fs/ceph/super.h

   1 #ifndef _FS_CEPH_SUPER_H
   2 #define _FS_CEPH_SUPER_H
   3
   4 #include "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
  15 #include "types.h"
  16 #include "messenger.h"
  17 #include "msgpool.h"
  18 #include "mon_client.h"
  19 #include "mds_client.h"
  20 #include "osd_client.h"
  21 #include "ceph_fs.h"
  22
  23 /* f_type in struct statfs */
  24 #define CEPH_SUPER_MAGIC 0x00c36400
  25
  26 /* large granularity for statfs utilization stats to facilitate
  27  * large volume sizes on 32-bit machines. */
  28 #define CEPH_BLOCK_SHIFT   20  /* 1 MB */
  29 #define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
  30
  31 /*
  32  * mount options
  33  */
  34 #define CEPH_OPT_FSID             (1<<0)
  35 #define CEPH_OPT_NOSHARE          (1<<1) /* don't share client with other sbs */
  36 #define CEPH_OPT_MYIP             (1<<2) /* specified my ip */
  37 #define CEPH_OPT_DIRSTAT          (1<<4) /* funky `cat dirname` for stats */
  38 #define CEPH_OPT_RBYTES           (1<<5) /* dir st_bytes = rbytes */
  39 #define CEPH_OPT_NOCRC            (1<<6) /* no data crc on writes */
  40 #define CEPH_OPT_NOASYNCREADDIR   (1<<7) /* no dcache readdir */
  41
  42 #define CEPH_OPT_DEFAULT   (CEPH_OPT_RBYTES)
  43
  44 #define ceph_set_opt(client, opt) \
  45         (client)->mount_args->flags |= CEPH_OPT_##opt;
  46 #define ceph_test_opt(client, opt) \
  47         (!!((client)->mount_args->flags & CEPH_OPT_##opt))
  48
  49
  50 struct ceph_mount_args {
  51         int sb_flags;
  52         int num_mon;
  53         struct ceph_entity_addr *mon_addr;
  54         int flags;
  55         int mount_timeout;
  56         int caps_wanted_delay_min, caps_wanted_delay_max;
  57         struct ceph_fsid fsid;
  58         struct ceph_entity_addr my_addr;
  59         int wsize;
  60         int rsize;            /* max readahead */
  61         int max_readdir;      /* max readdir size */
  62         int osd_timeout;
  63         char *snapdir_name;   /* default ".snap" */
  64         char *secret;
  65         int cap_release_safety;
  66 };
  67
  68 /*
  69  * defaults
  70  */
  71 #define CEPH_MOUNT_TIMEOUT_DEFAULT  60
  72 #define CEPH_MOUNT_RSIZE_DEFAULT    (512*1024) /* readahead */
  73
  74 #define CEPH_MSG_MAX_FRONT_LEN  (16*1024*1024)
  75 #define CEPH_MSG_MAX_DATA_LEN   (16*1024*1024)
  76
  77 #define CEPH_SNAPDIRNAME_DEFAULT ".snap"
  78
  79 /*
  80  * Delay telling the MDS we no longer want caps, in case we reopen
  81  * the file.  Delay a minimum amount of time, even if we send a cap
  82  * message for some other reason.  Otherwise, take the oppotunity to
  83  * update the mds to avoid sending another message later.
  84  */
  85 #define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
  86 #define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
  87
  88
  89 /* mount state */
  90 enum {
  91         CEPH_MOUNT_MOUNTING,
  92         CEPH_MOUNT_MOUNTED,
  93         CEPH_MOUNT_UNMOUNTING,
  94         CEPH_MOUNT_UNMOUNTED,
  95         CEPH_MOUNT_SHUTDOWN,
  96 };
  97
  98 /*
  99  * subtract jiffies
 100  */
 101 static inline unsigned long time_sub(unsigned long a, unsigned long b)
 102 {
 103         BUG_ON(time_after(b, a));
 104         return (long)a - (long)b;
 105 }
 106
 107 /*
 108  * per-filesystem client state
 109  *
 110  * possibly shared by multiple mount points, if they are
 111  * mounting the same ceph filesystem/cluster.
 112  */
 113 struct ceph_client {
 114         __s64 whoami;                   /* my client number */
 115 #ifdef CONFIG_DEBUG_FS
 116         struct dentry *debugfs_monmap;
 117         struct dentry *debugfs_mdsmap, *debugfs_osdmap;
 118         struct dentry *debugfs_dir, *debugfs_dentry_lru, *debugfs_caps;
 119 #endif
 120
 121         struct mutex mount_mutex;       /* serialize mount attempts */
 122         struct ceph_mount_args *mount_args;
 123         struct ceph_fsid fsid;
 124
 125         struct super_block *sb;
 126
 127         unsigned long mount_state;
 128         wait_queue_head_t mount_wq;
 129
 130         int mount_err;
 131         void *signed_ticket;           /* our keys to the kingdom */
 132         int signed_ticket_len;
 133
 134         struct ceph_messenger *msgr;   /* messenger instance */
 135         struct ceph_mon_client monc;
 136         struct ceph_mds_client mdsc;
 137         struct ceph_osd_client osdc;
 138
 139         /* writeback */
 140         mempool_t *wb_pagevec_pool;
 141         struct workqueue_struct *wb_wq;
 142         struct workqueue_struct *pg_inv_wq;
 143         struct workqueue_struct *trunc_wq;
 144
 145         struct backing_dev_info backing_dev_info;
 146 };
 147
 148 static inline struct ceph_client *ceph_client(struct super_block *sb)
 149 {
 150         return sb->s_fs_info;
 151 }
 152
 153
 154 /*
 155  * File i/o capability.  This tracks shared state with the metadata
 156  * server that allows us to cache or writeback attributes or to read
 157  * and write data.  For any given inode, we should have one or more
 158  * capabilities, one issued by each metadata server, and our
 159  * cumulative access is the OR of all issued capabilities.
 160  *
 161  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
 162  * session capability lists.
 163  */
 164 struct ceph_cap {
 165         struct ceph_inode_info *ci;
 166         struct rb_node ci_node;          /* per-ci cap tree */
 167         struct ceph_mds_session *session;
 168         struct list_head session_caps;   /* per-session caplist */
 169         int mds;
 170         u64 cap_id;       /* unique cap id (mds provided) */
 171         int issued;       /* latest, from the mds */
 172         int implemented;  /* implemented superset of issued (for revocation) */
 173         int mds_wanted;
 174         u32 seq, issue_seq, mseq;
 175         u32 cap_gen;      /* active/stale cycle */
 176         unsigned long last_used;
 177         struct list_head caps_item;
 178 };
 179
 180 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
 181 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
 182 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
 183
 184 /*
 185  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 186  * we first complete any in-process sync writes and writeback any dirty
 187  * data before flushing the snapped state (tracked here) back to the MDS.
 188  */
 189 struct ceph_cap_snap {
 190         atomic_t nref;
 191         struct ceph_inode_info *ci;
 192         struct list_head ci_item, flushing_item;
 193
 194         u64 follows, flush_tid;
 195         int issued, dirty;
 196         struct ceph_snap_context *context;
 197
 198         mode_t mode;
 199         uid_t uid;
 200         gid_t gid;
 201
 202         void *xattr_blob;
 203         int xattr_len;
 204         u64 xattr_version;
 205
 206         u64 size;
 207         struct timespec mtime, atime, ctime;
 208         u64 time_warp_seq;
 209         int writing;   /* a sync write is still in progress */
 210         int dirty_pages;     /* dirty pages awaiting writeback */
 211 };
 212
 213 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 214 {
 215         if (atomic_dec_and_test(&capsnap->nref))
 216                 kfree(capsnap);
 217 }
 218
 219 /*
 220  * The frag tree describes how a directory is fragmented, potentially across
 221  * multiple metadata servers.  It is also used to indicate points where
 222  * metadata authority is delegated, and whether/where metadata is replicated.
 223  *
 224  * A _leaf_ frag will be present in the i_fragtree IFF there is
 225  * delegation info.  That is, if mds >= 0 || ndist > 0.
 226  */
 227 #define CEPH_MAX_DIRFRAG_REP 4
 228
 229 struct ceph_inode_frag {
 230         struct rb_node node;
 231
 232         /* fragtree state */
 233         u32 frag;
 234         int split_by;         /* i.e. 2^(split_by) children */
 235
 236         /* delegation and replication info */
 237         int mds;              /* -1 if same authority as parent */
 238         int ndist;            /* >0 if replicated */
 239         int dist[CEPH_MAX_DIRFRAG_REP];
 240 };
 241
 242 /*
 243  * We cache inode xattrs as an encoded blob until they are first used,
 244  * at which point we parse them into an rbtree.
 245  */
 246 struct ceph_inode_xattr {
 247         struct rb_node node;
 248
 249         const char *name;
 250         int name_len;
 251         const char *val;
 252         int val_len;
 253         int dirty;
 254
 255         int should_free_name;
 256         int should_free_val;
 257 };
 258
 259 struct ceph_inode_xattrs_info {
 260         /*
 261          * (still encoded) xattr blob. we avoid the overhead of parsing
 262          * this until someone actually calls getxattr, etc.
 263          *
 264          * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 265          * NULL means we don't know.
 266         */
 267         struct ceph_buffer *blob, *prealloc_blob;
 268
 269         struct rb_root index;
 270         bool dirty;
 271         int count;
 272         int names_size;
 273         int vals_size;
 274         u64 version, index_version;
 275 };
 276
 277 /*
 278  * Ceph inode.
 279  */
 280 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
 281 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 282 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 283
 284 struct ceph_inode_info {
 285         struct ceph_vino i_vino;   /* ceph ino + snap */
 286
 287         u64 i_version;
 288         u32 i_time_warp_seq;
 289
 290         unsigned i_ceph_flags;
 291         unsigned long i_release_count;
 292
 293         struct ceph_file_layout i_layout;
 294         char *i_symlink;
 295
 296         /* for dirs */
 297         struct timespec i_rctime;
 298         u64 i_rbytes, i_rfiles, i_rsubdirs;
 299         u64 i_files, i_subdirs;
 300         u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
 301
 302         struct rb_root i_fragtree;
 303         struct mutex i_fragtree_mutex;
 304
 305         struct ceph_inode_xattrs_info i_xattrs;
 306
 307         /* capabilities.  protected _both_ by i_lock and cap->session's
 308          * s_mutex. */
 309         struct rb_root i_caps;           /* cap list */
 310         struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 311         unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 312         struct list_head i_dirty_item, i_flushing_item;
 313         u64 i_cap_flush_seq;
 314         /* we need to track cap writeback on a per-cap-bit basis, to allow
 315          * overlapping, pipelined cap flushes to the mds.  we can probably
 316          * reduce the tid to 8 bits if we're concerned about inode size. */
 317         u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
 318         wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 319         unsigned long i_hold_caps_min; /* jiffies */
 320         unsigned long i_hold_caps_max; /* jiffies */
 321         struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 322         int i_cap_exporting_mds;         /* to handle cap migration between */
 323         unsigned i_cap_exporting_mseq;   /*  mds's. */
 324         unsigned i_cap_exporting_issued;
 325         struct ceph_cap_reservation i_cap_migration_resv;
 326         struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 327         struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 */
 328         unsigned i_snap_caps;           /* cap bits for snapped files */
 329
 330         int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 331
 332         u32 i_truncate_seq;        /* last truncate to smaller size */
 333         u64 i_truncate_size;       /*  and the size we last truncated down to */
 334         int i_truncate_pending;    /*  still need to call vmtruncate */
 335
 336         u64 i_max_size;            /* max file size authorized by mds */
 337         u64 i_reported_size; /* (max_)size reported to or requested of mds */
 338         u64 i_wanted_max_size;     /* offset we'd like to write too */
 339         u64 i_requested_max_size;  /* max_size we've requested */
 340
 341         /* held references to caps */
 342         int i_pin_ref;
 343         int i_rd_ref, i_rdcache_ref, i_wr_ref;
 344         int i_wrbuffer_ref, i_wrbuffer_ref_head;
 345         u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 346         u32 i_rdcache_gen;      /* we increment this each time we get
 347                                    FILE_CACHE.  If it's non-zero, we
 348                                    _may_ have cached pages. */
 349         u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 350
 351         struct list_head i_unsafe_writes; /* uncommitted sync writes */
 352         struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 353         spinlock_t i_unsafe_lock;
 354
 355         struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 356         int i_snap_realm_counter; /* snap realm (if caps) */
 357         struct list_head i_snap_realm_item;
 358         struct list_head i_snap_flush_item;
 359
 360         struct work_struct i_wb_work;  /* writeback work */
 361         struct work_struct i_pg_inv_work;  /* page invalidation work */
 362
 363         struct work_struct i_vmtruncate_work;
 364
 365         struct inode vfs_inode; /* at end */
 366 };
 367
 368 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 369 {
 370         return container_of(inode, struct ceph_inode_info, vfs_inode);
 371 }
 372
 373 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 374 {
 375         struct ceph_inode_info *ci = ceph_inode(inode);
 376
 377         spin_lock(&inode->i_lock);
 378         ci->i_ceph_flags &= ~mask;
 379         spin_unlock(&inode->i_lock);
 380 }
 381
 382 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 383 {
 384         struct ceph_inode_info *ci = ceph_inode(inode);
 385
 386         spin_lock(&inode->i_lock);
 387         ci->i_ceph_flags |= mask;
 388         spin_unlock(&inode->i_lock);
 389 }
 390
 391 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
 392 {
 393         struct ceph_inode_info *ci = ceph_inode(inode);
 394         bool r;
 395
 396         smp_mb();
 397         r = (ci->i_ceph_flags & mask) == mask;
 398         return r;
 399 }
 400
 401
 402 /* find a specific frag @f */
 403 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 404                                                 u32 f);
 405
 406 /*
 407  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 408  * exists
 409  */
 410 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 411                             struct ceph_inode_frag *pfrag,
 412                             int *found);
 413
 414 /*
 415  * Ceph dentry state
 416  */
 417 struct ceph_dentry_info {
 418         struct ceph_mds_session *lease_session;
 419         u32 lease_gen, lease_shared_gen;
 420         u32 lease_seq;
 421         unsigned long lease_renew_after, lease_renew_from;
 422         struct list_head lru;
 423         struct dentry *dentry;
 424         u64 time;
 425         u64 offset;
 426 };
 427
 428 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 429 {
 430         return (struct ceph_dentry_info *)dentry->d_fsdata;
 431 }
 432
 433 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 434 {
 435         return ((loff_t)frag << 32) | (loff_t)off;
 436 }
 437
 438 /*
 439  * ino_t is <64 bits on many architectures, blech.
 440  *
 441  * don't include snap in ino hash, at least for now.
 442  */
 443 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 444 {
 445         ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
 446 #if BITS_PER_LONG == 32
 447         ino ^= vino.ino >> (sizeof(u64)-sizeof(ino_t)) * 8;
 448         if (!ino)
 449                 ino = 1;
 450 #endif
 451         return ino;
 452 }
 453
 454 static inline int ceph_set_ino_cb(struct inode *inode, void *data)
 455 {
 456         ceph_inode(inode)->i_vino = *(struct ceph_vino *)data;
 457         inode->i_ino = ceph_vino_to_ino(*(struct ceph_vino *)data);
 458         return 0;
 459 }
 460
 461 static inline struct ceph_vino ceph_vino(struct inode *inode)
 462 {
 463         return ceph_inode(inode)->i_vino;
 464 }
 465
 466 /* for printf-style formatting */
 467 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 468
 469 static inline u64 ceph_ino(struct inode *inode)
 470 {
 471         return ceph_inode(inode)->i_vino.ino;
 472 }
 473 static inline u64 ceph_snap(struct inode *inode)
 474 {
 475         return ceph_inode(inode)->i_vino.snap;
 476 }
 477
 478 static inline int ceph_ino_compare(struct inode *inode, void *data)
 479 {
 480         struct ceph_vino *pvino = (struct ceph_vino *)data;
 481         struct ceph_inode_info *ci = ceph_inode(inode);
 482         return ci->i_vino.ino == pvino->ino &&
 483                 ci->i_vino.snap == pvino->snap;
 484 }
 485
 486 static inline struct inode *ceph_find_inode(struct super_block *sb,
 487                                             struct ceph_vino vino)
 488 {
 489         ino_t t = ceph_vino_to_ino(vino);
 490         return ilookup5(sb, t, ceph_ino_compare, &vino);
 491 }
 492
 493
 494 /*
 495  * caps helpers
 496  */
 497 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 498 {
 499         return !RB_EMPTY_ROOT(&ci->i_caps);
 500 }
 501
 502 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 503 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 504 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 505                                     struct ceph_cap *cap);
 506
 507 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 508 {
 509         int issued;
 510         spin_lock(&ci->vfs_inode.i_lock);
 511         issued = __ceph_caps_issued(ci, NULL);
 512         spin_unlock(&ci->vfs_inode.i_lock);
 513         return issued;
 514 }
 515
 516 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 517                                         int touch)
 518 {
 519         int r;
 520         spin_lock(&ci->vfs_inode.i_lock);
 521         r = __ceph_caps_issued_mask(ci, mask, touch);
 522         spin_unlock(&ci->vfs_inode.i_lock);
 523         return r;
 524 }
 525
 526 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 527 {
 528         return ci->i_dirty_caps | ci->i_flushing_caps;
 529 }
 530 extern void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 531
 532 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 533 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 534
 535 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 536
 537 /*
 538  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 539  */
 540 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 541 {
 542         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 543         if (w & CEPH_CAP_FILE_BUFFER)
 544                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 545         return w;
 546 }
 547
 548 /* what the mds thinks we want */
 549 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 550
 551 extern void ceph_caps_init(void);
 552 extern void ceph_caps_finalize(void);
 553 extern int ceph_reserve_caps(struct ceph_cap_reservation *ctx, int need);
 554 extern int ceph_unreserve_caps(struct ceph_cap_reservation *ctx);
 555 extern void ceph_reservation_status(struct ceph_client *client,
 556                                     int *total, int *avail, int *used,
 557                                     int *reserved);
 558
 559 static inline struct ceph_client *ceph_inode_to_client(struct inode *inode)
 560 {
 561         return (struct ceph_client *)inode->i_sb->s_fs_info;
 562 }
 563
 564 static inline struct ceph_client *ceph_sb_to_client(struct super_block *sb)
 565 {
 566         return (struct ceph_client *)sb->s_fs_info;
 567 }
 568
 569 static inline int ceph_queue_writeback(struct inode *inode)
 570 {
 571         return queue_work(ceph_inode_to_client(inode)->wb_wq,
 572                    &ceph_inode(inode)->i_wb_work);
 573 }
 574
 575 static inline int ceph_queue_page_invalidation(struct inode *inode)
 576 {
 577         return queue_work(ceph_inode_to_client(inode)->pg_inv_wq,
 578                    &ceph_inode(inode)->i_pg_inv_work);
 579 }
 580
 581
 582 /*
 583  * we keep buffered readdir results attached to file->private_data
 584  */
 585 struct ceph_file_info {
 586         int fmode;     /* initialized on open */
 587
 588         /* readdir: position within the dir */
 589         u32 frag;
 590         struct ceph_mds_request *last_readdir;
 591         int at_end;
 592
 593         /* readdir: position within a frag */
 594         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 595         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 596         char *last_name;       /* last entry in previous chunk */
 597         struct dentry *dentry; /* next dentry (for dcache readdir) */
 598         unsigned long dir_release_count;
 599
 600         /* used for -o dirstat read() on directory thing */
 601         char *dir_info;
 602         int dir_info_len;
 603 };
 604
 605
 606
 607 /*
 608  * snapshots
 609  */
 610
 611 /*
 612  * A "snap context" is the set of existing snapshots when we
 613  * write data.  It is used by the OSD to guide its COW behavior.
 614  *
 615  * The ceph_snap_context is refcounted, and attached to each dirty
 616  * page, indicating which context the dirty data belonged when it was
 617  * dirtied.
 618  */
 619 struct ceph_snap_context {
 620         atomic_t nref;
 621         u64 seq;
 622         int num_snaps;
 623         u64 snaps[];
 624 };
 625
 626 static inline struct ceph_snap_context *
 627 ceph_get_snap_context(struct ceph_snap_context *sc)
 628 {
 629         /*
 630         printk("get_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
 631                atomic_read(&sc->nref)+1);
 632         */
 633         if (sc)
 634                 atomic_inc(&sc->nref);
 635         return sc;
 636 }
 637
 638 static inline void ceph_put_snap_context(struct ceph_snap_context *sc)
 639 {
 640         if (!sc)
 641                 return;
 642         /*
 643         printk("put_snap_context %p %d -> %d\n", sc, atomic_read(&sc->nref),
 644                atomic_read(&sc->nref)-1);
 645         */
 646         if (atomic_dec_and_test(&sc->nref)) {
 647                 /*printk(" deleting snap_context %p\n", sc);*/
 648                 kfree(sc);
 649         }
 650 }
 651
 652 /*
 653  * A "snap realm" describes a subset of the file hierarchy sharing
 654  * the same set of snapshots that apply to it.  The realms themselves
 655  * are organized into a hierarchy, such that children inherit (some of)
 656  * the snapshots of their parents.
 657  *
 658  * All inodes within the realm that have capabilities are linked into a
 659  * per-realm list.
 660  */
 661 struct ceph_snap_realm {
 662         u64 ino;
 663         atomic_t nref;
 664         u64 created, seq;
 665         u64 parent_ino;
 666         u64 parent_since;   /* snapid when our current parent became so */
 667
 668         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 669         int num_prior_parent_snaps;   /*  had prior to parent_since */
 670         u64 *snaps;                   /* snaps specific to this realm */
 671         int num_snaps;
 672
 673         struct ceph_snap_realm *parent;
 674         struct list_head children;       /* list of child realms */
 675         struct list_head child_item;
 676
 677         struct list_head empty_item;     /* if i have ref==0 */
 678
 679         /* the current set of snaps for this realm */
 680         struct ceph_snap_context *cached_context;
 681
 682         struct list_head inodes_with_caps;
 683         spinlock_t inodes_with_caps_lock;
 684 };
 685
 686
 687
 688 /*
 689  * calculate the number of pages a given length and offset map onto,
 690  * if we align the data.
 691  */
 692 static inline int calc_pages_for(u64 off, u64 len)
 693 {
 694         return ((off+len+PAGE_CACHE_SIZE-1) >> PAGE_CACHE_SHIFT) -
 695                 (off >> PAGE_CACHE_SHIFT);
 696 }
 697
 698
 699
 700 /* snap.c */
 701 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 702                                                u64 ino);
 703 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 704                                 struct ceph_snap_realm *realm);
 705 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 706                                 struct ceph_snap_realm *realm);
 707 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 708                                   void *p, void *e, bool deletion);
 709 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 710                              struct ceph_msg *msg);
 711 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci,
 712                                 struct ceph_snap_context *snapc);
 713 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 714                                   struct ceph_cap_snap *capsnap);
 715 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 716
 717 /*
 718  * a cap_snap is "pending" if it is still awaiting an in-progress
 719  * sync write (that may/may not still update size, mtime, etc.).
 720  */
 721 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 722 {
 723         return !list_empty(&ci->i_cap_snaps) &&
 724                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 725                            ci_item)->writing;
 726 }
 727
 728
 729 /* super.c */
 730 extern struct kmem_cache *ceph_inode_cachep;
 731 extern struct kmem_cache *ceph_cap_cachep;
 732 extern struct kmem_cache *ceph_dentry_cachep;
 733 extern struct kmem_cache *ceph_file_cachep;
 734
 735 extern const char *ceph_msg_type_name(int type);
 736
 737 #define FSID_FORMAT "%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x-" \
 738         "%02x%02x%02x%02x%02x%02x"
 739 #define PR_FSID(f) (f)->fsid[0], (f)->fsid[1], (f)->fsid[2], (f)->fsid[3], \
 740                 (f)->fsid[4], (f)->fsid[5], (f)->fsid[6], (f)->fsid[7],    \
 741                 (f)->fsid[8], (f)->fsid[9], (f)->fsid[10], (f)->fsid[11],  \
 742                 (f)->fsid[12], (f)->fsid[13], (f)->fsid[14], (f)->fsid[15]
 743
 744 /* inode.c */
 745 extern const struct inode_operations ceph_file_iops;
 746
 747 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 748 extern void ceph_destroy_inode(struct inode *inode);
 749
 750 extern struct inode *ceph_get_inode(struct super_block *sb,
 751                                     struct ceph_vino vino);
 752 extern struct inode *ceph_get_snapdir(struct inode *parent);
 753 extern int ceph_fill_file_size(struct inode *inode, int issued,
 754                                u32 truncate_seq, u64 truncate_size, u64 size);
 755 extern void ceph_fill_file_time(struct inode *inode, int issued,
 756                                 u64 time_warp_seq, struct timespec *ctime,
 757                                 struct timespec *mtime, struct timespec *atime);
 758 extern int ceph_fill_trace(struct super_block *sb,
 759                            struct ceph_mds_request *req,
 760                            struct ceph_mds_session *session);
 761 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 762                                     struct ceph_mds_session *session);
 763
 764 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 765
 766 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 767 extern void ceph_inode_writeback(struct work_struct *work);
 768 extern void ceph_vmtruncate_work(struct work_struct *work);
 769 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 770 extern void __ceph_queue_vmtruncate(struct inode *inode);
 771
 772 extern int ceph_do_getattr(struct inode *inode, int mask);
 773 extern int ceph_permission(struct inode *inode, int mask);
 774 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 775 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 776                         struct kstat *stat);
 777
 778 /* xattr.c */
 779 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 780                          size_t, int);
 781 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 782 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 783 extern int ceph_removexattr(struct dentry *, const char *);
 784 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 785 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 786
 787 /* caps.c */
 788 extern const char *ceph_cap_string(int c);
 789 extern void ceph_handle_caps(struct ceph_mds_session *session,
 790                              struct ceph_msg *msg);
 791 extern int ceph_add_cap(struct inode *inode,
 792                         struct ceph_mds_session *session, u64 cap_id,
 793                         int fmode, unsigned issued, unsigned wanted,
 794                         unsigned cap, unsigned seq, u64 realmino, int flags,
 795                         struct ceph_cap_reservation *caps_reservation);
 796 extern void __ceph_remove_cap(struct ceph_cap *cap,
 797                               struct ceph_cap_reservation *ctx);
 798 static inline void ceph_remove_cap(struct ceph_cap *cap)
 799 {
 800         struct inode *inode = &cap->ci->vfs_inode;
 801         spin_lock(&inode->i_lock);
 802         __ceph_remove_cap(cap, NULL);
 803         spin_unlock(&inode->i_lock);
 804 }
 805
 806 extern void ceph_queue_caps_release(struct inode *inode);
 807 extern int ceph_write_inode(struct inode *inode, int unused);
 808 extern int ceph_fsync(struct file *file, struct dentry *dentry, int datasync);
 809 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 810                                     struct ceph_mds_session *session);
 811 extern int ceph_get_cap_mds(struct inode *inode);
 812 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 813 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 814 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 815                                        struct ceph_snap_context *snapc);
 816 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 817                                struct ceph_mds_session **psession);
 818 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 819                             struct ceph_mds_session *session);
 820 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 821 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 822
 823 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 824                                      int mds, int drop, int unless, int force);
 825 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 826                                       int mds, int drop, int unless);
 827
 828 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 829                          int *got, loff_t endoff);
 830
 831 /* for counting open files by mode */
 832 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 833 {
 834         ci->i_nr_by_mode[mode]++;
 835 }
 836 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 837
 838 /* addr.c */
 839 extern const struct address_space_operations ceph_aops;
 840 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 841
 842 /* file.c */
 843 extern const struct file_operations ceph_file_fops;
 844 extern const struct address_space_operations ceph_aops;
 845 extern int ceph_open(struct inode *inode, struct file *file);
 846 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 847                                        struct nameidata *nd, int mode,
 848                                        int locked_dir);
 849 extern int ceph_release(struct inode *inode, struct file *filp);
 850 extern void ceph_release_page_vector(struct page **pages, int num_pages);
 851
 852 /* dir.c */
 853 extern const struct file_operations ceph_dir_fops;
 854 extern const struct inode_operations ceph_dir_iops;
 855 extern struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 856         ceph_snapdir_dentry_ops;
 857
 858 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 859 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 860                                          struct dentry *dentry, int err);
 861
 862 extern void ceph_dentry_lru_add(struct dentry *dn);
 863 extern void ceph_dentry_lru_touch(struct dentry *dn);
 864 extern void ceph_dentry_lru_del(struct dentry *dn);
 865
 866 /*
 867  * our d_ops vary depending on whether the inode is live,
 868  * snapshotted (read-only), or a virtual ".snap" directory.
 869  */
 870 int ceph_init_dentry(struct dentry *dentry);
 871
 872
 873 /* ioctl.c */
 874 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 875
 876 /* export.c */
 877 extern const struct export_operations ceph_export_ops;
 878
 879 /* debugfs.c */
 880 extern int ceph_debugfs_init(void);
 881 extern void ceph_debugfs_cleanup(void);
 882 extern int ceph_debugfs_client_init(struct ceph_client *client);
 883 extern void ceph_debugfs_client_cleanup(struct ceph_client *client);
 884
 885 static inline struct inode *get_dentry_parent_inode(struct dentry *dentry)
 886 {
 887         if (dentry && dentry->d_parent)
 888                 return dentry->d_parent->d_inode;
 889
 890         return NULL;
 891 }
 892
 893 #endif /* _FS_CEPH_SUPER_H */