linux/fs/ceph/super.h
<<
>>
Prefs
   1/* SPDX-License-Identifier: GPL-2.0 */
   2#ifndef _FS_CEPH_SUPER_H
   3#define _FS_CEPH_SUPER_H
   4
   5#include <linux/ceph/ceph_debug.h>
   6
   7#include <asm/unaligned.h>
   8#include <linux/backing-dev.h>
   9#include <linux/completion.h>
  10#include <linux/exportfs.h>
  11#include <linux/fs.h>
  12#include <linux/mempool.h>
  13#include <linux/pagemap.h>
  14#include <linux/wait.h>
  15#include <linux/writeback.h>
  16#include <linux/slab.h>
  17#include <linux/posix_acl.h>
  18#include <linux/refcount.h>
  19#include <linux/security.h>
  20
  21#include <linux/ceph/libceph.h>
  22
  23#ifdef CONFIG_CEPH_FSCACHE
  24#define FSCACHE_USE_NEW_IO_API
  25#include <linux/fscache.h>
  26#endif
  27
  28/* f_type in struct statfs */
  29#define CEPH_SUPER_MAGIC 0x00c36400
  30
  31/* large granularity for statfs utilization stats to facilitate
  32 * large volume sizes on 32-bit machines. */
  33#define CEPH_BLOCK_SHIFT   22  /* 4 MB */
  34#define CEPH_BLOCK         (1 << CEPH_BLOCK_SHIFT)
  35
  36#define CEPH_MOUNT_OPT_CLEANRECOVER    (1<<1) /* auto reonnect (clean mode) after blocklisted */
  37#define CEPH_MOUNT_OPT_DIRSTAT         (1<<4) /* `cat dirname` for stats */
  38#define CEPH_MOUNT_OPT_RBYTES          (1<<5) /* dir st_bytes = rbytes */
  39#define CEPH_MOUNT_OPT_NOASYNCREADDIR  (1<<7) /* no dcache readdir */
  40#define CEPH_MOUNT_OPT_INO32           (1<<8) /* 32 bit inos */
  41#define CEPH_MOUNT_OPT_DCACHE          (1<<9) /* use dcache for readdir etc */
  42#define CEPH_MOUNT_OPT_FSCACHE         (1<<10) /* use fscache */
  43#define CEPH_MOUNT_OPT_NOPOOLPERM      (1<<11) /* no pool permission check */
  44#define CEPH_MOUNT_OPT_MOUNTWAIT       (1<<12) /* mount waits if no mds is up */
  45#define CEPH_MOUNT_OPT_NOQUOTADF       (1<<13) /* no root dir quota in statfs */
  46#define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
  47#define CEPH_MOUNT_OPT_ASYNC_DIROPS    (1<<15) /* allow async directory ops */
  48
  49#define CEPH_MOUNT_OPT_DEFAULT                  \
  50        (CEPH_MOUNT_OPT_DCACHE |                \
  51         CEPH_MOUNT_OPT_NOCOPYFROM)
  52
  53#define ceph_set_mount_opt(fsc, opt) \
  54        (fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
  55#define ceph_clear_mount_opt(fsc, opt) \
  56        (fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
  57#define ceph_test_mount_opt(fsc, opt) \
  58        (!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
  59
  60/* max size of osd read request, limited by libceph */
  61#define CEPH_MAX_READ_SIZE              CEPH_MSG_MAX_DATA_LEN
  62/* osd has a configurable limitaion of max write size.
  63 * CEPH_MSG_MAX_DATA_LEN should be small enough. */
  64#define CEPH_MAX_WRITE_SIZE             CEPH_MSG_MAX_DATA_LEN
  65#define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
  66#define CEPH_MAX_READDIR_DEFAULT        1024
  67#define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
  68#define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
  69
  70/*
  71 * Delay telling the MDS we no longer want caps, in case we reopen
  72 * the file.  Delay a minimum amount of time, even if we send a cap
  73 * message for some other reason.  Otherwise, take the oppotunity to
  74 * update the mds to avoid sending another message later.
  75 */
  76#define CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT      5  /* cap release delay */
  77#define CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT     60  /* cap release delay */
  78
  79struct ceph_mount_options {
  80        unsigned int flags;
  81
  82        unsigned int wsize;            /* max write size */
  83        unsigned int rsize;            /* max read size */
  84        unsigned int rasize;           /* max readahead */
  85        unsigned int congestion_kb;    /* max writeback in flight */
  86        unsigned int caps_wanted_delay_min, caps_wanted_delay_max;
  87        int caps_max;
  88        unsigned int max_readdir;       /* max readdir result (entries) */
  89        unsigned int max_readdir_bytes; /* max readdir result (bytes) */
  90
  91        /*
  92         * everything above this point can be memcmp'd; everything below
  93         * is handled in compare_mount_options()
  94         */
  95
  96        char *snapdir_name;   /* default ".snap" */
  97        char *mds_namespace;  /* default NULL */
  98        char *server_path;    /* default NULL (means "/") */
  99        char *fscache_uniq;   /* default NULL */
 100};
 101
 102struct ceph_fs_client {
 103        struct super_block *sb;
 104
 105        struct list_head metric_wakeup;
 106
 107        struct ceph_mount_options *mount_options;
 108        struct ceph_client *client;
 109
 110        int mount_state;
 111
 112        bool blocklisted;
 113
 114        bool have_copy_from2;
 115
 116        u32 filp_gen;
 117        loff_t max_file_size;
 118
 119        struct ceph_mds_client *mdsc;
 120
 121        atomic_long_t writeback_count;
 122
 123        struct workqueue_struct *inode_wq;
 124        struct workqueue_struct *cap_wq;
 125
 126#ifdef CONFIG_DEBUG_FS
 127        struct dentry *debugfs_dentry_lru, *debugfs_caps;
 128        struct dentry *debugfs_congestion_kb;
 129        struct dentry *debugfs_bdi;
 130        struct dentry *debugfs_mdsc, *debugfs_mdsmap;
 131        struct dentry *debugfs_metric;
 132        struct dentry *debugfs_status;
 133        struct dentry *debugfs_mds_sessions;
 134#endif
 135
 136#ifdef CONFIG_CEPH_FSCACHE
 137        struct fscache_cookie *fscache;
 138#endif
 139};
 140
 141
 142/*
 143 * File i/o capability.  This tracks shared state with the metadata
 144 * server that allows us to cache or writeback attributes or to read
 145 * and write data.  For any given inode, we should have one or more
 146 * capabilities, one issued by each metadata server, and our
 147 * cumulative access is the OR of all issued capabilities.
 148 *
 149 * Each cap is referenced by the inode's i_caps rbtree and by per-mds
 150 * session capability lists.
 151 */
 152struct ceph_cap {
 153        struct ceph_inode_info *ci;
 154        struct rb_node ci_node;          /* per-ci cap tree */
 155        struct ceph_mds_session *session;
 156        struct list_head session_caps;   /* per-session caplist */
 157        u64 cap_id;       /* unique cap id (mds provided) */
 158        union {
 159                /* in-use caps */
 160                struct {
 161                        int issued;       /* latest, from the mds */
 162                        int implemented;  /* implemented superset of
 163                                             issued (for revocation) */
 164                        int mds;          /* mds index for this cap */
 165                        int mds_wanted;   /* caps wanted from this mds */
 166                };
 167                /* caps to release */
 168                struct {
 169                        u64 cap_ino;
 170                        int queue_release;
 171                };
 172        };
 173        u32 seq, issue_seq, mseq;
 174        u32 cap_gen;      /* active/stale cycle */
 175        unsigned long last_used;
 176        struct list_head caps_item;
 177};
 178
 179#define CHECK_CAPS_AUTHONLY   1  /* only check auth cap */
 180#define CHECK_CAPS_FLUSH      2  /* flush any dirty caps */
 181#define CHECK_CAPS_NOINVAL    4  /* don't invalidate pagecache */
 182
 183struct ceph_cap_flush {
 184        u64 tid;
 185        int caps;
 186        bool wake; /* wake up flush waiters when finish ? */
 187        bool is_capsnap; /* true means capsnap */
 188        struct list_head g_list; // global
 189        struct list_head i_list; // per inode
 190};
 191
 192/*
 193 * Snapped cap state that is pending flush to mds.  When a snapshot occurs,
 194 * we first complete any in-process sync writes and writeback any dirty
 195 * data before flushing the snapped state (tracked here) back to the MDS.
 196 */
 197struct ceph_cap_snap {
 198        refcount_t nref;
 199        struct list_head ci_item;
 200
 201        struct ceph_cap_flush cap_flush;
 202
 203        u64 follows;
 204        int issued, dirty;
 205        struct ceph_snap_context *context;
 206
 207        umode_t mode;
 208        kuid_t uid;
 209        kgid_t gid;
 210
 211        struct ceph_buffer *xattr_blob;
 212        u64 xattr_version;
 213
 214        u64 size;
 215        u64 change_attr;
 216        struct timespec64 mtime, atime, ctime, btime;
 217        u64 time_warp_seq;
 218        u64 truncate_size;
 219        u32 truncate_seq;
 220        int writing;   /* a sync write is still in progress */
 221        int dirty_pages;     /* dirty pages awaiting writeback */
 222        bool inline_data;
 223        bool need_flush;
 224};
 225
 226static inline void ceph_put_cap_snap(struct ceph_cap_snap *capsnap)
 227{
 228        if (refcount_dec_and_test(&capsnap->nref)) {
 229                if (capsnap->xattr_blob)
 230                        ceph_buffer_put(capsnap->xattr_blob);
 231                kfree(capsnap);
 232        }
 233}
 234
 235/*
 236 * The frag tree describes how a directory is fragmented, potentially across
 237 * multiple metadata servers.  It is also used to indicate points where
 238 * metadata authority is delegated, and whether/where metadata is replicated.
 239 *
 240 * A _leaf_ frag will be present in the i_fragtree IFF there is
 241 * delegation info.  That is, if mds >= 0 || ndist > 0.
 242 */
 243#define CEPH_MAX_DIRFRAG_REP 4
 244
 245struct ceph_inode_frag {
 246        struct rb_node node;
 247
 248        /* fragtree state */
 249        u32 frag;
 250        int split_by;         /* i.e. 2^(split_by) children */
 251
 252        /* delegation and replication info */
 253        int mds;              /* -1 if same authority as parent */
 254        int ndist;            /* >0 if replicated */
 255        int dist[CEPH_MAX_DIRFRAG_REP];
 256};
 257
 258/*
 259 * We cache inode xattrs as an encoded blob until they are first used,
 260 * at which point we parse them into an rbtree.
 261 */
 262struct ceph_inode_xattr {
 263        struct rb_node node;
 264
 265        const char *name;
 266        int name_len;
 267        const char *val;
 268        int val_len;
 269        int dirty;
 270
 271        int should_free_name;
 272        int should_free_val;
 273};
 274
 275/*
 276 * Ceph dentry state
 277 */
 278struct ceph_dentry_info {
 279        struct dentry *dentry;
 280        struct ceph_mds_session *lease_session;
 281        struct list_head lease_list;
 282        unsigned flags;
 283        int lease_shared_gen;
 284        u32 lease_gen;
 285        u32 lease_seq;
 286        unsigned long lease_renew_after, lease_renew_from;
 287        unsigned long time;
 288        u64 offset;
 289};
 290
 291#define CEPH_DENTRY_REFERENCED          1
 292#define CEPH_DENTRY_LEASE_LIST          2
 293#define CEPH_DENTRY_SHRINK_LIST         4
 294#define CEPH_DENTRY_PRIMARY_LINK        8
 295
 296struct ceph_inode_xattrs_info {
 297        /*
 298         * (still encoded) xattr blob. we avoid the overhead of parsing
 299         * this until someone actually calls getxattr, etc.
 300         *
 301         * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
 302         * NULL means we don't know.
 303        */
 304        struct ceph_buffer *blob, *prealloc_blob;
 305
 306        struct rb_root index;
 307        bool dirty;
 308        int count;
 309        int names_size;
 310        int vals_size;
 311        u64 version, index_version;
 312};
 313
 314/*
 315 * Ceph inode.
 316 */
 317struct ceph_inode_info {
 318        struct ceph_vino i_vino;   /* ceph ino + snap */
 319
 320        spinlock_t i_ceph_lock;
 321
 322        u64 i_version;
 323        u64 i_inline_version;
 324        u32 i_time_warp_seq;
 325
 326        unsigned long i_ceph_flags;
 327        atomic64_t i_release_count;
 328        atomic64_t i_ordered_count;
 329        atomic64_t i_complete_seq[2];
 330
 331        struct ceph_dir_layout i_dir_layout;
 332        struct ceph_file_layout i_layout;
 333        struct ceph_file_layout i_cached_layout;        // for async creates
 334        char *i_symlink;
 335
 336        /* for dirs */
 337        struct timespec64 i_rctime;
 338        u64 i_rbytes, i_rfiles, i_rsubdirs, i_rsnaps;
 339        u64 i_files, i_subdirs;
 340
 341        /* quotas */
 342        u64 i_max_bytes, i_max_files;
 343
 344        s32 i_dir_pin;
 345
 346        struct rb_root i_fragtree;
 347        int i_fragtree_nsplits;
 348        struct mutex i_fragtree_mutex;
 349
 350        struct ceph_inode_xattrs_info i_xattrs;
 351
 352        /* capabilities.  protected _both_ by i_ceph_lock and cap->session's
 353         * s_mutex. */
 354        struct rb_root i_caps;           /* cap list */
 355        struct ceph_cap *i_auth_cap;     /* authoritative cap, if any */
 356        unsigned i_dirty_caps, i_flushing_caps;     /* mask of dirtied fields */
 357
 358        /*
 359         * Link to the auth cap's session's s_cap_dirty list. s_cap_dirty
 360         * is protected by the mdsc->cap_dirty_lock, but each individual item
 361         * is also protected by the inode's i_ceph_lock. Walking s_cap_dirty
 362         * requires the mdsc->cap_dirty_lock. List presence for an item can
 363         * be tested under the i_ceph_lock. Changing anything requires both.
 364         */
 365        struct list_head i_dirty_item;
 366
 367        /*
 368         * Link to session's s_cap_flushing list. Protected in a similar
 369         * fashion to i_dirty_item, but also by the s_mutex for changes. The
 370         * s_cap_flushing list can be walked while holding either the s_mutex
 371         * or msdc->cap_dirty_lock. List presence can also be checked while
 372         * holding the i_ceph_lock for this inode.
 373         */
 374        struct list_head i_flushing_item;
 375
 376        /* we need to track cap writeback on a per-cap-bit basis, to allow
 377         * overlapping, pipelined cap flushes to the mds.  we can probably
 378         * reduce the tid to 8 bits if we're concerned about inode size. */
 379        struct ceph_cap_flush *i_prealloc_cap_flush;
 380        struct list_head i_cap_flush_list;
 381        wait_queue_head_t i_cap_wq;      /* threads waiting on a capability */
 382        unsigned long i_hold_caps_max; /* jiffies */
 383        struct list_head i_cap_delay_list;  /* for delayed cap release to mds */
 384        struct ceph_cap_reservation i_cap_migration_resv;
 385        struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
 386        struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
 387                                                    dirty|flushing caps */
 388        unsigned i_snap_caps;           /* cap bits for snapped files */
 389
 390        unsigned long i_last_rd;
 391        unsigned long i_last_wr;
 392        int i_nr_by_mode[CEPH_FILE_MODE_BITS];  /* open file counts */
 393
 394        struct mutex i_truncate_mutex;
 395        u32 i_truncate_seq;        /* last truncate to smaller size */
 396        u64 i_truncate_size;       /*  and the size we last truncated down to */
 397        int i_truncate_pending;    /*  still need to call vmtruncate */
 398
 399        u64 i_max_size;            /* max file size authorized by mds */
 400        u64 i_reported_size; /* (max_)size reported to or requested of mds */
 401        u64 i_wanted_max_size;     /* offset we'd like to write too */
 402        u64 i_requested_max_size;  /* max_size we've requested */
 403
 404        /* held references to caps */
 405        int i_pin_ref;
 406        int i_rd_ref, i_rdcache_ref, i_wr_ref, i_wb_ref, i_fx_ref;
 407        int i_wrbuffer_ref, i_wrbuffer_ref_head;
 408        atomic_t i_filelock_ref;
 409        atomic_t i_shared_gen;       /* increment each time we get FILE_SHARED */
 410        u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 411        u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 412
 413        struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 414        struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
 415        spinlock_t i_unsafe_lock;
 416
 417        union {
 418                struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
 419                struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
 420        };
 421        int i_snap_realm_counter; /* snap realm (if caps) */
 422        struct list_head i_snap_realm_item;
 423        struct list_head i_snap_flush_item;
 424        struct timespec64 i_btime;
 425        struct timespec64 i_snap_btime;
 426
 427        struct work_struct i_work;
 428        unsigned long  i_work_mask;
 429
 430#ifdef CONFIG_CEPH_FSCACHE
 431        struct fscache_cookie *fscache;
 432#endif
 433        errseq_t i_meta_err;
 434
 435        struct inode vfs_inode; /* at end */
 436};
 437
 438static inline struct ceph_inode_info *
 439ceph_inode(const struct inode *inode)
 440{
 441        return container_of(inode, struct ceph_inode_info, vfs_inode);
 442}
 443
 444static inline struct ceph_fs_client *
 445ceph_inode_to_client(const struct inode *inode)
 446{
 447        return (struct ceph_fs_client *)inode->i_sb->s_fs_info;
 448}
 449
 450static inline struct ceph_fs_client *
 451ceph_sb_to_client(const struct super_block *sb)
 452{
 453        return (struct ceph_fs_client *)sb->s_fs_info;
 454}
 455
 456static inline struct ceph_mds_client *
 457ceph_sb_to_mdsc(const struct super_block *sb)
 458{
 459        return (struct ceph_mds_client *)ceph_sb_to_client(sb)->mdsc;
 460}
 461
 462static inline struct ceph_vino
 463ceph_vino(const struct inode *inode)
 464{
 465        return ceph_inode(inode)->i_vino;
 466}
 467
 468static inline u32 ceph_ino_to_ino32(u64 vino)
 469{
 470        u32 ino = vino & 0xffffffff;
 471        ino ^= vino >> 32;
 472        if (!ino)
 473                ino = 2;
 474        return ino;
 475}
 476
 477/*
 478 * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
 479 * some arches. We generally do not use this value inside the ceph driver, but
 480 * we do want to set it to something, so that generic vfs code has an
 481 * appropriate value for tracepoints and the like.
 482 */
 483static inline ino_t ceph_vino_to_ino_t(struct ceph_vino vino)
 484{
 485        if (sizeof(ino_t) == sizeof(u32))
 486                return ceph_ino_to_ino32(vino.ino);
 487        return (ino_t)vino.ino;
 488}
 489
 490/* for printf-style formatting */
 491#define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
 492
 493static inline u64 ceph_ino(struct inode *inode)
 494{
 495        return ceph_inode(inode)->i_vino.ino;
 496}
 497
 498static inline u64 ceph_snap(struct inode *inode)
 499{
 500        return ceph_inode(inode)->i_vino.snap;
 501}
 502
 503/**
 504 * ceph_present_ino - format an inode number for presentation to userland
 505 * @sb: superblock where the inode lives
 506 * @ino: inode number to (possibly) convert
 507 *
 508 * If the user mounted with the ino32 option, then the 64-bit value needs
 509 * to be converted to something that can fit inside 32 bits. Note that
 510 * internal kernel code never uses this value, so this is entirely for
 511 * userland consumption.
 512 */
 513static inline u64 ceph_present_ino(struct super_block *sb, u64 ino)
 514{
 515        if (unlikely(ceph_test_mount_opt(ceph_sb_to_client(sb), INO32)))
 516                return ceph_ino_to_ino32(ino);
 517        return ino;
 518}
 519
 520static inline u64 ceph_present_inode(struct inode *inode)
 521{
 522        return ceph_present_ino(inode->i_sb, ceph_ino(inode));
 523}
 524
 525static inline int ceph_ino_compare(struct inode *inode, void *data)
 526{
 527        struct ceph_vino *pvino = (struct ceph_vino *)data;
 528        struct ceph_inode_info *ci = ceph_inode(inode);
 529        return ci->i_vino.ino == pvino->ino &&
 530                ci->i_vino.snap == pvino->snap;
 531}
 532
 533/*
 534 * The MDS reserves a set of inodes for its own usage. These should never
 535 * be accessible by clients, and so the MDS has no reason to ever hand these
 536 * out. The range is CEPH_MDS_INO_MDSDIR_OFFSET..CEPH_INO_SYSTEM_BASE.
 537 *
 538 * These come from src/mds/mdstypes.h in the ceph sources.
 539 */
 540#define CEPH_MAX_MDS            0x100
 541#define CEPH_NUM_STRAY          10
 542#define CEPH_MDS_INO_MDSDIR_OFFSET      (1 * CEPH_MAX_MDS)
 543#define CEPH_INO_SYSTEM_BASE            ((6*CEPH_MAX_MDS) + (CEPH_MAX_MDS * CEPH_NUM_STRAY))
 544
 545static inline bool ceph_vino_is_reserved(const struct ceph_vino vino)
 546{
 547        if (vino.ino < CEPH_INO_SYSTEM_BASE &&
 548            vino.ino >= CEPH_MDS_INO_MDSDIR_OFFSET) {
 549                WARN_RATELIMIT(1, "Attempt to access reserved inode number 0x%llx", vino.ino);
 550                return true;
 551        }
 552        return false;
 553}
 554
 555static inline struct inode *ceph_find_inode(struct super_block *sb,
 556                                            struct ceph_vino vino)
 557{
 558        if (ceph_vino_is_reserved(vino))
 559                return NULL;
 560
 561        /*
 562         * NB: The hashval will be run through the fs/inode.c hash function
 563         * anyway, so there is no need to squash the inode number down to
 564         * 32-bits first. Just use low-order bits on arches with 32-bit long.
 565         */
 566        return ilookup5(sb, (unsigned long)vino.ino, ceph_ino_compare, &vino);
 567}
 568
 569
 570/*
 571 * Ceph inode.
 572 */
 573#define CEPH_I_DIR_ORDERED      (1 << 0)  /* dentries in dir are ordered */
 574#define CEPH_I_FLUSH            (1 << 2)  /* do not delay flush of dirty metadata */
 575#define CEPH_I_POOL_PERM        (1 << 3)  /* pool rd/wr bits are valid */
 576#define CEPH_I_POOL_RD          (1 << 4)  /* can read from pool */
 577#define CEPH_I_POOL_WR          (1 << 5)  /* can write to pool */
 578#define CEPH_I_SEC_INITED       (1 << 6)  /* security initialized */
 579#define CEPH_I_KICK_FLUSH       (1 << 7)  /* kick flushing caps */
 580#define CEPH_I_FLUSH_SNAPS      (1 << 8)  /* need flush snapss */
 581#define CEPH_I_ERROR_WRITE      (1 << 9) /* have seen write errors */
 582#define CEPH_I_ERROR_FILELOCK   (1 << 10) /* have seen file lock errors */
 583#define CEPH_I_ODIRECT          (1 << 11) /* inode in direct I/O mode */
 584#define CEPH_ASYNC_CREATE_BIT   (12)      /* async create in flight for this */
 585#define CEPH_I_ASYNC_CREATE     (1 << CEPH_ASYNC_CREATE_BIT)
 586
 587/*
 588 * Masks of ceph inode work.
 589 */
 590#define CEPH_I_WORK_WRITEBACK           0
 591#define CEPH_I_WORK_INVALIDATE_PAGES    1
 592#define CEPH_I_WORK_VMTRUNCATE          2
 593#define CEPH_I_WORK_CHECK_CAPS          3
 594#define CEPH_I_WORK_FLUSH_SNAPS         4
 595
 596/*
 597 * We set the ERROR_WRITE bit when we start seeing write errors on an inode
 598 * and then clear it when they start succeeding. Note that we do a lockless
 599 * check first, and only take the lock if it looks like it needs to be changed.
 600 * The write submission code just takes this as a hint, so we're not too
 601 * worried if a few slip through in either direction.
 602 */
 603static inline void ceph_set_error_write(struct ceph_inode_info *ci)
 604{
 605        if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {
 606                spin_lock(&ci->i_ceph_lock);
 607                ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;
 608                spin_unlock(&ci->i_ceph_lock);
 609        }
 610}
 611
 612static inline void ceph_clear_error_write(struct ceph_inode_info *ci)
 613{
 614        if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {
 615                spin_lock(&ci->i_ceph_lock);
 616                ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;
 617                spin_unlock(&ci->i_ceph_lock);
 618        }
 619}
 620
 621static inline void __ceph_dir_set_complete(struct ceph_inode_info *ci,
 622                                           long long release_count,
 623                                           long long ordered_count)
 624{
 625        /*
 626         * Makes sure operations that setup readdir cache (update page
 627         * cache and i_size) are strongly ordered w.r.t. the following
 628         * atomic64_set() operations.
 629         */
 630        smp_mb();
 631        atomic64_set(&ci->i_complete_seq[0], release_count);
 632        atomic64_set(&ci->i_complete_seq[1], ordered_count);
 633}
 634
 635static inline void __ceph_dir_clear_complete(struct ceph_inode_info *ci)
 636{
 637        atomic64_inc(&ci->i_release_count);
 638}
 639
 640static inline void __ceph_dir_clear_ordered(struct ceph_inode_info *ci)
 641{
 642        atomic64_inc(&ci->i_ordered_count);
 643}
 644
 645static inline bool __ceph_dir_is_complete(struct ceph_inode_info *ci)
 646{
 647        return atomic64_read(&ci->i_complete_seq[0]) ==
 648                atomic64_read(&ci->i_release_count);
 649}
 650
 651static inline bool __ceph_dir_is_complete_ordered(struct ceph_inode_info *ci)
 652{
 653        return  atomic64_read(&ci->i_complete_seq[0]) ==
 654                atomic64_read(&ci->i_release_count) &&
 655                atomic64_read(&ci->i_complete_seq[1]) ==
 656                atomic64_read(&ci->i_ordered_count);
 657}
 658
 659static inline void ceph_dir_clear_complete(struct inode *inode)
 660{
 661        __ceph_dir_clear_complete(ceph_inode(inode));
 662}
 663
 664static inline void ceph_dir_clear_ordered(struct inode *inode)
 665{
 666        __ceph_dir_clear_ordered(ceph_inode(inode));
 667}
 668
 669static inline bool ceph_dir_is_complete_ordered(struct inode *inode)
 670{
 671        bool ret = __ceph_dir_is_complete_ordered(ceph_inode(inode));
 672        smp_rmb();
 673        return ret;
 674}
 675
 676/* find a specific frag @f */
 677extern struct ceph_inode_frag *__ceph_find_frag(struct ceph_inode_info *ci,
 678                                                u32 f);
 679
 680/*
 681 * choose fragment for value @v.  copy frag content to pfrag, if leaf
 682 * exists
 683 */
 684extern u32 ceph_choose_frag(struct ceph_inode_info *ci, u32 v,
 685                            struct ceph_inode_frag *pfrag,
 686                            int *found);
 687
 688static inline struct ceph_dentry_info *ceph_dentry(const struct dentry *dentry)
 689{
 690        return (struct ceph_dentry_info *)dentry->d_fsdata;
 691}
 692
 693/*
 694 * caps helpers
 695 */
 696static inline bool __ceph_is_any_real_caps(struct ceph_inode_info *ci)
 697{
 698        return !RB_EMPTY_ROOT(&ci->i_caps);
 699}
 700
 701extern int __ceph_caps_issued(struct ceph_inode_info *ci, int *implemented);
 702extern int __ceph_caps_issued_mask(struct ceph_inode_info *ci, int mask, int t);
 703extern int __ceph_caps_issued_mask_metric(struct ceph_inode_info *ci, int mask,
 704                                          int t);
 705extern int __ceph_caps_issued_other(struct ceph_inode_info *ci,
 706                                    struct ceph_cap *cap);
 707
 708static inline int ceph_caps_issued(struct ceph_inode_info *ci)
 709{
 710        int issued;
 711        spin_lock(&ci->i_ceph_lock);
 712        issued = __ceph_caps_issued(ci, NULL);
 713        spin_unlock(&ci->i_ceph_lock);
 714        return issued;
 715}
 716
 717static inline int ceph_caps_issued_mask_metric(struct ceph_inode_info *ci,
 718                                               int mask, int touch)
 719{
 720        int r;
 721        spin_lock(&ci->i_ceph_lock);
 722        r = __ceph_caps_issued_mask_metric(ci, mask, touch);
 723        spin_unlock(&ci->i_ceph_lock);
 724        return r;
 725}
 726
 727static inline int __ceph_caps_dirty(struct ceph_inode_info *ci)
 728{
 729        return ci->i_dirty_caps | ci->i_flushing_caps;
 730}
 731extern struct ceph_cap_flush *ceph_alloc_cap_flush(void);
 732extern void ceph_free_cap_flush(struct ceph_cap_flush *cf);
 733extern int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
 734                                  struct ceph_cap_flush **pcf);
 735
 736extern int __ceph_caps_revoking_other(struct ceph_inode_info *ci,
 737                                      struct ceph_cap *ocap, int mask);
 738extern int ceph_caps_revoking(struct ceph_inode_info *ci, int mask);
 739extern int __ceph_caps_used(struct ceph_inode_info *ci);
 740
 741static inline bool __ceph_is_file_opened(struct ceph_inode_info *ci)
 742{
 743        return ci->i_nr_by_mode[0];
 744}
 745extern int __ceph_caps_file_wanted(struct ceph_inode_info *ci);
 746extern int __ceph_caps_wanted(struct ceph_inode_info *ci);
 747
 748/* what the mds thinks we want */
 749extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
 750
 751extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 752extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
 753extern void ceph_adjust_caps_max_min(struct ceph_mds_client *mdsc,
 754                                     struct ceph_mount_options *fsopt);
 755extern int ceph_reserve_caps(struct ceph_mds_client *mdsc,
 756                             struct ceph_cap_reservation *ctx, int need);
 757extern void ceph_unreserve_caps(struct ceph_mds_client *mdsc,
 758                               struct ceph_cap_reservation *ctx);
 759extern void ceph_reservation_status(struct ceph_fs_client *client,
 760                                    int *total, int *avail, int *used,
 761                                    int *reserved, int *min);
 762
 763
 764
 765/*
 766 * we keep buffered readdir results attached to file->private_data
 767 */
 768#define CEPH_F_SYNC     1
 769#define CEPH_F_ATEND    2
 770
 771struct ceph_file_info {
 772        short fmode;     /* initialized on open */
 773        short flags;     /* CEPH_F_* */
 774
 775        spinlock_t rw_contexts_lock;
 776        struct list_head rw_contexts;
 777
 778        errseq_t meta_err;
 779        u32 filp_gen;
 780        atomic_t num_locks;
 781};
 782
 783struct ceph_dir_file_info {
 784        struct ceph_file_info file_info;
 785
 786        /* readdir: position within the dir */
 787        u32 frag;
 788        struct ceph_mds_request *last_readdir;
 789
 790        /* readdir: position within a frag */
 791        unsigned next_offset;  /* offset of next chunk (last_name's + 1) */
 792        char *last_name;       /* last entry in previous chunk */
 793        long long dir_release_count;
 794        long long dir_ordered_count;
 795        int readdir_cache_idx;
 796
 797        /* used for -o dirstat read() on directory thing */
 798        char *dir_info;
 799        int dir_info_len;
 800};
 801
 802struct ceph_rw_context {
 803        struct list_head list;
 804        struct task_struct *thread;
 805        int caps;
 806};
 807
 808#define CEPH_DEFINE_RW_CONTEXT(_name, _caps)    \
 809        struct ceph_rw_context _name = {        \
 810                .thread = current,              \
 811                .caps = _caps,                  \
 812        }
 813
 814static inline void ceph_add_rw_context(struct ceph_file_info *cf,
 815                                       struct ceph_rw_context *ctx)
 816{
 817        spin_lock(&cf->rw_contexts_lock);
 818        list_add(&ctx->list, &cf->rw_contexts);
 819        spin_unlock(&cf->rw_contexts_lock);
 820}
 821
 822static inline void ceph_del_rw_context(struct ceph_file_info *cf,
 823                                       struct ceph_rw_context *ctx)
 824{
 825        spin_lock(&cf->rw_contexts_lock);
 826        list_del(&ctx->list);
 827        spin_unlock(&cf->rw_contexts_lock);
 828}
 829
 830static inline struct ceph_rw_context*
 831ceph_find_rw_context(struct ceph_file_info *cf)
 832{
 833        struct ceph_rw_context *ctx, *found = NULL;
 834        spin_lock(&cf->rw_contexts_lock);
 835        list_for_each_entry(ctx, &cf->rw_contexts, list) {
 836                if (ctx->thread == current) {
 837                        found = ctx;
 838                        break;
 839                }
 840        }
 841        spin_unlock(&cf->rw_contexts_lock);
 842        return found;
 843}
 844
 845struct ceph_readdir_cache_control {
 846        struct page  *page;
 847        struct dentry **dentries;
 848        int index;
 849};
 850
 851/*
 852 * A "snap realm" describes a subset of the file hierarchy sharing
 853 * the same set of snapshots that apply to it.  The realms themselves
 854 * are organized into a hierarchy, such that children inherit (some of)
 855 * the snapshots of their parents.
 856 *
 857 * All inodes within the realm that have capabilities are linked into a
 858 * per-realm list.
 859 */
 860struct ceph_snap_realm {
 861        u64 ino;
 862        struct inode *inode;
 863        atomic_t nref;
 864        struct rb_node node;
 865
 866        u64 created, seq;
 867        u64 parent_ino;
 868        u64 parent_since;   /* snapid when our current parent became so */
 869
 870        u64 *prior_parent_snaps;      /* snaps inherited from any parents we */
 871        u32 num_prior_parent_snaps;   /*  had prior to parent_since */
 872        u64 *snaps;                   /* snaps specific to this realm */
 873        u32 num_snaps;
 874
 875        struct ceph_snap_realm *parent;
 876        struct list_head children;       /* list of child realms */
 877        struct list_head child_item;
 878
 879        struct list_head empty_item;     /* if i have ref==0 */
 880
 881        struct list_head dirty_item;     /* if realm needs new context */
 882
 883        /* the current set of snaps for this realm */
 884        struct ceph_snap_context *cached_context;
 885
 886        struct list_head inodes_with_caps;
 887        spinlock_t inodes_with_caps_lock;
 888};
 889
 890static inline int default_congestion_kb(void)
 891{
 892        int congestion_kb;
 893
 894        /*
 895         * Copied from NFS
 896         *
 897         * congestion size, scale with available memory.
 898         *
 899         *  64MB:    8192k
 900         * 128MB:   11585k
 901         * 256MB:   16384k
 902         * 512MB:   23170k
 903         *   1GB:   32768k
 904         *   2GB:   46340k
 905         *   4GB:   65536k
 906         *   8GB:   92681k
 907         *  16GB:  131072k
 908         *
 909         * This allows larger machines to have larger/more transfers.
 910         * Limit the default to 256M
 911         */
 912        congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);
 913        if (congestion_kb > 256*1024)
 914                congestion_kb = 256*1024;
 915
 916        return congestion_kb;
 917}
 918
 919
 920/* super.c */
 921extern int ceph_force_reconnect(struct super_block *sb);
 922/* snap.c */
 923struct ceph_snap_realm *ceph_lookup_snap_realm(struct ceph_mds_client *mdsc,
 924                                               u64 ino);
 925extern void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
 926                                struct ceph_snap_realm *realm);
 927extern void ceph_put_snap_realm(struct ceph_mds_client *mdsc,
 928                                struct ceph_snap_realm *realm);
 929extern int ceph_update_snap_trace(struct ceph_mds_client *m,
 930                                  void *p, void *e, bool deletion,
 931                                  struct ceph_snap_realm **realm_ret);
 932extern void ceph_handle_snap(struct ceph_mds_client *mdsc,
 933                             struct ceph_mds_session *session,
 934                             struct ceph_msg *msg);
 935extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 936                                  struct ceph_cap_snap *capsnap);
 937extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 938
 939extern struct ceph_snapid_map *ceph_get_snapid_map(struct ceph_mds_client *mdsc,
 940                                                   u64 snap);
 941extern void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
 942                                struct ceph_snapid_map *sm);
 943extern void ceph_trim_snapid_map(struct ceph_mds_client *mdsc);
 944extern void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc);
 945
 946
 947/*
 948 * a cap_snap is "pending" if it is still awaiting an in-progress
 949 * sync write (that may/may not still update size, mtime, etc.).
 950 */
 951static inline bool __ceph_have_pending_cap_snap(struct ceph_inode_info *ci)
 952{
 953        return !list_empty(&ci->i_cap_snaps) &&
 954               list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,
 955                               ci_item)->writing;
 956}
 957
 958/* inode.c */
 959struct ceph_mds_reply_info_in;
 960struct ceph_mds_reply_dirfrag;
 961
 962extern const struct inode_operations ceph_file_iops;
 963
 964extern struct inode *ceph_alloc_inode(struct super_block *sb);
 965extern void ceph_evict_inode(struct inode *inode);
 966extern void ceph_free_inode(struct inode *inode);
 967
 968extern struct inode *ceph_get_inode(struct super_block *sb,
 969                                    struct ceph_vino vino);
 970extern struct inode *ceph_get_snapdir(struct inode *parent);
 971extern int ceph_fill_file_size(struct inode *inode, int issued,
 972                               u32 truncate_seq, u64 truncate_size, u64 size);
 973extern void ceph_fill_file_time(struct inode *inode, int issued,
 974                                u64 time_warp_seq, struct timespec64 *ctime,
 975                                struct timespec64 *mtime,
 976                                struct timespec64 *atime);
 977extern int ceph_fill_inode(struct inode *inode, struct page *locked_page,
 978                    struct ceph_mds_reply_info_in *iinfo,
 979                    struct ceph_mds_reply_dirfrag *dirinfo,
 980                    struct ceph_mds_session *session, int cap_fmode,
 981                    struct ceph_cap_reservation *caps_reservation);
 982extern int ceph_fill_trace(struct super_block *sb,
 983                           struct ceph_mds_request *req);
 984extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 985                                    struct ceph_mds_session *session);
 986
 987extern int ceph_inode_holds_cap(struct inode *inode, int mask);
 988
 989extern bool ceph_inode_set_size(struct inode *inode, loff_t size);
 990extern void __ceph_do_pending_vmtruncate(struct inode *inode);
 991
 992void ceph_queue_inode_work(struct inode *inode, int work_bit);
 993
 994static inline void ceph_queue_vmtruncate(struct inode *inode)
 995{
 996        ceph_queue_inode_work(inode, CEPH_I_WORK_VMTRUNCATE);
 997}
 998
 999static inline void ceph_queue_invalidate(struct inode *inode)
1000{
1001        ceph_queue_inode_work(inode, CEPH_I_WORK_INVALIDATE_PAGES);
1002}
1003
1004static inline void ceph_queue_writeback(struct inode *inode)
1005{
1006        ceph_queue_inode_work(inode, CEPH_I_WORK_WRITEBACK);
1007}
1008
1009static inline void ceph_queue_check_caps(struct inode *inode)
1010{
1011        ceph_queue_inode_work(inode, CEPH_I_WORK_CHECK_CAPS);
1012}
1013
1014static inline void ceph_queue_flush_snaps(struct inode *inode)
1015{
1016        ceph_queue_inode_work(inode, CEPH_I_WORK_FLUSH_SNAPS);
1017}
1018
1019extern int __ceph_do_getattr(struct inode *inode, struct page *locked_page,
1020                             int mask, bool force);
1021static inline int ceph_do_getattr(struct inode *inode, int mask, bool force)
1022{
1023        return __ceph_do_getattr(inode, NULL, mask, force);
1024}
1025extern int ceph_permission(struct user_namespace *mnt_userns,
1026                           struct inode *inode, int mask);
1027extern int __ceph_setattr(struct inode *inode, struct iattr *attr);
1028extern int ceph_setattr(struct user_namespace *mnt_userns,
1029                        struct dentry *dentry, struct iattr *attr);
1030extern int ceph_getattr(struct user_namespace *mnt_userns,
1031                        const struct path *path, struct kstat *stat,
1032                        u32 request_mask, unsigned int flags);
1033
1034/* xattr.c */
1035int __ceph_setxattr(struct inode *, const char *, const void *, size_t, int);
1036ssize_t __ceph_getxattr(struct inode *, const char *, void *, size_t);
1037extern ssize_t ceph_listxattr(struct dentry *, char *, size_t);
1038extern struct ceph_buffer *__ceph_build_xattrs_blob(struct ceph_inode_info *ci);
1039extern void __ceph_destroy_xattrs(struct ceph_inode_info *ci);
1040extern const struct xattr_handler *ceph_xattr_handlers[];
1041
1042struct ceph_acl_sec_ctx {
1043#ifdef CONFIG_CEPH_FS_POSIX_ACL
1044        void *default_acl;
1045        void *acl;
1046#endif
1047#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1048        void *sec_ctx;
1049        u32 sec_ctxlen;
1050#endif
1051        struct ceph_pagelist *pagelist;
1052};
1053
1054#ifdef CONFIG_SECURITY
1055extern bool ceph_security_xattr_deadlock(struct inode *in);
1056extern bool ceph_security_xattr_wanted(struct inode *in);
1057#else
1058static inline bool ceph_security_xattr_deadlock(struct inode *in)
1059{
1060        return false;
1061}
1062static inline bool ceph_security_xattr_wanted(struct inode *in)
1063{
1064        return false;
1065}
1066#endif
1067
1068#ifdef CONFIG_CEPH_FS_SECURITY_LABEL
1069extern int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1070                                     struct ceph_acl_sec_ctx *ctx);
1071static inline void ceph_security_invalidate_secctx(struct inode *inode)
1072{
1073        security_inode_invalidate_secctx(inode);
1074}
1075#else
1076static inline int ceph_security_init_secctx(struct dentry *dentry, umode_t mode,
1077                                            struct ceph_acl_sec_ctx *ctx)
1078{
1079        return 0;
1080}
1081static inline void ceph_security_invalidate_secctx(struct inode *inode)
1082{
1083}
1084#endif
1085
1086void ceph_release_acl_sec_ctx(struct ceph_acl_sec_ctx *as_ctx);
1087
1088/* acl.c */
1089#ifdef CONFIG_CEPH_FS_POSIX_ACL
1090
1091struct posix_acl *ceph_get_acl(struct inode *, int);
1092int ceph_set_acl(struct user_namespace *mnt_userns,
1093                 struct inode *inode, struct posix_acl *acl, int type);
1094int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1095                       struct ceph_acl_sec_ctx *as_ctx);
1096void ceph_init_inode_acls(struct inode *inode,
1097                          struct ceph_acl_sec_ctx *as_ctx);
1098
1099static inline void ceph_forget_all_cached_acls(struct inode *inode)
1100{
1101       forget_all_cached_acls(inode);
1102}
1103
1104#else
1105
1106#define ceph_get_acl NULL
1107#define ceph_set_acl NULL
1108
1109static inline int ceph_pre_init_acls(struct inode *dir, umode_t *mode,
1110                                     struct ceph_acl_sec_ctx *as_ctx)
1111{
1112        return 0;
1113}
1114static inline void ceph_init_inode_acls(struct inode *inode,
1115                                        struct ceph_acl_sec_ctx *as_ctx)
1116{
1117}
1118static inline int ceph_acl_chmod(struct dentry *dentry, struct inode *inode)
1119{
1120        return 0;
1121}
1122
1123static inline void ceph_forget_all_cached_acls(struct inode *inode)
1124{
1125}
1126
1127#endif
1128
1129/* caps.c */
1130extern const char *ceph_cap_string(int c);
1131extern void ceph_handle_caps(struct ceph_mds_session *session,
1132                             struct ceph_msg *msg);
1133extern struct ceph_cap *ceph_get_cap(struct ceph_mds_client *mdsc,
1134                                     struct ceph_cap_reservation *ctx);
1135extern void ceph_add_cap(struct inode *inode,
1136                         struct ceph_mds_session *session, u64 cap_id,
1137                         unsigned issued, unsigned wanted,
1138                         unsigned cap, unsigned seq, u64 realmino, int flags,
1139                         struct ceph_cap **new_cap);
1140extern void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release);
1141extern void __ceph_remove_caps(struct ceph_inode_info *ci);
1142extern void ceph_put_cap(struct ceph_mds_client *mdsc,
1143                         struct ceph_cap *cap);
1144extern int ceph_is_any_caps(struct inode *inode);
1145
1146extern int ceph_write_inode(struct inode *inode, struct writeback_control *wbc);
1147extern int ceph_fsync(struct file *file, loff_t start, loff_t end,
1148                      int datasync);
1149extern void ceph_early_kick_flushing_caps(struct ceph_mds_client *mdsc,
1150                                          struct ceph_mds_session *session);
1151extern void ceph_kick_flushing_caps(struct ceph_mds_client *mdsc,
1152                                    struct ceph_mds_session *session);
1153void ceph_kick_flushing_inode_caps(struct ceph_mds_session *session,
1154                                   struct ceph_inode_info *ci);
1155extern struct ceph_cap *ceph_get_cap_for_mds(struct ceph_inode_info *ci,
1156                                             int mds);
1157extern void ceph_take_cap_refs(struct ceph_inode_info *ci, int caps,
1158                                bool snap_rwsem_locked);
1159extern void ceph_get_cap_refs(struct ceph_inode_info *ci, int caps);
1160extern void ceph_put_cap_refs(struct ceph_inode_info *ci, int had);
1161extern void ceph_put_cap_refs_async(struct ceph_inode_info *ci, int had);
1162extern void ceph_put_cap_refs_no_check_caps(struct ceph_inode_info *ci,
1163                                            int had);
1164extern void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
1165                                       struct ceph_snap_context *snapc);
1166extern void ceph_flush_snaps(struct ceph_inode_info *ci,
1167                             struct ceph_mds_session **psession);
1168extern bool __ceph_should_report_size(struct ceph_inode_info *ci);
1169extern void ceph_check_caps(struct ceph_inode_info *ci, int flags,
1170                            struct ceph_mds_session *session);
1171extern unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc);
1172extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
1173extern int  ceph_drop_caps_for_unlink(struct inode *inode);
1174extern int ceph_encode_inode_release(void **p, struct inode *inode,
1175                                     int mds, int drop, int unless, int force);
1176extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
1177                                      struct inode *dir,
1178                                      int mds, int drop, int unless);
1179
1180extern int ceph_get_caps(struct file *filp, int need, int want,
1181                         loff_t endoff, int *got);
1182extern int ceph_try_get_caps(struct inode *inode,
1183                             int need, int want, bool nonblock, int *got);
1184
1185/* for counting open files by mode */
1186extern void ceph_get_fmode(struct ceph_inode_info *ci, int mode, int count);
1187extern void ceph_put_fmode(struct ceph_inode_info *ci, int mode, int count);
1188extern void __ceph_touch_fmode(struct ceph_inode_info *ci,
1189                               struct ceph_mds_client *mdsc, int fmode);
1190
1191/* addr.c */
1192extern const struct address_space_operations ceph_aops;
1193extern int ceph_mmap(struct file *file, struct vm_area_struct *vma);
1194extern int ceph_uninline_data(struct file *filp, struct page *locked_page);
1195extern int ceph_pool_perm_check(struct inode *inode, int need);
1196extern void ceph_pool_perm_destroy(struct ceph_mds_client* mdsc);
1197
1198/* file.c */
1199extern const struct file_operations ceph_file_fops;
1200
1201extern int ceph_renew_caps(struct inode *inode, int fmode);
1202extern int ceph_open(struct inode *inode, struct file *file);
1203extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
1204                            struct file *file, unsigned flags, umode_t mode);
1205extern int ceph_release(struct inode *inode, struct file *filp);
1206extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1207                                  char *data, size_t len);
1208
1209/* dir.c */
1210extern const struct file_operations ceph_dir_fops;
1211extern const struct file_operations ceph_snapdir_fops;
1212extern const struct inode_operations ceph_dir_iops;
1213extern const struct inode_operations ceph_snapdir_iops;
1214extern const struct dentry_operations ceph_dentry_ops;
1215
1216extern loff_t ceph_make_fpos(unsigned high, unsigned off, bool hash_order);
1217extern int ceph_handle_notrace_create(struct inode *dir, struct dentry *dentry);
1218extern struct dentry *ceph_handle_snapdir(struct ceph_mds_request *req,
1219                               struct dentry *dentry);
1220extern struct dentry *ceph_finish_lookup(struct ceph_mds_request *req,
1221                                         struct dentry *dentry, int err);
1222
1223extern void __ceph_dentry_lease_touch(struct ceph_dentry_info *di);
1224extern void __ceph_dentry_dir_lease_touch(struct ceph_dentry_info *di);
1225extern void ceph_invalidate_dentry_lease(struct dentry *dentry);
1226extern int ceph_trim_dentries(struct ceph_mds_client *mdsc);
1227extern unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn);
1228extern void ceph_readdir_cache_release(struct ceph_readdir_cache_control *ctl);
1229
1230/* ioctl.c */
1231extern long ceph_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1232
1233/* export.c */
1234extern const struct export_operations ceph_export_ops;
1235struct inode *ceph_lookup_inode(struct super_block *sb, u64 ino);
1236
1237/* locks.c */
1238extern __init void ceph_flock_init(void);
1239extern int ceph_lock(struct file *file, int cmd, struct file_lock *fl);
1240extern int ceph_flock(struct file *file, int cmd, struct file_lock *fl);
1241extern void ceph_count_locks(struct inode *inode, int *p_num, int *f_num);
1242extern int ceph_encode_locks_to_buffer(struct inode *inode,
1243                                       struct ceph_filelock *flocks,
1244                                       int num_fcntl_locks,
1245                                       int num_flock_locks);
1246extern int ceph_locks_to_pagelist(struct ceph_filelock *flocks,
1247                                  struct ceph_pagelist *pagelist,
1248                                  int num_fcntl_locks, int num_flock_locks);
1249
1250/* debugfs.c */
1251extern void ceph_fs_debugfs_init(struct ceph_fs_client *client);
1252extern void ceph_fs_debugfs_cleanup(struct ceph_fs_client *client);
1253
1254/* quota.c */
1255static inline bool __ceph_has_any_quota(struct ceph_inode_info *ci)
1256{
1257        return ci->i_max_files || ci->i_max_bytes;
1258}
1259
1260extern void ceph_adjust_quota_realms_count(struct inode *inode, bool inc);
1261
1262static inline void __ceph_update_quota(struct ceph_inode_info *ci,
1263                                       u64 max_bytes, u64 max_files)
1264{
1265        bool had_quota, has_quota;
1266        had_quota = __ceph_has_any_quota(ci);
1267        ci->i_max_bytes = max_bytes;
1268        ci->i_max_files = max_files;
1269        has_quota = __ceph_has_any_quota(ci);
1270
1271        if (had_quota != has_quota)
1272                ceph_adjust_quota_realms_count(&ci->vfs_inode, has_quota);
1273}
1274
1275extern void ceph_handle_quota(struct ceph_mds_client *mdsc,
1276                              struct ceph_mds_session *session,
1277                              struct ceph_msg *msg);
1278extern bool ceph_quota_is_max_files_exceeded(struct inode *inode);
1279extern bool ceph_quota_is_same_realm(struct inode *old, struct inode *new);
1280extern bool ceph_quota_is_max_bytes_exceeded(struct inode *inode,
1281                                             loff_t newlen);
1282extern bool ceph_quota_is_max_bytes_approaching(struct inode *inode,
1283                                                loff_t newlen);
1284extern bool ceph_quota_update_statfs(struct ceph_fs_client *fsc,
1285                                     struct kstatfs *buf);
1286extern void ceph_cleanup_quotarealms_inodes(struct ceph_mds_client *mdsc);
1287
1288#endif /* _FS_CEPH_SUPER_H */
1289