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