_snapdir(struct inode *parent);
 648 extern int ceph_fill_file_size(struct inode *inode, int issued,
 649                                u32 truncate_seq, u64 truncate_size, u64 size);
 650 extern void ceph_fill_file_time(struct inode *inode, int issued,
 651                                 u64 time_warp_seq, struct timespec *ctime,
 652                                 struct timespec *mtime, struct timespec *atime);
 653 extern int ceph_fill_trace(struct super_block *sb,
 654                            struct ceph_mds_request *req,
 655                            struct ceph_mds_session *session);
 656 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 657                                     struct ceph_mds_session *session);
 658
 659 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 660
 661 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 662 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 663 extern void ceph_queue_vmtruncate(struct inode *inode);
 664
 665 extern void ceph_queue_invalidate(struct inode *inode);
 666 extern void ceph_queue_writeback(struct inode *inode);
 667
 668 extern int ceph_do_getattr(struct inode *inode, int mask);
 669 extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
 670 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 671 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 672                         struct kstat *stat);
 673
 674 /* xattr.c */
 675 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 676                          size_t, int);
 677 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 678 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 679 extern int ceph_removexattr(struct dentry *, const char *);
 680 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 681 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 682
 683 /* caps.c */
 684 extern const char *ceph_cap_string(int c);
 685 extern void ceph_handle_caps(struct ceph_mds_session *session,
 686                              struct ceph_msg *msg);
 687 extern int ceph_add_cap(struct inode *inode,
 688                         struct ceph_mds_session *session, u64 cap_id,
 689                         int fmode, unsigned issued, unsigned wanted,
 690                         unsigned cap, unsigned seq, u64 realmino, int flags,
 691                         struct ceph_cap_reservation *caps_reservation);
 692 extern void __ceph_remove_cap(struct ceph_cap *cap);
 693 static inline void ceph_remove_cap(struct ceph_cap *cap)
 694 {
 695         struct inode *inode = &cap->ci->vfs_inode;
 696         spin_lock(&inode->i_lock);
 697         __ceph_remove_cap(cap);
 698         spin_unlock(&inode->i_lock);
 699 }
 700 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
 701                          struct ceph_cap *cap);
 702
 703 extern void ceph_queue_caps_release(struct inode *inode);
 704 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
 705 extern int ceph_fsync(struct file *file, int datasync);
 706 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 707                                     struct ceph_mds_session *session);
 708 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
 709                                              int mds);
 710 extern int ceph_get_cap_mds(struct inode *inode);
 711 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 712 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 713 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 714                                        struct ceph_snap_context *snapc);
 715 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 716                                struct ceph_mds_session **psession,
 717                                int again);
 718 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 719                             struct ceph_mds_session *session);
 720 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 721 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 722
 723 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 724                                      int mds, int drop, int unless, int force);
 725 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 726                                       int mds, int drop, int unless);
 727
 728 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 729                          int *got, loff_t endoff);
 730
 731 /* for counting open files by mode */
 732 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 733 {
 734         ci->i_nr_by_mode[mode]++;
 735 }
 736 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 737
 738 /* addr.c */
 739 extern const struct address_space_operations ceph_aops;
 740 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 741
 742 /* file.c */
 743 extern const struct file_operations ceph_file_fops;
 744 extern const struct address_space_operations ceph_aops;
 745 extern int ceph_copy_to_page_vector(struct page **pages,
 746                                     const char *data,
 747                                     loff_t off, size_t len);
 748 extern int ceph_copy_from_page_vector(struct page **pages,
 749                                     char *data,
 750                                     loff_t off, size_t len);
 751 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
 752 extern int ceph_open(struct inode *inode, struct file *file);
 753 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 754                                        struct nameidata *nd, int mode,
 755                                        int locked_dir);
 756 extern int ceph_release(struct inode *inode, struct file *filp);
 757
 758 /* dir.c */
 759 extern const struct file_operations ceph_dir_fops;
 760 extern const struct inode_operations ceph_dir_iops;
 761 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 762         ceph_snapdir_dentry_ops;
 763
 764 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 765 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 766                                          struct dentry *dentry, int err);
 767
 768 extern void ceph_dentry_lru_add(struct dentry *dn);
 769 extern void ceph_dentry_lru_touch(struct dentry *dn);
 770 extern void ceph_dentry_lru_del(struct dentry *dn);
 771 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 772 extern unsigned ceph_dentry_hash(struct dentry *dn);
 773
 774 /*
 775  * our d_ops vary depending on whether the inode is live,
 776  * snapshotted (read-only), or a virtual ".snap" directory.
 777  */
 778 int ceph_init_dentry(struct dentry *dentry);
 779
 780
 781 /* ioctl.c */
 782 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 783
 784 /* export.c */
 785 extern const struct export_operations ceph_export_ops;
 786
 787 /* locks.c */
 788 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
 789 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
 790 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
 791 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
 792                              int p_locks, int f_locks);
 793 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 794
 795 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
 796 {
 797         if (dentry && dentry->d_parent)
 798                 return dentry->d_parent->d_inode;
 799
 800         return NULL;
 801 }
 802
 803 /* debugfs.c */
 804 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 805 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
 806
 807 #endif /* _FS_CEPH_SUPER_H */