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(ino_t ino)
 350 {
 351         ino ^= ino >> (sizeof(ino) * 8 - 32);
 352         if (!ino)
 353                 ino = 1;
 354         return ino;
 355 }
 356
 357 /*
 358  * kernel i_ino value
 359  */
 360 static inline ino_t ceph_vino_to_ino(struct ceph_vino vino)
 361 {
 362         ino_t ino = (ino_t)vino.ino;  /* ^ (vino.snap << 20); */
 363 #if BITS_PER_LONG == 32
 364         ino = ceph_ino_to_ino32(ino);
 365 #endif
 366         return ino;
 367 }
 368
 369 /*
 370  * user-visible ino (stat, filldir)
 371  */
 372 #if BITS_PER_LONG == 32
 373 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
 374 {
 375         return ino;
 376 }
 377 #else
 378 static inline ino_t ceph_translate_ino(struct super_block *sb, ino_t ino)
 379 {
 380         if (ceph_test_mount_opt(ceph_sb_to_client(sb), INO32))
 381                 ino = ceph_ino_to_ino32(ino);
 382         return ino;
 383 }
 384 #endif
 385
 386
 387 /* for printf-style formatting */
 388 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 389
 390 static inline u64 ceph_ino(struct inode *inode)
 391 {
 392         return ceph_inode(inode)->i_vino.ino;
 393 }
 394 static inline u64 ceph_snap(struct inode *inode)
 395 {
 396         return ceph_inode(inode)->i_vino.snap;
 397 }
 398
 399 static inline int ceph_ino_compare(struct inode *inode, void *data)
 400 {
 401         struct ceph_vino *pvino = (struct ceph_vino *)data;
 402         struct ceph_inode_info *ci = ceph_inode(inode);
 403         return ci->i_vino.ino == pvino->ino &&
 404                 ci->i_vino.snap == pvino->snap;
 405 }
 406
 407 static inline struct inode *ceph_find_inode(struct super_block *sb,
 408                                             struct ceph_vino vino)
 409 {
 410         ino_t t = ceph_vino_to_ino(vino);
 411         return ilookup5(sb, t, ceph_ino_compare, &vino);
 412 }
 413
 414
 415 /*
 416  * Ceph inode.
 417  */
 418 #define CEPH_I_COMPLETE  1  /* we have complete directory cached */
 419 #define CEPH_I_NODELAY   4  /* do not delay cap release */
 420 #define CEPH_I_FLUSH     8  /* do not delay flush of dirty metadata */
 421 #define CEPH_I_NOFLUSH  16  /* do not flush dirty caps */
 422
 423 static inline void ceph_i_clear(struct inode *inode, unsigned mask)
 424 {
 425         struct ceph_inode_info *ci = ceph_inode(inode);
 426
 427         spin_lock(&inode->i_lock);
 428         ci->i_ceph_flags &= ~mask;
 429         spin_unlock(&inode->i_lock);
 430 }
 431
 432 static inline void ceph_i_set(struct inode *inode, unsigned mask)
 433 {
 434         struct ceph_inode_info *ci = ceph_inode(inode);
 435
 436         spin_lock(&inode->i_lock);
 437         ci->i_ceph_flags |= mask;
 438         spin_unlock(&inode->i_lock);
 439 }
 440
 441 static inline bool ceph_i_test(struct inode *inode, unsigned mask)
 442 {
 443         struct ceph_inode_info *ci = ceph_inode(inode);
 444         bool r;
 445
 446         spin_lock(&inode->i_lock);
 447         r = (ci->i_ceph_flags & mask) == mask;
 448         spin_unlock(&inode->i_lock);
 449         return r;
 450 }
 451
 452
 453 /* find a specific frag @f */
 454 extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 455                                                 u32 f);
 456
 457 /*
 458  * choose fragment for value @v.  copy frag content to pfrag, if leaf
 459  * exists
 460  */
 461 extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 462                             struct ceph_inode_frag *pfrag,
 463                             int *found);
 464
 465 static inline struct ceph_dentry_info *ceph_dentry(struct dentry *dentry)
 466 {
 467         return (struct ceph_dentry_info *)dentry->d_fsdata;
 468 }
 469
 470 static inline loff_t ceph_make_fpos(unsigned frag, unsigned off)
 471 {
 472         return ((loff_t)frag << 32) | (loff_t)off;
 473 }
 474
 475 /*
 476  * caps helpers
 477  */
 478 static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 479 {
 480         return !RB_EMPTY_ROOT(&ci->i_caps);
 481 }
 482
 483 extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 484 extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 485 extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 486                                     struct ceph_cap *cap);
 487
 488 static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 489 {
 490         int issued;
 491         spin_lock(&ci->vfs_inode.i_lock);
 492         issued = __ceph_caps_issued(ci, NULL);
 493         spin_unlock(&ci->vfs_inode.i_lock);
 494         return issued;
 495 }
 496
 497 static inline int ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask,
 498                                         int touch)
 499 {
 500         int r;
 501         spin_lock(&ci->vfs_inode.i_lock);
 502         r = __ceph_caps_issued_mask(ci, mask, touch);
 503         spin_unlock(&ci->vfs_inode.i_lock);
 504         return r;
 505 }
 506
 507 static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 508 {
 509         return ci->i_dirty_caps | ci->i_flushing_caps;
 510 }
 511 extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask);
 512
 513 extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 514 extern int __ceph_caps_used(struct ceph_inode_info *ci);
 515
 516 extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 517
 518 /*
 519  * wanted, by virtue of open file modes AND cap refs (buffered/cached data)
 520  */
 521 static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 522 {
 523         int w = __ceph_caps_file_wanted(ci) | __ceph_caps_used(ci);
 524         if (w & CEPH_CAP_FILE_BUFFER)
 525                 w |= CEPH_CAP_FILE_EXCL;  /* we want EXCL if dirty data */
 526         return w;
 527 }
 528
 529 /* what the mds thinks we want */
 530 extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
 531
 532 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 533 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 534 extern void ceph_adjust_min_caps(struct ceph_mds_client *mdsc, int delta);
 535 extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 536                              struct ceph_cap_reservation *ctx, int need);
 537 extern int ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 538                                struct ceph_cap_reservation *ctx);
 539 extern void ceph_reservation_status(struct ceph_fs_client *client,
 540                                     int *total, int *avail, int *used,
 541                                     int *reserved, int *min);
 542
 543
 544
 545 /*
 546  * we keep buffered readdir results attached to file->private_data
 547  */
 548 #define CEPH_F_SYNC     1
 549 #define CEPH_F_ATEND    2
 550
 551 struct ceph_file_info {
 552         short fmode;     /* initialized on open */
 553         short flags;     /* CEPH_F_* */
 554
 555         /* readdir: position within the dir */
 556         u32 frag;
 557         struct ceph_mds_request *last_readdir;
 558
 559         /* readdir: position within a frag */
 560         unsigned offset;       /* offset of last chunk, adjusted for . and .. */
 561         u64 next_offset;       /* offset of next chunk (last_name's + 1) */
 562         char *last_name;       /* last entry in previous chunk */
 563         struct dentry *dentry; /* next dentry (for dcache readdir) */
 564         unsigned long dir_release_count;
 565
 566         /* used for -o dirstat read() on directory thing */
 567         char *dir_info;
 568         int dir_info_len;
 569 };
 570
 571
 572
 573 /*
 574  * A "snap realm" describes a subset of the file hierarchy sharing
 575  * the same set of snapshots that apply to it.  The realms themselves
 576  * are organized into a hierarchy, such that children inherit (some of)
 577  * the snapshots of their parents.
 578  *
 579  * All inodes within the realm that have capabilities are linked into a
 580  * per-realm list.
 581  */
 582 struct ceph_snap_realm {
 583         u64 ino;
 584         atomic_t nref;
 585         struct rb_node node;
 586
 587         u64 created, seq;
 588         u64 parent_ino;
 589         u64 parent_since;   /* snapid when our current parent became so */
 590
 591         u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 592         int num_prior_parent_snaps;   /*  had prior to parent_since */
 593         u64 *snaps;                   /* snaps specific to this realm */
 594         int num_snaps;
 595
 596         struct ceph_snap_realm *parent;
 597         struct list_head children;       /* list of child realms */
 598         struct list_head child_item;
 599
 600         struct list_head empty_item;     /* if i have ref==0 */
 601
 602         struct list_head dirty_item;     /* if realm needs new context */
 603
 604         /* the current set of snaps for this realm */
 605         struct ceph_snap_context *cached_context;
 606
 607         struct list_head inodes_with_caps;
 608         spinlock_t inodes_with_caps_lock;
 609 };
 610
 611 static inline int default_congestion_kb(void)
 612 {
 613         int congestion_kb;
 614
 615         /*
 616          * Copied from NFS
 617          *
 618          * congestion size, scale with available memory.
 619          *
 620          *  64MB:    8192k
 621          * 128MB:   11585k
 622          * 256MB:   16384k
 623          * 512MB:   23170k
 624          *   1GB:   32768k
 625          *   2GB:   46340k
 626          *   4GB:   65536k
 627          *   8GB:   92681k
 628          *  16GB:  131072k
 629          *
 630          * This allows larger machines to have larger/more transfers.
 631          * Limit the default to 256M
 632          */
 633         congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
 634         if (congestion_kb > 256*1024)
 635                 congestion_kb = 256*1024;
 636
 637         return congestion_kb;
 638 }
 639
 640
 641
 642 /* snap.c */
 643 struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 644                                                u64 ino);
 645 extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 646                                 struct ceph_snap_realm *realm);
 647 extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 648                                 struct ceph_snap_realm *realm);
 649 extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 650                                   void *p, void *e, bool deletion);
 651 extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 652                              struct ceph_mds_session *session,
 653                              struct ceph_msg *msg);
 654 extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 655 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 656                                   struct ceph_cap_snap *capsnap);
 657 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 658
 659 /*
 660  * a cap_snap is "pending" if it is still awaiting an in-progress
 661  * sync write (that may/may not still update size, mtime, etc.).
 662  */
 663 static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 664 {
 665         return !list_empty(&ci->i_cap_snaps) &&
 666                 list_entry(ci->i_cap_snaps.prev, struct ceph_cap_snap,
 667                            ci_item)->writing;
 668 }
 669
 670 /* inode.c */
 671 extern const struct inode_operations ceph_file_iops;
 672
 673 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 674 extern void ceph_destroy_inode(struct inode *inode);
 675
 676 extern struct inode *ceph_get_inode(struct super_block *sb,
 677                                     struct ceph_vino vino);
 678 extern struct inode *ceph_get_snapdir(struct inode *parent);
 679 extern int ceph_fill_file_size(struct inode *inode, int issued,
 680                                u32 truncate_seq, u64 truncate_size, u64 size);
 681 extern void ceph_fill_file_time(struct inode *inode, int issued,
 682                                 u64 time_warp_seq, struct timespec *ctime,
 683                                 struct timespec *mtime, struct timespec *atime);
 684 extern int ceph_fill_trace(struct super_block *sb,
 685                            struct ceph_mds_request *req,
 686                            struct ceph_mds_session *session);
 687 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 688                                     struct ceph_mds_session *session);
 689
 690 extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 691
 692 extern int ceph_inode_set_size(struct inode *inode, loff_t size);
 693 extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 694 extern void ceph_queue_vmtruncate(struct inode *inode);
 695
 696 extern void ceph_queue_invalidate(struct inode *inode);
 697 extern void ceph_queue_writeback(struct inode *inode);
 698
 699 extern int ceph_do_getattr(struct inode *inode, int mask);
 700 extern int ceph_permission(struct inode *inode, int mask);
 701 extern int ceph_setattr(struct dentry *dentry, struct iattr *attr);
 702 extern int ceph_getattr(struct vfsmount *mnt, struct dentry *dentry,
 703                         struct kstat *stat);
 704
 705 /* xattr.c */
 706 extern int ceph_setxattr(struct dentry *, const char *, const void *,
 707                          size_t, int);
 708 extern ssize_t ceph_getxattr(struct dentry *, const char *, void *, size_t);
 709 extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
 710 extern int ceph_removexattr(struct dentry *, const char *);
 711 extern void __ceph_build_xattrs_blob(struct ceph_inode_info *ci);
 712 extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
 713
 714 /* caps.c */
 715 extern const char *ceph_cap_string(int c);
 716 extern void ceph_handle_caps(struct ceph_mds_session *session,
 717                              struct ceph_msg *msg);
 718 extern int ceph_add_cap(struct inode *inode,
 719                         struct ceph_mds_session *session, u64 cap_id,
 720                         int fmode, unsigned issued, unsigned wanted,
 721                         unsigned cap, unsigned seq, u64 realmino, int flags,
 722                         struct ceph_cap_reservation *caps_reservation);
 723 extern void __ceph_remove_cap(struct ceph_cap *cap);
 724 static inline void ceph_remove_cap(struct ceph_cap *cap)
 725 {
 726         struct inode *inode = &cap->ci->vfs_inode;
 727         spin_lock(&inode->i_lock);
 728         __ceph_remove_cap(cap);
 729         spin_unlock(&inode->i_lock);
 730 }
 731 extern void ceph_put_cap(struct ceph_mds_client *mdsc,
 732                          struct ceph_cap *cap);
 733
 734 extern void ceph_queue_caps_release(struct inode *inode);
 735 extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
 736 extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
 737                       int datasync);
 738 extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
 739                                     struct ceph_mds_session *session);
 740 extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
 741                                              int mds);
 742 extern int ceph_get_cap_mds(struct inode *inode);
 743 extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
 744 extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
 745 extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 746                                        struct ceph_snap_context *snapc);
 747 extern void __ceph_flush_snaps(struct ceph_inode_info *ci,
 748                                struct ceph_mds_session **psession,
 749                                int again);
 750 extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 751                             struct ceph_mds_session *session);
 752 extern void ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
 753 extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 754
 755 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 756                                      int mds, int drop, int unless, int force);
 757 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
 758                                       int mds, int drop, int unless);
 759
 760 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
 761                          int *got, loff_t endoff);
 762
 763 /* for counting open files by mode */
 764 static inline void __ceph_get_fmode(struct ceph_inode_info *ci, int mode)
 765 {
 766         ci->i_nr_by_mode[mode]++;
 767 }
 768 extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode);
 769
 770 /* addr.c */
 771 extern const struct address_space_operations ceph_aops;
 772 extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
 773
 774 /* file.c */
 775 extern const struct file_operations ceph_file_fops;
 776 extern const struct address_space_operations ceph_aops;
 777 extern int ceph_copy_to_page_vector(struct page **pages,
 778                                     const char *data,
 779                                     loff_t off, size_t len);
 780 extern int ceph_copy_from_page_vector(struct page **pages,
 781                                     char *data,
 782                                     loff_t off, size_t len);
 783 extern struct page **ceph_alloc_page_vector(int num_pages, gfp_t flags);
 784 extern int ceph_open(struct inode *inode, struct file *file);
 785 extern struct dentry *ceph_lookup_open(struct inode *dir, struct dentry *dentry,
 786                                        struct nameidata *nd, int mode,
 787                                        int locked_dir);
 788 extern int ceph_release(struct inode *inode, struct file *filp);
 789
 790 /* dir.c */
 791 extern const struct file_operations ceph_dir_fops;
 792 extern const struct inode_operations ceph_dir_iops;
 793 extern const struct dentry_operations ceph_dentry_ops, ceph_snap_dentry_ops,
 794         ceph_snapdir_dentry_ops;
 795
 796 extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
 797 extern int ceph_handle_snapdir(struct ceph_mds_request *req,
 798                                struct dentry *dentry, int err);
 799 extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
 800                                          struct dentry *dentry, int err);
 801
 802 extern void ceph_dentry_lru_add(struct dentry *dn);
 803 extern void ceph_dentry_lru_touch(struct dentry *dn);
 804 extern void ceph_dentry_lru_del(struct dentry *dn);
 805 extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
 806 extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
 807 extern struct inode *ceph_get_dentry_parent_inode(struct dentry *dentry);
 808
 809 /*
 810  * our d_ops vary depending on whether the inode is live,
 811  * snapshotted (read-only), or a virtual ".snap" directory.
 812  */
 813 int ceph_init_dentry(struct dentry *dentry);
 814
 815
 816 /* ioctl.c */
 817 extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 818
 819 /* export.c */
 820 extern const struct export_operations ceph_export_ops;
 821
 822 /* locks.c */
 823 extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
 824 extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
 825 extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
 826 extern int ceph_encode_locks(struct inode *i, struct ceph_pagelist *p,
 827                              int p_locks, int f_locks);
 828 extern int lock_to_ceph_filelock(struct file_lock *fl, struct ceph_filelock *c);
 829
 830 /* debugfs.c */
 831 extern int ceph_fs_debugfs_init(struct ceph_fs_client *client);
 832 extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
 833
 834 #endif /* _FS_CEPH_SUPER_H */