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 #define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
  31
  32 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES)
  33
  34 #define ceph_set_mount_opt(fsc, opt) \
  35         (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
  36 #define ceph_test_mount_opt(fsc, opt) \
  37         (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
  38
  39 #define CEPH_RSIZE_DEFAULT             0           /* max read size */
  40 #define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
  41 #define CEPH_MAX_READDIR_DEFAULT        1024
  42 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
  43 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
  44
  45 struct ceph_mount_options {
  46         int flags;
  47         int sb_flags;
  48
  49         int wsize;            /* max write size */
  50         int rsize;            /* max read size */
  51         int rasize;           /* max readahead */
  52         int congestion_kb;    /* max writeback in flight */
  53         int caps_wanted_delay_min, caps_wanted_delay_max;
  54         int cap_release_safety;
  55         int max_readdir;       /* max readdir result (entires) */
  56         int max_readdir_bytes; /* max readdir result (bytes) */
  57
  58         /*
  59          * everything above this point can be memcmp'd; everything below
  60          * is handled in compare_mount_options()
  61          */
  62
  63         char *snapdir_name;   /* default ".snap" */
  64 };
  65
  66 struct ceph_fs_client {
  67         struct super_block *sb;
  68
  69         struct ceph_mount_options *mount_options;
  70         struct ceph_client *client;
  71
  72         unsigned long mount_state;
  73         int min_caps;                  /* min caps i added */
  74
  75         struct ceph_mds_client *mdsc;
  76
  77         /* writeback */
  78         mempool_t *wb_pagevec_pool;
  79         struct workqueue_struct *wb_wq;
  80         struct workqueue_struct *pg_inv_wq;
  81         struct workqueue_struct *trunc_wq;
  82         atomic_long_t writeback_count;
  83
  84         struct backing_dev_info backing_dev_info;
  85
  86 #ifdef CONFIG_DEBUG_FS
  87         struct dentry *debugfs_dentry_lru, *debugfs_caps;
  88         struct dentry *debugfs_congestion_kb;
  89         struct dentry *debugfs_bdi;
  90         struct dentry *debugfs_mdsc, *debugfs_mdsmap;
  91 #endif
  92 };
  93
  94
  95 /*
  96  * File i/o capability.  This tracks shared state with the metadata
  97  * server that allows us to cache or writeback attributes or to read
  98  * and write data.  For any given inode, we should have one or more
  99  * capabilities, one issued by each metadata server, and our
 100  * cumulative access is the OR of all issued capabilities.
 101  *
 102  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
 103  * session capability lists.
 104  */
 105 struct ceph_cap {
 106         struct ceph_inode_info *ci;
 107         struct rb_node ci_node;          /* per-ci cap tree */
 108         struct ceph_mds_session *session;
 109         struct list_head session_caps;   /* per-session caplist */
 110         int mds;
 111         u64 cap_id;       /* unique cap id (mds provided) */
 112         int issued;       /* latest, from the mds */
 113         int implemented;  /* implemented superset of issued (for revocation) */
 114         int mds_wanted;
 115         u32 seq, issue_seq, mseq;
 116         u32 cap_gen;      /* active/stale cycle */
 117         unsigned long last_used;
 118         struct list_head caps_item;
 119 };
 120
 121 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
 122 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
 123 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
 124
 125 /*
 126  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 127  * we first complete any in-process sync writes and writeback any dirty
 128  * data before flushing the snapped state (tracked here) back to the MDS.
 129  */
 130 struct ceph_cap_snap {
 131         atomic_t nref;
 132         struct ceph_inode_info *ci;
 133         struct list_head ci_item, flushing_item;
 134
 135         u64 follows, flush_tid;
 136         int issued, dirty;
 137         struct ceph_snap_context *context;
 138
 139         mode_t mode;
 140         uid_t uid;
 141         gid_t gid;
 142
 143         struct ceph_buffer *xattr_blob;
 144         u64 xattr_version;
 145
 146         u64 size;
 147         struct timespec mtime, atime, ctime;
 148         u64 time_warp_seq;
 149         int writing;   /* a sync write is still in progress */
 150         int dirty_pages;     /* dirty pages awaiting writeback */
 151 };
 152
 153 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 154 {
 155         if (atomic_dec_and_test(&capsnap->nref)) {
 156                 if (capsnap->xattr_blob)
 157                         ceph_buffer_put(capsnap->xattr_blob);
 158                 kfree(capsnap);
 159         }
 160 }
 161
 162 /*
 163  * The frag tree describes how a directory is fragmented, potentially across
 164  * multiple metadata servers.  It is also used to indicate points where
 165  * metadata authority is delegated, and whether/where metadata is replicated.
 166  *
 167  * A _leaf_ frag will be present in the i_fragtree IFF there is
 168  * delegation info.  That is, if mds >= 0 || ndist > 0.
 169  */
 170 #define CEPH_MAX_DIRFRAG_REP 4
 171
 172 struct ceph_inode_frag {
 173         struct rb_node node;
 174
 175         /* fragtree state */
 176         u32 frag;
 177         int split_by;         /* i.e. 2^(split_by) children */
 178
 179         /* delegation and replication info */
 180         int mds;              /* -1 if same authority as parent */
 181         int ndist;            /* >0 if replicated */
 182         int dist[CEPH_MAX_DIRFRAG_REP];
 183 };
 184
 185 /*
 186  * We cache inode xattrs as an encoded blob until they are first used,
 187  * at which point we parse them into an rbtree.
 188  */
 189 struct ceph_inode_xattr {
 190         struct rb_node node;
 191
 192         const char *name;
 193         int name_len;
 194         const char *val;
 195         int val_len;
 196         int dirty;
 197
 198         int should_free_name;
 199         int should_free_val;
 200 };
 201
 202 /*
 203  * Ceph dentry state
 204  */
 205 struct ceph_dentry_info {
 206         struct ceph_mds_session *lease_session;
 207         u32 lease_gen, lease_shared_gen;
 208         u32 lease_seq;
 209         unsigned long lease_renew_after, lease_renew_from;
 210         struct list_head lru;
 211         struct dentry *dentry;
 212         u64 time;
 213         u64 offset;
 214 };
 215
 216 struct ceph_inode_xattrs_info {
 217         /*
 218          * (still encoded) xattr blob. we avoid the overhead of parsing
 219          * this until someone actually calls getxattr, etc.
 220          *
 221          * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 222          * NULL means we don't know.
 223         */
 224         struct ceph_buffer *blob, *prealloc_blob;
 225
 226         struct rb_root index;
 227         bool dirty;
 228         int count;
 229         int names_size;
 230         int vals_size;
 231         u64 version, index_version;
 232 };
 233
 234 /*
 235  * Ceph inode.
 236  */
 237 struct ceph_inode_info {
 238         struct ceph_vino i_vino;   /* ceph ino + snap */
 239
 240         u64 i_version;
 241         u32 i_time_warp_seq;
 242
 243         unsigned i_ceph_flags;
 244         unsigned long i_release_count;
 245
 246         struct ceph_dir_layout i_dir_layout;
 247         struct ceph_file_layout i_layout;
 248         char *i_symlink;
 249
 250         /* for dirs */
 251         struct timespec i_rctime;
 252         u64 i_rbytes, i_rfiles, i_rsubdirs;
 253         u64 i_files, i_subdirs;
 254         u64 i_max_offset;  /* largest readdir offset, set with I_COMPLETE */
 255
 256         struct rb_root i_fragtree;
 257         struct mutex i_fragtree_mutex;
 258
 259         struct ceph_inode_xattrs_info i_xattrs;
 260
 261         /* capabilities.  protected _both_ by i_lock and cap->session's
 262          * s_mutex. */
 263         struct rb_root i_caps;           /* cap list */
 264         struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 265         unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 266         struct list_head i_dirty_item, i_flushing_item;
 267         u64 i_cap_flush_seq;
 268         /* we need to track cap writeback on a per-cap-bit basis, to allow
 269          * overlapping, pipelined cap flushes to the mds.  we can probably
 270          * reduce the tid to 8 bits if we're concerned about inode size. */
 271         u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
 272         wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 273         unsigned long i_hold_caps_min; /* jiffies */
 274         unsigned long i_hold_caps_max; /* jiffies */
 275         struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 276         int i_cap_exporting_mds;         /* to handle cap migration between */
 277         unsigned i_cap_exporting_mseq;   /*  mds's. */
 278         unsigned i_cap_exporting_issued;
 279         struct ceph_cap_reservation i_cap_migration_resv;
 280         struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 281         struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 282                                                     dirty|flushing caps */
 283         unsigned i_snap_caps;           /* cap bits for snapped files */
 284
 285         int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 286
 287         u32 i_truncate_seq;        /* last truncate to smaller size */
 288         u64 i_truncate_size;       /*  and the size we last truncated down to */
 289         int i_truncate_pending;    /*  still need to call vmtruncate */
 290
 291         u64 i_max_size;            /* max file size authorized by mds */
 292         u64 i_reported_size; /* (max_)size reported to or requested of mds */
 293         u64 i_wanted_max_size;     /* offset we'd like to write too */
 294         u64 i_requested_max_size;  /* max_size we've requested */
 295
 296         /* held references to caps */
 297         int i_pin_ref;
 298         int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
 299         int i_wrbuffer_ref, i_wrbuffer_ref_head;
 300         u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 301         u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 302         u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 303
 304         struct list_head i_unsafe_writes; /* uncommitted sync writes */
 305         struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 306         spinlock_t i_unsafe_lock;
 307
 308         struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 309         int i_snap_realm_counter; /* snap realm (if caps) */
 310         struct list_head i_snap_realm_item;
 311         struct list_head i_snap_flush_item;
 312
 313         struct work_struct i_wb_work;  /* writeback work */
 314         struct work_struct i_pg_inv_work;  /* page invalidation work */
 315
 316         struct work_struct i_vmtruncate_work;
 317
 318         struct inode vfs_inode; /* at end */
 319 };
 320
 321 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 322 {
 323         return container_of(inode, struct ceph_inode_info, vfs_inode);
 324 }
 325
 326 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
 327 {
 328         return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 329 }
 330
 331 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
 332 {
 333         return (struct ceph_fs_client *)sb->s_fs_info;
 334 }
 335
 336 static inline struct ceph_vino ceph_vino(struct inode *inode)
 337 {
 338         return ceph_inode(inode)->i_vino;
 339 }
 340
 341 /*
 342  * ino_t is <64 bits on many architectures, blech.
 343  *
 344  *               i_ino (kernel inode)   st_ino (userspace)
 345  * i386          32                     32
 346  * x86_64+ino32  64                     32
 347  * x86_64        64                     64
 348  */
 349 static inline u32 ceph_ino_to_ino32(__u64 vino)
 350 {
 351         u32 ino = vino & 0xffffffff;
 352         ino ^= vino >> 32;
 353         if (!ino)
 354                 ino = 1;
 355         return ino;
 356 }
 357
 358 /*
 359  * kernel i_ino value
 360  */
 361 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 362 {
 363 #if BITS_PER_LONG == 32
 364         return ceph_ino_to_ino32(vino.ino);
 365 #else
 366         return (ino_t)vino.ino;
 367 #endif
 368 }
 369
 370 /*
 371  * user-visible ino (stat, filldir)
 372  */
 373 #if BITS_PER_LONG == 32
 374 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
 375 {
 376         return ino;
 377 }
 378 #else
 379 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
 380 {
 381         if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
 382                 ino = ceph_ino_to_ino32(ino);
 383         return ino;
 384 }
 385 #endif
 386
 387
 388 /* for printf-style formatting */
 389 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 390
 391 static inline u64 ceph_ino(struct inode *inode)
 392 {
 393         return ceph_inode(inode)->i_vino.ino;
 394 }
 395 static inline u64 ceph_snap(struct inode *inode)
 396 {
 397         return ceph_inode(inode)->i_vino.snap;
 398 }
 399
 400 static inline int ceph_ino_compare(struct inode *inode, void *data)
 401 {
 402         struct ceph_vino *pvino = (struct ceph_vino *)data;
 403         struct ceph_inode_info *ci = ceph_inode(inode);
 404         return ci->i_vino.ino == pvino->ino &&
 405                 ci->i_vino.snap == pvino->snap;
 406 }
 407
 408 static inline struct inode *ceph_find_inode(struct super_block *sb,
 409                                             struct ceph_vino vino)
 410 {
 411         ino_t t = ceph_vino_to_ino(vino);
 412         return ilookup5(sb, t, ceph_ino_compare, &vino);
 413 }
 414
 415
 416 /*
 417  * Ceph inode.
 418  */
 419 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
 420 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 421 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 422 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
 423
 424 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 425 {
 426         struct ceph_inode_info *ci = ceph_inode(inode);
 427
 428         spin_lock(&inode->i_lock);
 429         ci->i_ceph_flags &= ~mask;
 430         spin_unlock(&inode->i_lock);
 431 }
 432
 433 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 434 {
 435         struct ceph_inode_info *ci = ceph_inode(inode);
 436
 437         spin_lock(&inode->i_lock);
 438         ci->i_ceph_flags |= mask;
 439         spin_unlock(&inode->i_lock);
 440 }
 441
 442 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
 443 {
 444         struct ceph_inode_info *ci = ceph_inode(inode);
 445         bool r;
 446
 447         spin_lock(&inode->i_lock);
 448         r = (ci->i_ceph_flags & mask) == mask;
 449         spin_unlock(&inode->i_lock);
 450         return r;
 451 }
 452
 453
 454 /* find a specific frag @f */
 455 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 456                                                 u32 f);
 457
 458 /*
 459  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 460  * exists
 461  */
 462 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 463                             struct ceph_inode_frag *pfrag,
 464                             int *found);
 465
 466 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 467 {
 468         return (struct ceph_dentry_info *)dentry->d_fsdata;
 469 }
 470
 471 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 472 {
 473         return ((loff_t)frag << 32) | (loff_t)off;
 474 }
 475
 476 /*
 477  * caps helpers
 478  */
 479 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 480 {
 481         return !RB_EMPTY_ROOT(&ci->i_caps);
 482 }
 483
 484 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 485 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 486 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 487                                     struct ceph_cap *cap);
 488
 489 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 490 {
 491         int issued;
 492         spin_lock(&ci->vfs_inode.i_lock);
 493         issued = __ceph_caps_issued(ci, NULL);
 494         spin_unlock(&ci->vfs_inode.i_lock);
 495         return issued;
 496 }
 497
 498 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 499                                         int touch)
 500 {
 501         int r;
 502         spin_lock(&ci->vfs_inode.i_lock);
 503         r = __ceph_caps_issued_mask(ci, mask, touch);
 504         spin_unlock(&ci->vfs_inode.i_lock);
 505         return r;
 506 }
 507
 508 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 509 {
 510         return ci->i_dirty_caps | ci->i_flushing_caps;
 511 }
 512 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 513
 514 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 515 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 516
 517 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 518
 519 /*
 520  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 521  */
 522 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 523 {
 524         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 525         if (w & CEPH_CAP_FILE_BUFFER)
 526                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 527         return w;
 528 }
 529
 530 /* what the mds thinks we want */
 531 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 532
 533 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 534 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 535 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
 536 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 537                              struct ceph_cap_reservation *ctx, int need);
 538 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 539                                struct ceph_cap_reservation *ctx);
 540 extern void ceph_reservation_status(struct ceph_fs_client *client,
 541                                     int *total, int *avail, int *used,
 542                                     int *reserved, int *min);
 543
 544
 545
 546 /*
 547  * we keep buffered readdir results attached to file->private_data
 548  */
 549 #define CEPH_F_SYNC     1
 550 #define CEPH_F_ATEND    2
 551
 552 struct ceph_file_info {
 553         short fmode;     /* initialized on open */
 554         short flags;     /* CEPH_F_* */
 555
 556         /* readdir: position within the dir */
 557         u32 frag;
 558         struct ceph_mds_request *last_readdir;
 559
 560         /* readdir: position within a frag */
 561         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 562         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 563         char *last_name;       /* last entry in previous chunk */
 564         struct dentry *dentry; /* next dentry (for dcache readdir) */
 565         unsigned long dir_release_count;
 566
 567         /* used for -o dirstat read() on directory thing */
 568         char *dir_info;
 569         int dir_info_len;
 570 };
 571
 572
 573
 574 /*
 575  * A "snap realm" describes a subset of the file hierarchy sharing
 576  * the same set of snapshots that apply to it.  The realms themselves
 577  * are organized into a hierarchy, such that children inherit (some of)
 578  * the snapshots of their parents.
 579  *
 580  * All inodes within the realm that have capabilities are linked into a
 581  * per-realm list.
 582  */
 583 struct ceph_snap_realm {
 584         u64 ino;
 585         atomic_t nref;
 586         struct rb_node node;
 587
 588         u64 created, seq;
 589         u64 parent_ino;
 590         u64 parent_since;   /* snapid when our current parent became so */
 591
 592         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 593         int num_prior_parent_snaps;   /*  had prior to parent_since */
 594         u64 *snaps;                   /* snaps specific to this realm */
 595         int num_snaps;
 596
 597         struct ceph_snap_realm *parent;
 598         struct list_head children;       /* list of child realms */
 599         struct list_head child_item;
 600
 601         struct list_head empty_item;     /* if i have ref==0 */
 602
 603         struct list_head dirty_item;     /* if realm needs new context */
 604
 605         /* the current set of snaps for this realm */
 606         struct ceph_snap_context *cached_context;
 607
 608         struct list_head inodes_with_caps;
 609         spinlock_t inodes_with_caps_lock;
 610 };
 611
 612 static inline int default_congestion_kb(void)
 613 {
 614         int congestion_kb;
 615
 616         /*
 617          * Copied from NFS
 618          *
 619          * congestion size, scale with available memory.
 620          *
 621          *  64MB:    8192k
 622          * 128MB:   11585k
 623          * 256MB:   16384k
 624          * 512MB:   23170k
 625          *   1GB:   32768k
 626          *   2GB:   46340k
 627          *   4GB:   65536k
 628          *   8GB:   92681k
 629          *  16GB:  131072k
 630          *
 631          * This allows larger machines to have larger/more transfers.
 632          * Limit the default to 256M
 633          */
 634         congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
 635         if (congestion_kb > 256*1024)
 636                 congestion_kb = 256*1024;
 637
 638         return congestion_kb;
 639 }
 640
 641
 642
 643 /* snap.c */
 644 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 645                                                u64 ino);
 646 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 647                                 struct ceph_snap_realm *realm);
 648 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 649                                 struct ceph_snap_realm *realm);
 650 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 651                                   void *p, void *e, bool deletion);
 652 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 653                              struct ceph_mds_session *session,
 654                              struct ceph_msg *msg);
 655 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 656 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 657                                   struct ceph_cap_snap *capsnap);
 658 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 659
 660 /*
 661  * a cap_snap is "pending" if it is still awaiting an in-progress
 662  * sync write (that may/may not still update size, mtime, etc.).
 663  */
 664 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 665 {
 666         return !list_empty(&ci->i_cap_snaps) &&
 667                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 668                            ci_item)->writing;
 669 }
 670
 671 /* inode.c */
 672 extern const struct inode_operations ceph_file_iops;
 673
 674 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 675 extern void ceph_destroy_inode(struct inode *inode);
 676
 677 extern struct inode *ceph_get_inode(struct super_block *sb,
 678                                     struct ceph_vino vino);
 679 extern struct inode *ceph_get_snapdir(struct inode *parent);
 680 extern int ceph_fill_file_size(struct inode *inode, int issued,
 681                                u32 truncate_seq, u64 truncate_size, u64 size);
 682 extern void ceph_fill_file_time(struct inode *inode, int issued,
 683                                 u64 time_warp_seq, struct timespec *ctime,
 684                                 struct timespec *mtime, struct timespec *atime);
 685 extern int ceph_fill_trace(struct super_block *sb,
 686                            struct ceph_mds_request *req,
 687                            struct ceph_mds_session *session);
 688 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 689                                     struct ceph_mds_session *session);
 690
 691 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 692
 693 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 694 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 695 extern void ceph_queue_vmtruncate(struct inode *inode);
 696
 697 extern void ceph_queue_invalidate(struct inode *inode);
 698 extern void ceph_queue_writeback(struct inode *inode);
 699
 700 extern int ceph_do_getattr(struct inode *inode, int mask);
 701 extern int ceph_permission(struct inode *inode, int mask);
 702 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 703 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 704                         struct kstat *stat);
 705
 706 /* xattr.c */
 707 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 708                          size_t, int);
 709 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 710 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 711 extern int ceph_removexattr(struct dentry *, const char *);
 712 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 713 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 714
 715 /* caps.c */
 716 extern const char *ceph_cap_string(int c);
 717 extern void ceph_handle_caps(struct ceph_mds_session *session,
 718                              struct ceph_msg *msg);
 719 extern int ceph_add_cap(struct inode *inode,
 720                         struct ceph_mds_session *session, u64 cap_id,
 721                         int fmode, unsigned issued, unsigned wanted,
 722                         unsigned cap, unsigned seq, u64 realmino, int flags,
 723                         struct ceph_cap_reservation *caps_reservation);
 724 extern void __ceph_remove_cap(struct ceph_cap *cap);
 725 static inline void ceph_remove_cap(struct ceph_cap *cap)
 726 {
 727         struct inode *inode = &cap->ci->vfs_inode;
 728         spin_lock(&inode->i_lock);
 729         __ceph_remove_cap(cap);
 730         spin_unlock(&inode->i_lock);
 731 }
 732 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
 733                          struct ceph_cap *cap);
 734
 735 extern void ceph_queue_caps_release(struct inode *inode);
 736 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
 737 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
 738                       int datasync);
 739 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 740                                     struct ceph_mds_session *session);
 741 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
 742                                              int mds);
 743 extern int ceph_get_cap_mds(struct inode *inode);
 744 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 745 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 746 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 747                                        struct ceph_snap_context *snapc);
 748 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 749                                struct ceph_mds_session **psession,
 750                                int again);
 751 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 752                             struct ceph_mds_session *session);
 753 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 754 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 755
 756 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 757                                      int mds, int drop, int unless, int force);
 758 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 759                                       int mds, int drop, int unless);
 760
 761 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 762                          int *got, loff_t endoff);
 763
 764 /* for counting open files by mode */
 765 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 766 {
 767         ci->i_nr_by_mode[mode]++;
 768 }
 769 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 770
 771 /* addr.c */
 772 extern const struct address_space_operations ceph_aops;
 773 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 774
 775 /* file.c */
 776 extern const struct file_operations ceph_file_fops;
 777 extern const struct address_space_operations ceph_aops;
 778 extern int ceph_copy_to_page_vector(struct page **pages,
 779                                     const char *data,
 780                                     loff_t off, size_t len);
 781 extern int ceph_copy_from_page_vector(struct page **pages,
 782                                     char *data,
 783                                     loff_t off, size_t len);
 784 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
 785 extern int ceph_open(struct inode *inode, struct file *file);
 786 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 787                                        struct nameidata *nd, int mode,
 788                                        int locked_dir);
 789 extern int ceph_release(struct inode *inode, struct file *filp);
 790
 791 /* dir.c */
 792 extern const struct file_operations ceph_dir_fops;
 793 extern const struct inode_operations ceph_dir_iops;
 794 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 795         ceph_snapdir_dentry_ops;
 796
 797 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 798 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
 799                                struct dentry *dentry, int err);
 800 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 801                                          struct dentry *dentry, int err);
 802
 803 extern void ceph_dentry_lru_add(struct dentry *dn);
 804 extern void ceph_dentry_lru_touch(struct dentry *dn);
 805 extern void ceph_dentry_lru_del(struct dentry *dn);
 806 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 807 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
 808 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
 809
 810 /*
 811  * our d_ops vary depending on whether the inode is live,
 812  * snapshotted (read-only), or a virtual ".snap" directory.
 813  */
 814 int ceph_init_dentry(struct dentry *dentry);
 815
 816
 817 /* ioctl.c */
 818 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 819
 820 /* export.c */
 821 extern const struct export_operations ceph_export_ops;
 822
 823 /* locks.c */
 824 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
 825 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
 826 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
 827 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
 828                              int p_locks, int f_locks);
 829 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 830
 831 /* debugfs.c */
 832 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 833 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
 834
 835 #endif /* _FS_CEPH_SUPER_H */