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   22  /* 4 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 #define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
  32
  33 #define CEPH_MOUNT_OPT_DEFAULT    (CEPH_MOUNT_OPT_RBYTES)
  34
  35 #define ceph_set_mount_opt(fsc, opt) \
  36         (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt;
  37 #define ceph_test_mount_opt(fsc, opt) \
  38         (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
  39
  40 #define CEPH_RSIZE_DEFAULT             0           /* max read size */
  41 #define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
  42 #define CEPH_MAX_READDIR_DEFAULT        1024
  43 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
  44 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
  45
  46 struct ceph_mount_options {
  47         int flags;
  48         int sb_flags;
  49
  50         int wsize;            /* max write size */
  51         int rsize;            /* max read size */
  52         int rasize;           /* max readahead */
  53         int congestion_kb;    /* max writeback in flight */
  54         int caps_wanted_delay_min, caps_wanted_delay_max;
  55         int cap_release_safety;
  56         int max_readdir;       /* max readdir result (entires) */
  57         int max_readdir_bytes; /* max readdir result (bytes) */
  58
  59         /*
  60          * everything above this point can be memcmp'd; everything below
  61          * is handled in compare_mount_options()
  62          */
  63
  64         char *snapdir_name;   /* default ".snap" */
  65 };
  66
  67 struct ceph_fs_client {
  68         struct super_block *sb;
  69
  70         struct ceph_mount_options *mount_options;
  71         struct ceph_client *client;
  72
  73         unsigned long mount_state;
  74         int min_caps;                  /* min caps i added */
  75
  76         struct ceph_mds_client *mdsc;
  77
  78         /* writeback */
  79         mempool_t *wb_pagevec_pool;
  80         struct workqueue_struct *wb_wq;
  81         struct workqueue_struct *pg_inv_wq;
  82         struct workqueue_struct *trunc_wq;
  83         atomic_long_t writeback_count;
  84
  85         struct backing_dev_info backing_dev_info;
  86
  87 #ifdef CONFIG_DEBUG_FS
  88         struct dentry *debugfs_dentry_lru, *debugfs_caps;
  89         struct dentry *debugfs_congestion_kb;
  90         struct dentry *debugfs_bdi;
  91         struct dentry *debugfs_mdsc, *debugfs_mdsmap;
  92 #endif
  93 };
  94
  95
  96 /*
  97  * File i/o capability.  This tracks shared state with the metadata
  98  * server that allows us to cache or writeback attributes or to read
  99  * and write data.  For any given inode, we should have one or more
 100  * capabilities, one issued by each metadata server, and our
 101  * cumulative access is the OR of all issued capabilities.
 102  *
 103  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
 104  * session capability lists.
 105  */
 106 struct ceph_cap {
 107         struct ceph_inode_info *ci;
 108         struct rb_node ci_node;          /* per-ci cap tree */
 109         struct ceph_mds_session *session;
 110         struct list_head session_caps;   /* per-session caplist */
 111         int mds;
 112         u64 cap_id;       /* unique cap id (mds provided) */
 113         int issued;       /* latest, from the mds */
 114         int implemented;  /* implemented superset of issued (for revocation) */
 115         int mds_wanted;
 116         u32 seq, issue_seq, mseq;
 117         u32 cap_gen;      /* active/stale cycle */
 118         unsigned long last_used;
 119         struct list_head caps_item;
 120 };
 121
 122 #define CHECK_CAPS_NODELAY    1  /* do not delay any further */
 123 #define CHECK_CAPS_AUTHONLY   2  /* only check auth cap */
 124 #define CHECK_CAPS_FLUSH      4  /* flush any dirty caps */
 125
 126 /*
 127  * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 128  * we first complete any in-process sync writes and writeback any dirty
 129  * data before flushing the snapped state (tracked here) back to the MDS.
 130  */
 131 struct ceph_cap_snap {
 132         atomic_t nref;
 133         struct ceph_inode_info *ci;
 134         struct list_head ci_item, flushing_item;
 135
 136         u64 follows, flush_tid;
 137         int issued, dirty;
 138         struct ceph_snap_context *context;
 139
 140         umode_t mode;
 141         kuid_t uid;
 142         kgid_t gid;
 143
 144         struct ceph_buffer *xattr_blob;
 145         u64 xattr_version;
 146
 147         u64 size;
 148         struct timespec mtime, atime, ctime;
 149         u64 time_warp_seq;
 150         int writing;   /* a sync write is still in progress */
 151         int dirty_pages;     /* dirty pages awaiting writeback */
 152 };
 153
 154 static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 155 {
 156         if (atomic_dec_and_test(&capsnap->nref)) {
 157                 if (capsnap->xattr_blob)
 158                         ceph_buffer_put(capsnap->xattr_blob);
 159                 kfree(capsnap);
 160         }
 161 }
 162
 163 /*
 164  * The frag tree describes how a directory is fragmented, potentially across
 165  * multiple metadata servers.  It is also used to indicate points where
 166  * metadata authority is delegated, and whether/where metadata is replicated.
 167  *
 168  * A _leaf_ frag will be present in the i_fragtree IFF there is
 169  * delegation info.  That is, if mds >= 0 || ndist > 0.
 170  */
 171 #define CEPH_MAX_DIRFRAG_REP 4
 172
 173 struct ceph_inode_frag {
 174         struct rb_node node;
 175
 176         /* fragtree state */
 177         u32 frag;
 178         int split_by;         /* i.e. 2^(split_by) children */
 179
 180         /* delegation and replication info */
 181         int mds;              /* -1 if same authority as parent */
 182         int ndist;            /* >0 if replicated */
 183         int dist[CEPH_MAX_DIRFRAG_REP];
 184 };
 185
 186 /*
 187  * We cache inode xattrs as an encoded blob until they are first used,
 188  * at which point we parse them into an rbtree.
 189  */
 190 struct ceph_inode_xattr {
 191         struct rb_node node;
 192
 193         const char *name;
 194         int name_len;
 195         const char *val;
 196         int val_len;
 197         int dirty;
 198
 199         int should_free_name;
 200         int should_free_val;
 201 };
 202
 203 /*
 204  * Ceph dentry state
 205  */
 206 struct ceph_dentry_info {
 207         struct ceph_mds_session *lease_session;
 208         u32 lease_gen, lease_shared_gen;
 209         u32 lease_seq;
 210         unsigned long lease_renew_after, lease_renew_from;
 211         struct list_head lru;
 212         struct dentry *dentry;
 213         u64 time;
 214         u64 offset;
 215 };
 216
 217 struct ceph_inode_xattrs_info {
 218         /*
 219          * (still encoded) xattr blob. we avoid the overhead of parsing
 220          * this until someone actually calls getxattr, etc.
 221          *
 222          * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 223          * NULL means we don't know.
 224         */
 225         struct ceph_buffer *blob, *prealloc_blob;
 226
 227         struct rb_root index;
 228         bool dirty;
 229         int count;
 230         int names_size;
 231         int vals_size;
 232         u64 version, index_version;
 233 };
 234
 235 /*
 236  * Ceph inode.
 237  */
 238 struct ceph_inode_info {
 239         struct ceph_vino i_vino;   /* ceph ino + snap */
 240
 241         spinlock_t i_ceph_lock;
 242
 243         u64 i_version;
 244         u32 i_time_warp_seq;
 245
 246         unsigned i_ceph_flags;
 247         atomic_t i_release_count;
 248         atomic_t i_complete_count;
 249
 250         struct ceph_dir_layout i_dir_layout;
 251         struct ceph_file_layout i_layout;
 252         char *i_symlink;
 253
 254         /* for dirs */
 255         struct timespec i_rctime;
 256         u64 i_rbytes, i_rfiles, i_rsubdirs;
 257         u64 i_files, i_subdirs;
 258         u64 i_max_offset;  /* largest readdir offset, set with complete dir */
 259
 260         struct rb_root i_fragtree;
 261         struct mutex i_fragtree_mutex;
 262
 263         struct ceph_inode_xattrs_info i_xattrs;
 264
 265         /* capabilities.  protected _both_ by i_ceph_lock and cap->session's
 266          * s_mutex. */
 267         struct rb_root i_caps;           /* cap list */
 268         struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 269         unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 270         struct list_head i_dirty_item, i_flushing_item;
 271         u64 i_cap_flush_seq;
 272         /* we need to track cap writeback on a per-cap-bit basis, to allow
 273          * overlapping, pipelined cap flushes to the mds.  we can probably
 274          * reduce the tid to 8 bits if we're concerned about inode size. */
 275         u16 i_cap_flush_last_tid, i_cap_flush_tid[CEPH_CAP_BITS];
 276         wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 277         unsigned long i_hold_caps_min; /* jiffies */
 278         unsigned long i_hold_caps_max; /* jiffies */
 279         struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 280         int i_cap_exporting_mds;         /* to handle cap migration between */
 281         unsigned i_cap_exporting_mseq;   /*  mds's. */
 282         unsigned i_cap_exporting_issued;
 283         struct ceph_cap_reservation i_cap_migration_resv;
 284         struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 285         struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 286                                                     dirty|flushing caps */
 287         unsigned i_snap_caps;           /* cap bits for snapped files */
 288
 289         int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */
 290
 291         u32 i_truncate_seq;        /* last truncate to smaller size */
 292         u64 i_truncate_size;       /*  and the size we last truncated down to */
 293         int i_truncate_pending;    /*  still need to call vmtruncate */
 294
 295         u64 i_max_size;            /* max file size authorized by mds */
 296         u64 i_reported_size; /* (max_)size reported to or requested of mds */
 297         u64 i_wanted_max_size;     /* offset we'd like to write too */
 298         u64 i_requested_max_size;  /* max_size we've requested */
 299
 300         /* held references to caps */
 301         int i_pin_ref;
 302         int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref;
 303         int i_wrbuffer_ref, i_wrbuffer_ref_head;
 304         u32 i_shared_gen;       /* increment each time we get FILE_SHARED */
 305         u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 306         u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 307
 308         struct list_head i_unsafe_writes; /* uncommitted sync writes */
 309         struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 310         spinlock_t i_unsafe_lock;
 311
 312         struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 313         int i_snap_realm_counter; /* snap realm (if caps) */
 314         struct list_head i_snap_realm_item;
 315         struct list_head i_snap_flush_item;
 316
 317         struct work_struct i_wb_work;  /* writeback work */
 318         struct work_struct i_pg_inv_work;  /* page invalidation work */
 319
 320         struct work_struct i_vmtruncate_work;
 321
 322         struct inode vfs_inode; /* at end */
 323 };
 324
 325 static inline struct ceph_inode_info *ceph_inode(struct inode *inode)
 326 {
 327         return container_of(inode, struct ceph_inode_info, vfs_inode);
 328 }
 329
 330 static inline struct ceph_fs_client *ceph_inode_to_client(struct inode *inode)
 331 {
 332         return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 333 }
 334
 335 static inline struct ceph_fs_client *ceph_sb_to_client(struct super_block *sb)
 336 {
 337         return (struct ceph_fs_client *)sb->s_fs_info;
 338 }
 339
 340 static inline struct ceph_vino ceph_vino(struct inode *inode)
 341 {
 342         return ceph_inode(inode)->i_vino;
 343 }
 344
 345 /*
 346  * ino_t is <64 bits on many architectures, blech.
 347  *
 348  *               i_ino (kernel inode)   st_ino (userspace)
 349  * i386          32                     32
 350  * x86_64+ino32  64                     32
 351  * x86_64        64                     64
 352  */
 353 static inline u32 ceph_ino_to_ino32(__u64 vino)
 354 {
 355         u32 ino = vino & 0xffffffff;
 356         ino ^= vino >> 32;
 357         if (!ino)
 358                 ino = 2;
 359         return ino;
 360 }
 361
 362 /*
 363  * kernel i_ino value
 364  */
 365 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 366 {
 367 #if BITS_PER_LONG == 32
 368         return ceph_ino_to_ino32(vino.ino);
 369 #else
 370         return (ino_t)vino.ino;
 371 #endif
 372 }
 373
 374 /*
 375  * user-visible ino (stat, filldir)
 376  */
 377 #if BITS_PER_LONG == 32
 378 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
 379 {
 380         return ino;
 381 }
 382 #else
 383 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
 384 {
 385         if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
 386                 ino = ceph_ino_to_ino32(ino);
 387         return ino;
 388 }
 389 #endif
 390
 391
 392 /* for printf-style formatting */
 393 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 394
 395 static inline u64 ceph_ino(struct inode *inode)
 396 {
 397         return ceph_inode(inode)->i_vino.ino;
 398 }
 399 static inline u64 ceph_snap(struct inode *inode)
 400 {
 401         return ceph_inode(inode)->i_vino.snap;
 402 }
 403
 404 static inline int ceph_ino_compare(struct inode *inode, void *data)
 405 {
 406         struct ceph_vino *pvino = (struct ceph_vino *)data;
 407         struct ceph_inode_info *ci = ceph_inode(inode);
 408         return ci->i_vino.ino == pvino->ino &&
 409                 ci->i_vino.snap == pvino->snap;
 410 }
 411
 412 static inline struct inode *ceph_find_inode(struct super_block *sb,
 413                                             struct ceph_vino vino)
 414 {
 415         ino_t t = ceph_vino_to_ino(vino);
 416         return ilookup5(sb, t, ceph_ino_compare, &vino);
 417 }
 418
 419
 420 /*
 421  * Ceph inode.
 422  */
 423 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 424 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 425 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
 426
 427 static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
 428                                            int release_count)
 429 {
 430         atomic_set(&ci->i_complete_count, release_count);
 431 }
 432
 433 static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
 434 {
 435         atomic_inc(&ci->i_release_count);
 436 }
 437
 438 static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
 439 {
 440         return atomic_read(&ci->i_complete_count) ==
 441                 atomic_read(&ci->i_release_count);
 442 }
 443
 444 static inline void ceph_dir_clear_complete(struct inode *inode)
 445 {
 446         __ceph_dir_clear_complete(ceph_inode(inode));
 447 }
 448
 449 static inline bool ceph_dir_is_complete(struct inode *inode)
 450 {
 451         return __ceph_dir_is_complete(ceph_inode(inode));
 452 }
 453
 454
 455 /* find a specific frag @f */
 456 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 457                                                 u32 f);
 458
 459 /*
 460  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 461  * exists
 462  */
 463 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 464                             struct ceph_inode_frag *pfrag,
 465                             int *found);
 466
 467 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 468 {
 469         return (struct ceph_dentry_info *)dentry->d_fsdata;
 470 }
 471
 472 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 473 {
 474         return ((loff_t)frag << 32) | (loff_t)off;
 475 }
 476
 477 /*
 478  * caps helpers
 479  */
 480 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 481 {
 482         return !RB_EMPTY_ROOT(&ci->i_caps);
 483 }
 484
 485 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 486 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 487 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 488                                     struct ceph_cap *cap);
 489
 490 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 491 {
 492         int issued;
 493         spin_lock(&ci->i_ceph_lock);
 494         issued = __ceph_caps_issued(ci, NULL);
 495         spin_unlock(&ci->i_ceph_lock);
 496         return issued;
 497 }
 498
 499 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 500                                         int touch)
 501 {
 502         int r;
 503         spin_lock(&ci->i_ceph_lock);
 504         r = __ceph_caps_issued_mask(ci, mask, touch);
 505         spin_unlock(&ci->i_ceph_lock);
 506         return r;
 507 }
 508
 509 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 510 {
 511         return ci->i_dirty_caps | ci->i_flushing_caps;
 512 }
 513 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 514
 515 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 516 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 517
 518 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 519
 520 /*
 521  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 522  */
 523 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 524 {
 525         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 526         if (w & CEPH_CAP_FILE_BUFFER)
 527                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 528         return w;
 529 }
 530
 531 /* what the mds thinks we want */
 532 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 533
 534 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 535 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 536 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
 537 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 538                              struct ceph_cap_reservation *ctx, int need);
 539 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 540                                struct ceph_cap_reservation *ctx);
 541 extern void ceph_reservation_status(struct ceph_fs_client *client,
 542                                     int *total, int *avail, int *used,
 543                                     int *reserved, int *min);
 544
 545
 546
 547 /*
 548  * we keep buffered readdir results attached to file->private_data
 549  */
 550 #define CEPH_F_SYNC     1
 551 #define CEPH_F_ATEND    2
 552
 553 struct ceph_file_info {
 554         short fmode;     /* initialized on open */
 555         short flags;     /* CEPH_F_* */
 556
 557         /* readdir: position within the dir */
 558         u32 frag;
 559         struct ceph_mds_request *last_readdir;
 560
 561         /* readdir: position within a frag */
 562         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 563         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 564         char *last_name;       /* last entry in previous chunk */
 565         struct dentry *dentry; /* next dentry (for dcache readdir) */
 566         int dir_release_count;
 567
 568         /* used for -o dirstat read() on directory thing */
 569         char *dir_info;
 570         int dir_info_len;
 571 };
 572
 573
 574
 575 /*
 576  * A "snap realm" describes a subset of the file hierarchy sharing
 577  * the same set of snapshots that apply to it.  The realms themselves
 578  * are organized into a hierarchy, such that children inherit (some of)
 579  * the snapshots of their parents.
 580  *
 581  * All inodes within the realm that have capabilities are linked into a
 582  * per-realm list.
 583  */
 584 struct ceph_snap_realm {
 585         u64 ino;
 586         atomic_t nref;
 587         struct rb_node node;
 588
 589         u64 created, seq;
 590         u64 parent_ino;
 591         u64 parent_since;   /* snapid when our current parent became so */
 592
 593         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 594         u32 num_prior_parent_snaps;   /*  had prior to parent_since */
 595         u64 *snaps;                   /* snaps specific to this realm */
 596         u32 num_snaps;
 597
 598         struct ceph_snap_realm *parent;
 599         struct list_head children;       /* list of child realms */
 600         struct list_head child_item;
 601
 602         struct list_head empty_item;     /* if i have ref==0 */
 603
 604         struct list_head dirty_item;     /* if realm needs new context */
 605
 606         /* the current set of snaps for this realm */
 607         struct ceph_snap_context *cached_context;
 608
 609         struct list_head inodes_with_caps;
 610         spinlock_t inodes_with_caps_lock;
 611 };
 612
 613 static inline int default_congestion_kb(void)
 614 {
 615         int congestion_kb;
 616
 617         /*
 618          * Copied from NFS
 619          *
 620          * congestion size, scale with available memory.
 621          *
 622          *  64MB:    8192k
 623          * 128MB:   11585k
 624          * 256MB:   16384k
 625          * 512MB:   23170k
 626          *   1GB:   32768k
 627          *   2GB:   46340k
 628          *   4GB:   65536k
 629          *   8GB:   92681k
 630          *  16GB:  131072k
 631          *
 632          * This allows larger machines to have larger/more transfers.
 633          * Limit the default to 256M
 634          */
 635         congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
 636         if (congestion_kb > 256*1024)
 637                 congestion_kb = 256*1024;
 638
 639         return congestion_kb;
 640 }
 641
 642
 643
 644 /* snap.c */
 645 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 646                                                u64 ino);
 647 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 648                                 struct ceph_snap_realm *realm);
 649 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 650                                 struct ceph_snap_realm *realm);
 651 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 652                                   void *p, void *e, bool deletion);
 653 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 654                              struct ceph_mds_session *session,
 655                              struct ceph_msg *msg);
 656 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 657 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 658                                   struct ceph_cap_snap *capsnap);
 659 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 660
 661 /*
 662  * a cap_snap is "pending" if it is still awaiting an in-progress
 663  * sync write (that may/may not still update size, mtime, etc.).
 664  */
 665 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 666 {
 667         return !list_empty(&ci->i_cap_snaps) &&
 668                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 669                            ci_item)->writing;
 670 }
 671
 672 /* inode.c */
 673 extern const struct inode_operations ceph_file_iops;
 674
 675 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 676 extern void ceph_destroy_inode(struct inode *inode);
 677
 678 extern struct inode *ceph_get_inode(struct super_block *sb,
 679                                     struct ceph_vino vino);
 680 extern struct inode *ceph_get_snapdir(struct inode *parent);
 681 extern int ceph_fill_file_size(struct inode *inode, int issued,
 682                                u32 truncate_seq, u64 truncate_size, u64 size);
 683 extern void ceph_fill_file_time(struct inode *inode, int issued,
 684                                 u64 time_warp_seq, struct timespec *ctime,
 685                                 struct timespec *mtime, struct timespec *atime);
 686 extern int ceph_fill_trace(struct super_block *sb,
 687                            struct ceph_mds_request *req,
 688                            struct ceph_mds_session *session);
 689 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 690                                     struct ceph_mds_session *session);
 691
 692 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 693
 694 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 695 extern void __ceph_do_pending_vmtruncate(struct inode *inode, bool needlock);
 696 extern void ceph_queue_vmtruncate(struct inode *inode);
 697
 698 extern void ceph_queue_invalidate(struct inode *inode);
 699 extern void ceph_queue_writeback(struct inode *inode);
 700
 701 extern int ceph_do_getattr(struct inode *inode, int mask);
 702 extern int ceph_permission(struct inode *inode, int mask);
 703 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 704 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 705                         struct kstat *stat);
 706
 707 /* xattr.c */
 708 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 709                          size_t, int);
 710 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 711 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 712 extern int ceph_removexattr(struct dentry *, const char *);
 713 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 714 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 715 extern void __init ceph_xattr_init(void);
 716 extern void ceph_xattr_exit(void);
 717
 718 /* caps.c */
 719 extern const char *ceph_cap_string(int c);
 720 extern void ceph_handle_caps(struct ceph_mds_session *session,
 721                              struct ceph_msg *msg);
 722 extern int ceph_add_cap(struct inode *inode,
 723                         struct ceph_mds_session *session, u64 cap_id,
 724                         int fmode, unsigned issued, unsigned wanted,
 725                         unsigned cap, unsigned seq, u64 realmino, int flags,
 726                         struct ceph_cap_reservation *caps_reservation);
 727 extern void __ceph_remove_cap(struct ceph_cap *cap);
 728 static inline void ceph_remove_cap(struct ceph_cap *cap)
 729 {
 730         spin_lock(&cap->ci->i_ceph_lock);
 731         __ceph_remove_cap(cap);
 732         spin_unlock(&cap->ci->i_ceph_lock);
 733 }
 734 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
 735                          struct ceph_cap *cap);
 736
 737 extern void __queue_cap_release(struct ceph_mds_session *session, u64 ino,
 738                                 u64 cap_id, u32 migrate_seq, u32 issue_seq);
 739 extern void ceph_queue_caps_release(struct inode *inode);
 740 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
 741 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
 742                       int datasync);
 743 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 744                                     struct ceph_mds_session *session);
 745 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
 746                                              int mds);
 747 extern int ceph_get_cap_mds(struct inode *inode);
 748 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 749 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 750 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 751                                        struct ceph_snap_context *snapc);
 752 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 753                                struct ceph_mds_session **psession,
 754                                int again);
 755 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 756                             struct ceph_mds_session *session);
 757 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 758 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 759
 760 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 761                                      int mds, int drop, int unless, int force);
 762 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 763                                       int mds, int drop, int unless);
 764
 765 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 766                          int *got, loff_t endoff);
 767
 768 /* for counting open files by mode */
 769 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 770 {
 771         ci->i_nr_by_mode[mode]++;
 772 }
 773 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 774
 775 /* addr.c */
 776 extern const struct address_space_operations ceph_aops;
 777 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 778
 779 /* file.c */
 780 extern const struct file_operations ceph_file_fops;
 781 extern const struct address_space_operations ceph_aops;
 782
 783 extern int ceph_open(struct inode *inode, struct file *file);
 784 extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
 785                             struct file *file, unsigned flags, umode_t mode,
 786                             int *opened);
 787 extern int ceph_release(struct inode *inode, struct file *filp);
 788
 789 /* dir.c */
 790 extern const struct file_operations ceph_dir_fops;
 791 extern const struct inode_operations ceph_dir_iops;
 792 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 793         ceph_snapdir_dentry_ops;
 794
 795 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 796 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
 797                                struct dentry *dentry, int err);
 798 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 799                                          struct dentry *dentry, int err);
 800
 801 extern void ceph_dentry_lru_add(struct dentry *dn);
 802 extern void ceph_dentry_lru_touch(struct dentry *dn);
 803 extern void ceph_dentry_lru_del(struct dentry *dn);
 804 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 805 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
 806 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
 807
 808 /*
 809  * our d_ops vary depending on whether the inode is live,
 810  * snapshotted (read-only), or a virtual ".snap" directory.
 811  */
 812 int ceph_init_dentry(struct dentry *dentry);
 813
 814
 815 /* ioctl.c */
 816 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 817
 818 /* export.c */
 819 extern const struct export_operations ceph_export_ops;
 820
 821 /* locks.c */
 822 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
 823 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
 824 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
 825 extern int ceph_encode_locks_to_buffer(struct inode *inode,
 826                                        struct ceph_filelock *flocks,
 827                                        int num_fcntl_locks,
 828                                        int num_flock_locks);
 829 extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
 830                                   struct ceph_pagelist *pagelist,
 831                                   int num_fcntl_locks, int num_flock_locks);
 832 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 833
 834 /* debugfs.c */
 835 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 836 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
 837
 838 #endif /* _FS_CEPH_SUPER_H */