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_file_layout i_layout;
 243         char *i_symlink;
 244
 245         /* for dirs */
 246         struct timespec i_rctime;
 247         u64 i_rbytes, i_rfiles, i_rsubdirs;
 248         u64 i_files, i_subdirs;
 249         u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
 250
 251         struct rb_root i_fragtree;
 252         struct mutex i_fragtree_mutex;
 253
 254         struct ceph_inode_xattrs_info i_xattrs;
 255
 256         /* capabilities.  protected _both_ by i_lock and cap->session's
 257          * s_mutex. */
 258         struct rb_root i_caps;           /* cap list */
 259         struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 260         unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 261         struct list_head i_dirty_item, i_flushing_item;
 262         u64 i_cap_flush_seq;
 263         /* we need to track cap writeback on a per-cap-bit basis, to allow
 264          * overlapping, pipelined cap flushes to the mds.  we can probably
 265          * reduce the tid to 8 bits if we're concerned about inode size. */
 266         u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
 267         wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 268         unsigned long i_hold_caps_min; /* jiffies */
 269         unsigned long i_hold_caps_max; /* jiffies */
 270         struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 271         int i_cap_exporting_mds;         /* to handle cap migration between */
 272         unsigned i_cap_exporting_mseq;   /*  mds's. */
 273         unsigned i_cap_exporting_issued;
 274         struct ceph_cap_reservation i_cap_migration_resv;
 275         struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 276         struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 277                                                     dirty|flushing caps */
 278         unsigned i_snap_caps;           /* cap bits for snapped files */
 279
 280         int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 281
 282         u32 i_truncate_seq;        /* last truncate to smaller size */
 283         u64 i_truncate_size;       /*  and the size we last truncated down to */
 284         int i_truncate_pending;    /*  still need to call vmtruncate */
 285
 286         u64 i_max_size;            /* max file size authorized by mds */
 287         u64 i_reported_size; /* (max_)size reported to or requested of mds */
 288         u64 i_wanted_max_size;     /* offset we'd like to write too */
 289         u64 i_requested_max_size;  /* max_size we've requested */
 290
 291         /* held references to caps */
 292         int i_pin_ref;
 293         int i_rd_ref, i_rdcache_ref, i_wr_ref;
 294         int i_wrbuffer_ref, i_wrbuffer_ref_head;
 295         u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 296         u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 297         u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 298
 299         struct list_head i_unsafe_writes; /* uncommitted sync writes */
 300         struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 301         spinlock_t i_unsafe_lock;
 302
 303         struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 304         int i_snap_realm_counter; /* snap realm (if caps) */
 305         struct list_head i_snap_realm_item;
 306         struct list_head i_snap_flush_item;
 307
 308         struct work_struct i_wb_work;  /* writeback work */
 309         struct work_struct i_pg_inv_work;  /* page invalidation work */
 310
 311         struct work_struct i_vmtruncate_work;
 312
 313         struct inode vfs_inode; /* at end */
 314 };
 315
 316 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 317 {
 318         return container_of(inode, struct ceph_inode_info, vfs_inode);
 319 }
 320
 321 static inline struct ceph_vino ceph_vino(struct inode *inode)
 322 {
 323         return ceph_inode(inode)->i_vino;
 324 }
 325
 326 /*
 327  * ino_t is <64 bits on many architectures, blech.
 328  *
 329  * don't include snap in ino hash, at least for now.
 330  */
 331 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 332 {
 333         ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
 334 #if BITS_PER_LONG == 32
 335         ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
 336         if (!ino)
 337                 ino = 1;
 338 #endif
 339         return ino;
 340 }
 341
 342 /* for printf-style formatting */
 343 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 344
 345 static inline u64 ceph_ino(struct inode *inode)
 346 {
 347         return ceph_inode(inode)->i_vino.ino;
 348 }
 349 static inline u64 ceph_snap(struct inode *inode)
 350 {
 351         return ceph_inode(inode)->i_vino.snap;
 352 }
 353
 354 static inline int ceph_ino_compare(struct inode *inode, void *data)
 355 {
 356         struct ceph_vino *pvino = (struct ceph_vino *)data;
 357         struct ceph_inode_info *ci = ceph_inode(inode);
 358         return ci->i_vino.ino == pvino->ino &&
 359                 ci->i_vino.snap == pvino->snap;
 360 }
 361
 362 static inline struct inode *ceph_find_inode(struct super_block *sb,
 363                                             struct ceph_vino vino)
 364 {
 365         ino_t t = ceph_vino_to_ino(vino);
 366         return ilookup5(sb, t, ceph_ino_compare, &vino);
 367 }
 368
 369
 370 /*
 371  * Ceph inode.
 372  */
 373 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
 374 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 375 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 376 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
 377
 378 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 379 {
 380         struct ceph_inode_info *ci = ceph_inode(inode);
 381
 382         spin_lock(&inode->i_lock);
 383         ci->i_ceph_flags &= ~mask;
 384         spin_unlock(&inode->i_lock);
 385 }
 386
 387 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 388 {
 389         struct ceph_inode_info *ci = ceph_inode(inode);
 390
 391         spin_lock(&inode->i_lock);
 392         ci->i_ceph_flags |= mask;
 393         spin_unlock(&inode->i_lock);
 394 }
 395
 396 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
 397 {
 398         struct ceph_inode_info *ci = ceph_inode(inode);
 399         bool r;
 400
 401         spin_lock(&inode->i_lock);
 402         r = (ci->i_ceph_flags & mask) == mask;
 403         spin_unlock(&inode->i_lock);
 404         return r;
 405 }
 406
 407
 408 /* find a specific frag @f */
 409 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 410                                                 u32 f);
 411
 412 /*
 413  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 414  * exists
 415  */
 416 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 417                             struct ceph_inode_frag *pfrag,
 418                             int *found);
 419
 420 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 421 {
 422         return (struct ceph_dentry_info *)dentry->d_fsdata;
 423 }
 424
 425 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 426 {
 427         return ((loff_t)frag << 32) | (loff_t)off;
 428 }
 429
 430 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
 431 {
 432         ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
 433         inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
 434         return 0;
 435 }
 436
 437 /*
 438  * caps helpers
 439  */
 440 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 441 {
 442         return !RB_EMPTY_ROOT(&ci->i_caps);
 443 }
 444
 445 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 446 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 447 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 448                                     struct ceph_cap *cap);
 449
 450 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 451 {
 452         int issued;
 453         spin_lock(&ci->vfs_inode.i_lock);
 454         issued = __ceph_caps_issued(ci, NULL);
 455         spin_unlock(&ci->vfs_inode.i_lock);
 456         return issued;
 457 }
 458
 459 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 460                                         int touch)
 461 {
 462         int r;
 463         spin_lock(&ci->vfs_inode.i_lock);
 464         r = __ceph_caps_issued_mask(ci, mask, touch);
 465         spin_unlock(&ci->vfs_inode.i_lock);
 466         return r;
 467 }
 468
 469 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 470 {
 471         return ci->i_dirty_caps | ci->i_flushing_caps;
 472 }
 473 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 474
 475 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 476 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 477
 478 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 479
 480 /*
 481  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 482  */
 483 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 484 {
 485         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 486         if (w & CEPH_CAP_FILE_BUFFER)
 487                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 488         return w;
 489 }
 490
 491 /* what the mds thinks we want */
 492 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 493
 494 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 495 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 496 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
 497 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 498                              struct ceph_cap_reservation *ctx, int need);
 499 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 500                                struct ceph_cap_reservation *ctx);
 501 extern void ceph_reservation_status(struct ceph_fs_client *client,
 502                                     int *total, int *avail, int *used,
 503                                     int *reserved, int *min);
 504
 505 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
 506 {
 507         return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 508 }
 509
 510 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
 511 {
 512         return (struct ceph_fs_client *)sb->s_fs_info;
 513 }
 514
 515
 516 /*
 517  * we keep buffered readdir results attached to file->private_data
 518  */
 519 struct ceph_file_info {
 520         int fmode;     /* initialized on open */
 521
 522         /* readdir: position within the dir */
 523         u32 frag;
 524         struct ceph_mds_request *last_readdir;
 525         int at_end;
 526
 527         /* readdir: position within a frag */
 528         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 529         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 530         char *last_name;       /* last entry in previous chunk */
 531         struct dentry *dentry; /* next dentry (for dcache readdir) */
 532         unsigned long dir_release_count;
 533
 534         /* used for -o dirstat read() on directory thing */
 535         char *dir_info;
 536         int dir_info_len;
 537 };
 538
 539
 540
 541 /*
 542  * A "snap realm" describes a subset of the file hierarchy sharing
 543  * the same set of snapshots that apply to it.  The realms themselves
 544  * are organized into a hierarchy, such that children inherit (some of)
 545  * the snapshots of their parents.
 546  *
 547  * All inodes within the realm that have capabilities are linked into a
 548  * per-realm list.
 549  */
 550 struct ceph_snap_realm {
 551         u64 ino;
 552         atomic_t nref;
 553         struct rb_node node;
 554
 555         u64 created, seq;
 556         u64 parent_ino;
 557         u64 parent_since;   /* snapid when our current parent became so */
 558
 559         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 560         int num_prior_parent_snaps;   /*  had prior to parent_since */
 561         u64 *snaps;                   /* snaps specific to this realm */
 562         int num_snaps;
 563
 564         struct ceph_snap_realm *parent;
 565         struct list_head children;       /* list of child realms */
 566         struct list_head child_item;
 567
 568         struct list_head empty_item;     /* if i have ref==0 */
 569
 570         struct list_head dirty_item;     /* if realm needs new context */
 571
 572         /* the current set of snaps for this realm */
 573         struct ceph_snap_context *cached_context;
 574
 575         struct list_head inodes_with_caps;
 576         spinlock_t inodes_with_caps_lock;
 577 };
 578
 579 static inline int default_congestion_kb(void)
 580 {
 581         int congestion_kb;
 582
 583         /*
 584          * Copied from NFS
 585          *
 586          * congestion size, scale with available memory.
 587          *
 588          *  64MB:    8192k
 589          * 128MB:   11585k
 590          * 256MB:   16384k
 591          * 512MB:   23170k
 592          *   1GB:   32768k
 593          *   2GB:   46340k
 594          *   4GB:   65536k
 595          *   8GB:   92681k
 596          *  16GB:  131072k
 597          *
 598          * This allows larger machines to have larger/more transfers.
 599          * Limit the default to 256M
 600          */
 601         congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
 602         if (congestion_kb > 256*1024)
 603                 congestion_kb = 256*1024;
 604
 605         return congestion_kb;
 606 }
 607
 608
 609
 610 /* snap.c */
 611 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 612                                                u64 ino);
 613 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 614                                 struct ceph_snap_realm *realm);
 615 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 616                                 struct ceph_snap_realm *realm);
 617 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 618                                   void *p, void *e, bool deletion);
 619 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 620                              struct ceph_mds_session *session,
 621                              struct ceph_msg *msg);
 622 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 623 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 624                                   struct ceph_cap_snap *capsnap);
 625 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 626
 627 /*
 628  * a cap_snap is "pending" if it is still awaiting an in-progress
 629  * sync write (that may/may not still update size, mtime, etc.).
 630  */
 631 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 632 {
 633         return !list_empty(&ci->i_cap_snaps) &&
 634                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 635                            ci_item)->writing;
 636 }
 637
 638 /* inode.c */
 639 extern const struct inode_operations ceph_file_iops;
 640
 641 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 642 extern void ceph_destroy_inode(struct inode *inode);
 643
 644 extern struct inode *ceph_get_inode(struct super_block *sb,
 645                                     struct ceph_vino vino);
 646 extern struct inode *ceph_get_snapdir(struct inode *parent);
 647 extern int ceph_fill_file_size(struct inode *inode, int issued,
 648                                u32 truncate_seq, u64 truncate_size, u64 size);
 649 extern void ceph_fill_file_time(struct inode *inode, int issued,
 650                                 u64 time_warp_seq, struct timespec *ctime,
 651                                 struct timespec *mtime, struct timespec *atime);
 652 extern int ceph_fill_trace(struct super_block *sb,
 653                            struct ceph_mds_request *req,
 654                            struct ceph_mds_session *session);
 655 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 656                                     struct ceph_mds_session *session);
 657
 658 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 659
 660 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 661 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 662 extern void ceph_queue_vmtruncate(struct inode *inode);
 663
 664 extern void ceph_queue_invalidate(struct inode *inode);
 665 extern void ceph_queue_writeback(struct inode *inode);
 666
 667 extern int ceph_do_getattr(struct inode *inode, int mask);
 668 extern int ceph_permission(struct inode *inode, int mask);
 669 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 670 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 671                         struct kstat *stat);
 672
 673 /* xattr.c */
 674 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 675                          size_t, int);
 676 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 677 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 678 extern int ceph_removexattr(struct dentry *, const char *);
 679 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 680 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 681
 682 /* caps.c */
 683 extern const char *ceph_cap_string(int c);
 684 extern void ceph_handle_caps(struct ceph_mds_session *session,
 685                              struct ceph_msg *msg);
 686 extern int ceph_add_cap(struct inode *inode,
 687                         struct ceph_mds_session *session, u64 cap_id,
 688                         int fmode, unsigned issued, unsigned wanted,
 689                         unsigned cap, unsigned seq, u64 realmino, int flags,
 690                         struct ceph_cap_reservation *caps_reservation);
 691 extern void __ceph_remove_cap(struct ceph_cap *cap);
 692 static inline void ceph_remove_cap(struct ceph_cap *cap)
 693 {
 694         struct inode *inode = &cap->ci->vfs_inode;
 695         spin_lock(&inode->i_lock);
 696         __ceph_remove_cap(cap);
 697         spin_unlock(&inode->i_lock);
 698 }
 699 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
 700                          struct ceph_cap *cap);
 701
 702 extern void ceph_queue_caps_release(struct inode *inode);
 703 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
 704 extern int ceph_fsync(struct file *file, int datasync);
 705 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 706                                     struct ceph_mds_session *session);
 707 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
 708                                              int mds);
 709 extern int ceph_get_cap_mds(struct inode *inode);
 710 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 711 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 712 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 713                                        struct ceph_snap_context *snapc);
 714 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 715                                struct ceph_mds_session **psession,
 716                                int again);
 717 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 718                             struct ceph_mds_session *session);
 719 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 720 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 721
 722 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 723                                      int mds, int drop, int unless, int force);
 724 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 725                                       int mds, int drop, int unless);
 726
 727 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 728                          int *got, loff_t endoff);
 729
 730 /* for counting open files by mode */
 731 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 732 {
 733         ci->i_nr_by_mode[mode]++;
 734 }
 735 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 736
 737 /* addr.c */
 738 extern const struct address_space_operations ceph_aops;
 739 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 740
 741 /* file.c */
 742 extern const struct file_operations ceph_file_fops;
 743 extern const struct address_space_operations ceph_aops;
 744 extern int ceph_copy_to_page_vector(struct page **pages,
 745                                     const char *data,
 746                                     loff_t off, size_t len);
 747 extern int ceph_copy_from_page_vector(struct page **pages,
 748                                     char *data,
 749                                     loff_t off, size_t len);
 750 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
 751 extern int ceph_open(struct inode *inode, struct file *file);
 752 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 753                                        struct nameidata *nd, int mode,
 754                                        int locked_dir);
 755 extern int ceph_release(struct inode *inode, struct file *filp);
 756
 757 /* dir.c */
 758 extern const struct file_operations ceph_dir_fops;
 759 extern const struct inode_operations ceph_dir_iops;
 760 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 761         ceph_snapdir_dentry_ops;
 762
 763 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 764 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 765                                          struct dentry *dentry, int err);
 766
 767 extern void ceph_dentry_lru_add(struct dentry *dn);
 768 extern void ceph_dentry_lru_touch(struct dentry *dn);
 769 extern void ceph_dentry_lru_del(struct dentry *dn);
 770 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 771
 772 /*
 773  * our d_ops vary depending on whether the inode is live,
 774  * snapshotted (read-only), or a virtual ".snap" directory.
 775  */
 776 int ceph_init_dentry(struct dentry *dentry);
 777
 778
 779 /* ioctl.c */
 780 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 781
 782 /* export.c */
 783 extern const struct export_operations ceph_export_ops;
 784
 785 /* locks.c */
 786 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
 787 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
 788 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
 789 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
 790                              int p_locks, int f_locks);
 791 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 792
 793 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
 794 {
 795         if (dentry && dentry->d_parent)
 796                 return dentry->d_parent->d_inode;
 797
 798         return NULL;
 799 }
 800
 801 /* debugfs.c */
 802 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 803 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
 804
 805 #endif /* _FS_CEPH_SUPER_H */