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         struct inode *parent_inode;
 211 };
 212
 213 struct ceph_inode_xattrs_info {
 214         /*
 215          * (still encoded) xattr blob. we avoid the overhead of parsing
 216          * this until someone actually calls getxattr, etc.
 217          *
 218          * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 219          * NULL means we don't know.
 220         */
 221         struct ceph_buffer *blob, *prealloc_blob;
 222
 223         struct rb_root index;
 224         bool dirty;
 225         int count;
 226         int names_size;
 227         int vals_size;
 228         u64 version, index_version;
 229 };
 230
 231 /*
 232  * Ceph inode.
 233  */
 234 struct ceph_inode_info {
 235         struct ceph_vino i_vino;   /* ceph ino + snap */
 236
 237         u64 i_version;
 238         u32 i_time_warp_seq;
 239
 240         unsigned i_ceph_flags;
 241         unsigned long i_release_count;
 242
 243         struct ceph_dir_layout i_dir_layout;
 244         struct ceph_file_layout i_layout;
 245         char *i_symlink;
 246
 247         /* for dirs */
 248         struct timespec i_rctime;
 249         u64 i_rbytes, i_rfiles, i_rsubdirs;
 250         u64 i_files, i_subdirs;
 251         u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
 252
 253         struct rb_root i_fragtree;
 254         struct mutex i_fragtree_mutex;
 255
 256         struct ceph_inode_xattrs_info i_xattrs;
 257
 258         /* capabilities.  protected _both_ by i_lock and cap->session's
 259          * s_mutex. */
 260         struct rb_root i_caps;           /* cap list */
 261         struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 262         unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 263         struct list_head i_dirty_item, i_flushing_item;
 264         u64 i_cap_flush_seq;
 265         /* we need to track cap writeback on a per-cap-bit basis, to allow
 266          * overlapping, pipelined cap flushes to the mds.  we can probably
 267          * reduce the tid to 8 bits if we're concerned about inode size. */
 268         u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
 269         wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 270         unsigned long i_hold_caps_min; /* jiffies */
 271         unsigned long i_hold_caps_max; /* jiffies */
 272         struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 273         int i_cap_exporting_mds;         /* to handle cap migration between */
 274         unsigned i_cap_exporting_mseq;   /*  mds's. */
 275         unsigned i_cap_exporting_issued;
 276         struct ceph_cap_reservation i_cap_migration_resv;
 277         struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 278         struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 279                                                     dirty|flushing caps */
 280         unsigned i_snap_caps;           /* cap bits for snapped files */
 281
 282         int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 283
 284         u32 i_truncate_seq;        /* last truncate to smaller size */
 285         u64 i_truncate_size;       /*  and the size we last truncated down to */
 286         int i_truncate_pending;    /*  still need to call vmtruncate */
 287
 288         u64 i_max_size;            /* max file size authorized by mds */
 289         u64 i_reported_size; /* (max_)size reported to or requested of mds */
 290         u64 i_wanted_max_size;     /* offset we'd like to write too */
 291         u64 i_requested_max_size;  /* max_size we've requested */
 292
 293         /* held references to caps */
 294         int i_pin_ref;
 295         int i_rd_ref, i_rdcache_ref, i_wr_ref;
 296         int i_wrbuffer_ref, i_wrbuffer_ref_head;
 297         u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 298         u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 299         u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 300
 301         struct list_head i_unsafe_writes; /* uncommitted sync writes */
 302         struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 303         spinlock_t i_unsafe_lock;
 304
 305         struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 306         int i_snap_realm_counter; /* snap realm (if caps) */
 307         struct list_head i_snap_realm_item;
 308         struct list_head i_snap_flush_item;
 309
 310         struct work_struct i_wb_work;  /* writeback work */
 311         struct work_struct i_pg_inv_work;  /* page invalidation work */
 312
 313         struct work_struct i_vmtruncate_work;
 314
 315         struct inode vfs_inode; /* at end */
 316 };
 317
 318 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 319 {
 320         return container_of(inode, struct ceph_inode_info, vfs_inode);
 321 }
 322
 323 static inline struct ceph_vino ceph_vino(struct inode *inode)
 324 {
 325         return ceph_inode(inode)->i_vino;
 326 }
 327
 328 /*
 329  * ino_t is <64 bits on many architectures, blech.
 330  *
 331  * don't include snap in ino hash, at least for now.
 332  */
 333 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 334 {
 335         ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
 336 #if BITS_PER_LONG == 32
 337         ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
 338         if (!ino)
 339                 ino = 1;
 340 #endif
 341         return ino;
 342 }
 343
 344 /* for printf-style formatting */
 345 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 346
 347 static inline u64 ceph_ino(struct inode *inode)
 348 {
 349         return ceph_inode(inode)->i_vino.ino;
 350 }
 351 static inline u64 ceph_snap(struct inode *inode)
 352 {
 353         return ceph_inode(inode)->i_vino.snap;
 354 }
 355
 356 static inline int ceph_ino_compare(struct inode *inode, void *data)
 357 {
 358         struct ceph_vino *pvino = (struct ceph_vino *)data;
 359         struct ceph_inode_info *ci = ceph_inode(inode);
 360         return ci->i_vino.ino == pvino->ino &&
 361                 ci->i_vino.snap == pvino->snap;
 362 }
 363
 364 static inline struct inode *ceph_find_inode(struct super_block *sb,
 365                                             struct ceph_vino vino)
 366 {
 367         ino_t t = ceph_vino_to_ino(vino);
 368         return ilookup5(sb, t, ceph_ino_compare, &vino);
 369 }
 370
 371
 372 /*
 373  * Ceph inode.
 374  */
 375 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
 376 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 377 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 378 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
 379
 380 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 381 {
 382         struct ceph_inode_info *ci = ceph_inode(inode);
 383
 384         spin_lock(&inode->i_lock);
 385         ci->i_ceph_flags &= ~mask;
 386         spin_unlock(&inode->i_lock);
 387 }
 388
 389 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 390 {
 391         struct ceph_inode_info *ci = ceph_inode(inode);
 392
 393         spin_lock(&inode->i_lock);
 394         ci->i_ceph_flags |= mask;
 395         spin_unlock(&inode->i_lock);
 396 }
 397
 398 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
 399 {
 400         struct ceph_inode_info *ci = ceph_inode(inode);
 401         bool r;
 402
 403         spin_lock(&inode->i_lock);
 404         r = (ci->i_ceph_flags & mask) == mask;
 405         spin_unlock(&inode->i_lock);
 406         return r;
 407 }
 408
 409
 410 /* find a specific frag @f */
 411 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 412                                                 u32 f);
 413
 414 /*
 415  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 416  * exists
 417  */
 418 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 419                             struct ceph_inode_frag *pfrag,
 420                             int *found);
 421
 422 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 423 {
 424         return (struct ceph_dentry_info *)dentry->d_fsdata;
 425 }
 426
 427 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 428 {
 429         return ((loff_t)frag << 32) | (loff_t)off;
 430 }
 431
 432 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
 433 {
 434         ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
 435         inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
 436         return 0;
 437 }
 438
 439 /*
 440  * caps helpers
 441  */
 442 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 443 {
 444         return !RB_EMPTY_ROOT(&ci->i_caps);
 445 }
 446
 447 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 448 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 449 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 450                                     struct ceph_cap *cap);
 451
 452 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 453 {
 454         int issued;
 455         spin_lock(&ci->vfs_inode.i_lock);
 456         issued = __ceph_caps_issued(ci, NULL);
 457         spin_unlock(&ci->vfs_inode.i_lock);
 458         return issued;
 459 }
 460
 461 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 462                                         int touch)
 463 {
 464         int r;
 465         spin_lock(&ci->vfs_inode.i_lock);
 466         r = __ceph_caps_issued_mask(ci, mask, touch);
 467         spin_unlock(&ci->vfs_inode.i_lock);
 468         return r;
 469 }
 470
 471 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 472 {
 473         return ci->i_dirty_caps | ci->i_flushing_caps;
 474 }
 475 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 476
 477 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 478 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 479
 480 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 481
 482 /*
 483  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 484  */
 485 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 486 {
 487         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 488         if (w & CEPH_CAP_FILE_BUFFER)
 489                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 490         return w;
 491 }
 492
 493 /* what the mds thinks we want */
 494 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 495
 496 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 497 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 498 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
 499 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 500                              struct ceph_cap_reservation *ctx, int need);
 501 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 502                                struct ceph_cap_reservation *ctx);
 503 extern void ceph_reservation_status(struct ceph_fs_client *client,
 504                                     int *total, int *avail, int *used,
 505                                     int *reserved, int *min);
 506
 507 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
 508 {
 509         return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 510 }
 511
 512 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
 513 {
 514         return (struct ceph_fs_client *)sb->s_fs_info;
 515 }
 516
 517
 518 /*
 519  * we keep buffered readdir results attached to file->private_data
 520  */
 521 struct ceph_file_info {
 522         int fmode;     /* initialized on open */
 523
 524         /* readdir: position within the dir */
 525         u32 frag;
 526         struct ceph_mds_request *last_readdir;
 527         int at_end;
 528
 529         /* readdir: position within a frag */
 530         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 531         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 532         char *last_name;       /* last entry in previous chunk */
 533         struct dentry *dentry; /* next dentry (for dcache readdir) */
 534         unsigned long dir_release_count;
 535
 536         /* used for -o dirstat read() on directory thing */
 537         char *dir_info;
 538         int dir_info_len;
 539 };
 540
 541
 542
 543 /*
 544  * A "snap realm" describes a subset of the file hierarchy sharing
 545  * the same set of snapshots that apply to it.  The realms themselves
 546  * are organized into a hierarchy, such that children inherit (some of)
 547  * the snapshots of their parents.
 548  *
 549  * All inodes within the realm that have capabilities are linked into a
 550  * per-realm list.
 551  */
 552 struct ceph_snap_realm {
 553         u64 ino;
 554         atomic_t nref;
 555         struct rb_node node;
 556
 557         u64 created, seq;
 558         u64 parent_ino;
 559         u64 parent_since;   /* snapid when our current parent became so */
 560
 561         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 562         int num_prior_parent_snaps;   /*  had prior to parent_since */
 563         u64 *snaps;                   /* snaps specific to this realm */
 564         int num_snaps;
 565
 566         struct ceph_snap_realm *parent;
 567         struct list_head children;       /* list of child realms */
 568         struct list_head child_item;
 569
 570         struct list_head empty_item;     /* if i have ref==0 */
 571
 572         struct list_head dirty_item;     /* if realm needs new context */
 573
 574         /* the current set of snaps for this realm */
 575         struct ceph_snap_context *cached_context;
 576
 577         struct list_head inodes_with_caps;
 578         spinlock_t inodes_with_caps_lock;
 579 };
 580
 581 static inline int default_congestion_kb(void)
 582 {
 583         int congestion_kb;
 584
 585         /*
 586          * Copied from NFS
 587          *
 588          * congestion size, scale with available memory.
 589          *
 590          *  64MB:    8192k
 591          * 128MB:   11585k
 592          * 256MB:   16384k
 593          * 512MB:   23170k
 594          *   1GB:   32768k
 595          *   2GB:   46340k
 596          *   4GB:   65536k
 597          *   8GB:   92681k
 598          *  16GB:  131072k
 599          *
 600          * This allows larger machines to have larger/more transfers.
 601          * Limit the default to 256M
 602          */
 603         congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
 604         if (congestion_kb > 256*1024)
 605                 congestion_kb = 256*1024;
 606
 607         return congestion_kb;
 608 }
 609
 610
 611
 612 /* snap.c */
 613 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 614                                                u64 ino);
 615 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 616                                 struct ceph_snap_realm *realm);
 617 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 618                                 struct ceph_snap_realm *realm);
 619 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 620                                   void *p, void *e, bool deletion);
 621 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 622                              struct ceph_mds_session *session,
 623                              struct ceph_msg *msg);
 624 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 625 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 626                                   struct ceph_cap_snap *capsnap);
 627 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 628
 629 /*
 630  * a cap_snap is "pending" if it is still awaiting an in-progress
 631  * sync write (that may/may not still update size, mtime, etc.).
 632  */
 633 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 634 {
 635         return !list_empty(&ci->i_cap_snaps) &&
 636                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 637                            ci_item)->writing;
 638 }
 639
 640 /* inode.c */
 641 extern const struct inode_operations ceph_file_iops;
 642
 643 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 644 extern void ceph_destroy_inode(struct inode *inode);
 645
 646 extern struct inode *ceph_get_inode(struct super_block *sb,
 647                                     struct ceph_vino vino);
 648 extern struct inode *ceph_get_snapdir(struct inode *parent);
 649 extern int ceph_fill_file_size(struct inode *inode, int issued,
 650                                u32 truncate_seq, u64 truncate_size, u64 size);
 651 extern void ceph_fill_file_time(struct inode *inode, int issued,
 652                                 u64 time_warp_seq, struct timespec *ctime,
 653                                 struct timespec *mtime, struct timespec *atime);
 654 extern int ceph_fill_trace(struct super_block *sb,
 655                            struct ceph_mds_request *req,
 656                            struct ceph_mds_session *session);
 657 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 658                                     struct ceph_mds_session *session);
 659
 660 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 661
 662 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 663 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 664 extern void ceph_queue_vmtruncate(struct inode *inode);
 665
 666 extern void ceph_queue_invalidate(struct inode *inode);
 667 extern void ceph_queue_writeback(struct inode *inode);
 668
 669 extern int ceph_do_getattr(struct inode *inode, int mask);
 670 extern int ceph_permission(struct inode *inode, int mask, unsigned int flags);
 671 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 672 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 673                         struct kstat *stat);
 674
 675 /* xattr.c */
 676 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 677                          size_t, int);
 678 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 679 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 680 extern int ceph_removexattr(struct dentry *, const char *);
 681 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 682 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 683
 684 /* caps.c */
 685 extern const char *ceph_cap_string(int c);
 686 extern void ceph_handle_caps(struct ceph_mds_session *session,
 687                              struct ceph_msg *msg);
 688 extern int ceph_add_cap(struct inode *inode,
 689                         struct ceph_mds_session *session, u64 cap_id,
 690                         int fmode, unsigned issued, unsigned wanted,
 691                         unsigned cap, unsigned seq, u64 realmino, int flags,
 692                         struct ceph_cap_reservation *caps_reservation);
 693 extern void __ceph_remove_cap(struct ceph_cap *cap);
 694 static inline void ceph_remove_cap(struct ceph_cap *cap)
 695 {
 696         struct inode *inode = &cap->ci->vfs_inode;
 697         spin_lock(&inode->i_lock);
 698         __ceph_remove_cap(cap);
 699         spin_unlock(&inode->i_lock);
 700 }
 701 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
 702                          struct ceph_cap *cap);
 703
 704 extern void ceph_queue_caps_release(struct inode *inode);
 705 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
 706 extern int ceph_fsync(struct file *file, int datasync);
 707 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 708                                     struct ceph_mds_session *session);
 709 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
 710                                              int mds);
 711 extern int ceph_get_cap_mds(struct inode *inode);
 712 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 713 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 714 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 715                                        struct ceph_snap_context *snapc);
 716 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 717                                struct ceph_mds_session **psession,
 718                                int again);
 719 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 720                             struct ceph_mds_session *session);
 721 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 722 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 723
 724 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 725                                      int mds, int drop, int unless, int force);
 726 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 727                                       int mds, int drop, int unless);
 728
 729 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 730                          int *got, loff_t endoff);
 731
 732 /* for counting open files by mode */
 733 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 734 {
 735         ci->i_nr_by_mode[mode]++;
 736 }
 737 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 738
 739 /* addr.c */
 740 extern const struct address_space_operations ceph_aops;
 741 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 742
 743 /* file.c */
 744 extern const struct file_operations ceph_file_fops;
 745 extern const struct address_space_operations ceph_aops;
 746 extern int ceph_copy_to_page_vector(struct page **pages,
 747                                     const char *data,
 748                                     loff_t off, size_t len);
 749 extern int ceph_copy_from_page_vector(struct page **pages,
 750                                     char *data,
 751                                     loff_t off, size_t len);
 752 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
 753 extern int ceph_open(struct inode *inode, struct file *file);
 754 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 755                                        struct nameidata *nd, int mode,
 756                                        int locked_dir);
 757 extern int ceph_release(struct inode *inode, struct file *filp);
 758
 759 /* dir.c */
 760 extern const struct file_operations ceph_dir_fops;
 761 extern const struct inode_operations ceph_dir_iops;
 762 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 763         ceph_snapdir_dentry_ops;
 764
 765 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 766 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 767                                          struct dentry *dentry, int err);
 768
 769 extern void ceph_dentry_lru_add(struct dentry *dn);
 770 extern void ceph_dentry_lru_touch(struct dentry *dn);
 771 extern void ceph_dentry_lru_del(struct dentry *dn);
 772 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 773 extern unsigned ceph_dentry_hash(struct dentry *dn);
 774
 775 /*
 776  * our d_ops vary depending on whether the inode is live,
 777  * snapshotted (read-only), or a virtual ".snap" directory.
 778  */
 779 int ceph_init_dentry(struct dentry *dentry);
 780
 781
 782 /* ioctl.c */
 783 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 784
 785 /* export.c */
 786 extern const struct export_operations ceph_export_ops;
 787
 788 /* locks.c */
 789 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
 790 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
 791 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
 792 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
 793                              int p_locks, int f_locks);
 794 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 795
 796 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
 797 {
 798         if (dentry && dentry->d_parent)
 799                 return dentry->d_parent->d_inode;
 800
 801         return NULL;
 802 }
 803
 804 /* debugfs.c */
 805 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 806 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
 807
 808 #endif /* _FS_CEPH_SUPER_H */