linux/drivers/staging/lustre/lustre/llite/llite_internal.h
<<
>>
Prefs
   1/*
   2 * GPL HEADER START
   3 *
   4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5 *
   6 * This program is free software; you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License version 2 only,
   8 * as published by the Free Software Foundation.
   9 *
  10 * This program is distributed in the hope that it will be useful, but
  11 * WITHOUT ANY WARRANTY; without even the implied warranty of
  12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13 * General Public License version 2 for more details (a copy is included
  14 * in the LICENSE file that accompanied this code).
  15 *
  16 * You should have received a copy of the GNU General Public License
  17 * version 2 along with this program; If not, see
  18 * http://www.gnu.org/licenses/gpl-2.0.html
  19 *
  20 * GPL HEADER END
  21 */
  22/*
  23 * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  24 * Use is subject to license terms.
  25 *
  26 * Copyright (c) 2011, 2015, Intel Corporation.
  27 */
  28/*
  29 * This file is part of Lustre, http://www.lustre.org/
  30 * Lustre is a trademark of Sun Microsystems, Inc.
  31 */
  32
  33#ifndef LLITE_INTERNAL_H
  34#define LLITE_INTERNAL_H
  35#include "../include/lustre_debug.h"
  36#include "../include/lustre_ver.h"
  37#include "../include/lustre_disk.h"     /* for s2sbi */
  38#include "../include/lustre_linkea.h"
  39
  40/* for struct cl_lock_descr and struct cl_io */
  41#include "../include/lustre_patchless_compat.h"
  42#include "../include/lustre_compat.h"
  43#include "../include/cl_object.h"
  44#include "../include/lustre_lmv.h"
  45#include "../include/lustre_mdc.h"
  46#include "../include/lustre_intent.h"
  47#include <linux/compat.h>
  48#include <linux/namei.h>
  49#include <linux/xattr.h>
  50#include <linux/posix_acl_xattr.h>
  51#include "vvp_internal.h"
  52#include "range_lock.h"
  53
  54#ifndef FMODE_EXEC
  55#define FMODE_EXEC 0
  56#endif
  57
  58#ifndef VM_FAULT_RETRY
  59#define VM_FAULT_RETRY 0
  60#endif
  61
  62/** Only used on client-side for indicating the tail of dir hash/offset. */
  63#define LL_DIR_END_OFF    0x7fffffffffffffffULL
  64#define LL_DIR_END_OFF_32BIT    0x7fffffffUL
  65
  66/* 4UL * 1024 * 1024 */
  67#define LL_MAX_BLKSIZE_BITS 22
  68
  69#define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
  70#define LUSTRE_FPRIVATE(file) ((file)->private_data)
  71
  72struct ll_dentry_data {
  73        struct lookup_intent            *lld_it;
  74        unsigned int                    lld_sa_generation;
  75        unsigned int                    lld_invalid:1;
  76        unsigned int                    lld_nfs_dentry:1;
  77        struct rcu_head                 lld_rcu_head;
  78};
  79
  80#define ll_d2d(de) ((struct ll_dentry_data *)((de)->d_fsdata))
  81
  82#define LLI_INODE_MAGIC          0x111d0de5
  83#define LLI_INODE_DEAD            0xdeadd00d
  84
  85struct ll_getname_data {
  86        struct dir_context ctx;
  87        char        *lgd_name;      /* points to a buffer with NAME_MAX+1 size */
  88        struct lu_fid    lgd_fid;       /* target fid we are looking for */
  89        int           lgd_found;     /* inode matched? */
  90};
  91
  92struct ll_grouplock {
  93        struct lu_env   *lg_env;
  94        struct cl_io    *lg_io;
  95        struct cl_lock  *lg_lock;
  96        unsigned long    lg_gid;
  97};
  98
  99enum ll_file_flags {
 100        /* File data is modified. */
 101        LLIF_DATA_MODIFIED      = 0,
 102        /* File is being restored */
 103        LLIF_FILE_RESTORING     = 1,
 104        /* Xattr cache is attached to the file */
 105        LLIF_XATTR_CACHE        = 2,
 106};
 107
 108struct ll_inode_info {
 109        __u32                           lli_inode_magic;
 110
 111        spinlock_t                      lli_lock;
 112        unsigned long                   lli_flags;
 113        struct posix_acl                *lli_posix_acl;
 114
 115        /* identifying fields for both metadata and data stacks. */
 116        struct lu_fid              lli_fid;
 117        /* master inode fid for stripe directory */
 118        struct lu_fid              lli_pfid;
 119
 120        /* We need all three because every inode may be opened in different
 121         * modes
 122         */
 123        struct obd_client_handle       *lli_mds_read_och;
 124        struct obd_client_handle       *lli_mds_write_och;
 125        struct obd_client_handle       *lli_mds_exec_och;
 126        __u64                      lli_open_fd_read_count;
 127        __u64                      lli_open_fd_write_count;
 128        __u64                      lli_open_fd_exec_count;
 129        /* Protects access to och pointers and their usage counters */
 130        struct mutex                    lli_och_mutex;
 131
 132        struct inode                    lli_vfs_inode;
 133
 134        /* the most recent timestamps obtained from mds */
 135        s64                             lli_atime;
 136        s64                             lli_mtime;
 137        s64                             lli_ctime;
 138        spinlock_t                      lli_agl_lock;
 139
 140        /* Try to make the d::member and f::member are aligned. Before using
 141         * these members, make clear whether it is directory or not.
 142         */
 143        union {
 144                /* for directory */
 145                struct {
 146                        /* serialize normal readdir and statahead-readdir. */
 147                        struct mutex                    lli_readdir_mutex;
 148
 149                        /* metadata statahead */
 150                        /* since parent-child threads can share the same @file
 151                         * struct, "opendir_key" is the token when dir close for
 152                         * case of parent exit before child -- it is me should
 153                         * cleanup the dir readahead.
 154                         */
 155                        void                           *lli_opendir_key;
 156                        struct ll_statahead_info       *lli_sai;
 157                        /* protect statahead stuff. */
 158                        spinlock_t                      lli_sa_lock;
 159                        /* "opendir_pid" is the token when lookup/revalidate
 160                         * -- I am the owner of dir statahead.
 161                         */
 162                        pid_t                           lli_opendir_pid;
 163                        /* stat will try to access statahead entries or start
 164                         * statahead if this flag is set, and this flag will be
 165                         * set upon dir open, and cleared when dir is closed,
 166                         * statahead hit ratio is too low, or start statahead
 167                         * thread failed.
 168                         */
 169                        unsigned int                    lli_sa_enabled:1;
 170                        /* generation for statahead */
 171                        unsigned int                    lli_sa_generation;
 172                        /* directory stripe information */
 173                        struct lmv_stripe_md           *lli_lsm_md;
 174                        /* default directory stripe offset.  This is extracted
 175                         * from the "dmv" xattr in order to decide which MDT to
 176                         * create a subdirectory on.  The MDS itself fetches
 177                         * "dmv" and gets the rest of the default layout itself
 178                         * (count, hash, etc).
 179                         */
 180                        __u32                           lli_def_stripe_offset;
 181                };
 182
 183                /* for non-directory */
 184                struct {
 185                        struct mutex                    lli_size_mutex;
 186                        char                           *lli_symlink_name;
 187                        /*
 188                         * struct rw_semaphore {
 189                         *    signed long       count;     // align d.d_def_acl
 190                         *    spinlock_t        wait_lock; // align d.d_sa_lock
 191                         *    struct list_head wait_list;
 192                         * }
 193                         */
 194                        struct rw_semaphore             lli_trunc_sem;
 195                        struct range_lock_tree          lli_write_tree;
 196
 197                        struct rw_semaphore             lli_glimpse_sem;
 198                        unsigned long                   lli_glimpse_time;
 199                        struct list_head                lli_agl_list;
 200                        __u64                           lli_agl_index;
 201
 202                        /* for writepage() only to communicate to fsync */
 203                        int                             lli_async_rc;
 204
 205                        /*
 206                         * whenever a process try to read/write the file, the
 207                         * jobid of the process will be saved here, and it'll
 208                         * be packed into the write PRC when flush later.
 209                         *
 210                         * so the read/write statistics for jobid will not be
 211                         * accurate if the file is shared by different jobs.
 212                         */
 213                        char                            lli_jobid[LUSTRE_JOBID_SIZE];
 214                };
 215        };
 216
 217        /* XXX: For following frequent used members, although they maybe special
 218         *      used for non-directory object, it is some time-wasting to check
 219         *      whether the object is directory or not before using them. On the
 220         *      other hand, currently, sizeof(f) > sizeof(d), it cannot reduce
 221         *      the "ll_inode_info" size even if moving those members into u.f.
 222         *      So keep them out side.
 223         *
 224         *      In the future, if more members are added only for directory,
 225         *      some of the following members can be moved into u.f.
 226         */
 227        struct cl_object               *lli_clob;
 228
 229        /* mutex to request for layout lock exclusively. */
 230        struct mutex                    lli_layout_mutex;
 231        /* Layout version, protected by lli_layout_lock */
 232        __u32                           lli_layout_gen;
 233        spinlock_t                      lli_layout_lock;
 234
 235        struct rw_semaphore             lli_xattrs_list_rwsem;
 236        struct mutex                    lli_xattrs_enq_lock;
 237        struct list_head                lli_xattrs;/* ll_xattr_entry->xe_list */
 238};
 239
 240static inline __u32 ll_layout_version_get(struct ll_inode_info *lli)
 241{
 242        __u32 gen;
 243
 244        spin_lock(&lli->lli_layout_lock);
 245        gen = lli->lli_layout_gen;
 246        spin_unlock(&lli->lli_layout_lock);
 247
 248        return gen;
 249}
 250
 251static inline void ll_layout_version_set(struct ll_inode_info *lli, __u32 gen)
 252{
 253        spin_lock(&lli->lli_layout_lock);
 254        lli->lli_layout_gen = gen;
 255        spin_unlock(&lli->lli_layout_lock);
 256}
 257
 258int ll_xattr_cache_destroy(struct inode *inode);
 259
 260int ll_xattr_cache_get(struct inode *inode, const char *name,
 261                       char *buffer, size_t size, __u64 valid);
 262
 263int ll_init_security(struct dentry *dentry, struct inode *inode,
 264                     struct inode *dir);
 265
 266/*
 267 * Locking to guarantee consistency of non-atomic updates to long long i_size,
 268 * consistency between file size and KMS.
 269 *
 270 * Implemented by ->lli_size_mutex and ->lsm_lock, nested in that order.
 271 */
 272
 273void ll_inode_size_lock(struct inode *inode);
 274void ll_inode_size_unlock(struct inode *inode);
 275
 276/* FIXME: replace the name of this with LL_I to conform to kernel stuff */
 277/* static inline struct ll_inode_info *LL_I(struct inode *inode) */
 278static inline struct ll_inode_info *ll_i2info(struct inode *inode)
 279{
 280        return container_of(inode, struct ll_inode_info, lli_vfs_inode);
 281}
 282
 283/* default to about 64M of readahead on a given system. */
 284#define SBI_DEFAULT_READAHEAD_MAX       (64UL << (20 - PAGE_SHIFT))
 285
 286/* default to read-ahead full files smaller than 2MB on the second read */
 287#define SBI_DEFAULT_READAHEAD_WHOLE_MAX (2UL << (20 - PAGE_SHIFT))
 288
 289enum ra_stat {
 290        RA_STAT_HIT = 0,
 291        RA_STAT_MISS,
 292        RA_STAT_DISTANT_READPAGE,
 293        RA_STAT_MISS_IN_WINDOW,
 294        RA_STAT_FAILED_GRAB_PAGE,
 295        RA_STAT_FAILED_MATCH,
 296        RA_STAT_DISCARDED,
 297        RA_STAT_ZERO_LEN,
 298        RA_STAT_ZERO_WINDOW,
 299        RA_STAT_EOF,
 300        RA_STAT_MAX_IN_FLIGHT,
 301        RA_STAT_WRONG_GRAB_PAGE,
 302        RA_STAT_FAILED_REACH_END,
 303        _NR_RA_STAT,
 304};
 305
 306struct ll_ra_info {
 307        atomic_t              ra_cur_pages;
 308        unsigned long        ra_max_pages;
 309        unsigned long        ra_max_pages_per_file;
 310        unsigned long        ra_max_read_ahead_whole_pages;
 311};
 312
 313/* ra_io_arg will be filled in the beginning of ll_readahead with
 314 * ras_lock, then the following ll_read_ahead_pages will read RA
 315 * pages according to this arg, all the items in this structure are
 316 * counted by page index.
 317 */
 318struct ra_io_arg {
 319        unsigned long ria_start;  /* start offset of read-ahead*/
 320        unsigned long ria_end;    /* end offset of read-ahead*/
 321        unsigned long ria_reserved; /* reserved pages for read-ahead */
 322        unsigned long ria_end_min;  /* minimum end to cover current read */
 323        bool ria_eof;               /* reach end of file */
 324        /* If stride read pattern is detected, ria_stoff means where
 325         * stride read is started. Note: for normal read-ahead, the
 326         * value here is meaningless, and also it will not be accessed
 327         */
 328        pgoff_t ria_stoff;
 329        /* ria_length and ria_pages are the length and pages length in the
 330         * stride I/O mode. And they will also be used to check whether
 331         * it is stride I/O read-ahead in the read-ahead pages
 332         */
 333        unsigned long ria_length;
 334        unsigned long ria_pages;
 335};
 336
 337/* LL_HIST_MAX=32 causes an overflow */
 338#define LL_HIST_MAX 28
 339#define LL_HIST_START 12 /* buckets start at 2^12 = 4k */
 340#define LL_PROCESS_HIST_MAX 10
 341struct per_process_info {
 342        pid_t pid;
 343        struct obd_histogram pp_r_hist;
 344        struct obd_histogram pp_w_hist;
 345};
 346
 347/* pp_extents[LL_PROCESS_HIST_MAX] will hold the combined process info */
 348struct ll_rw_extents_info {
 349        struct per_process_info pp_extents[LL_PROCESS_HIST_MAX + 1];
 350};
 351
 352#define LL_OFFSET_HIST_MAX 100
 353struct ll_rw_process_info {
 354        pid_t                rw_pid;
 355        int                    rw_op;
 356        loff_t              rw_range_start;
 357        loff_t              rw_range_end;
 358        loff_t              rw_last_file_pos;
 359        loff_t              rw_offset;
 360        size_t              rw_smallest_extent;
 361        size_t              rw_largest_extent;
 362        struct ll_file_data      *rw_last_file;
 363};
 364
 365enum stats_track_type {
 366        STATS_TRACK_ALL = 0,  /* track all processes */
 367        STATS_TRACK_PID,      /* track process with this pid */
 368        STATS_TRACK_PPID,     /* track processes with this ppid */
 369        STATS_TRACK_GID,      /* track processes with this gid */
 370        STATS_TRACK_LAST,
 371};
 372
 373/* flags for sbi->ll_flags */
 374#define LL_SBI_NOLCK         0x01 /* DLM locking disabled (directio-only) */
 375#define LL_SBI_CHECKSUM   0x02 /* checksum each page as it's written */
 376#define LL_SBI_FLOCK         0x04
 377#define LL_SBI_USER_XATTR       0x08 /* support user xattr */
 378#define LL_SBI_ACL             0x10 /* support ACL */
 379/* LL_SBI_RMT_CLIENT             0x40    remote client */
 380#define LL_SBI_MDS_CAPA          0x80 /* support mds capa, obsolete */
 381#define LL_SBI_OSS_CAPA         0x100 /* support oss capa, obsolete */
 382#define LL_SBI_LOCALFLOCK       0x200 /* Local flocks support by kernel */
 383#define LL_SBI_LRU_RESIZE       0x400 /* lru resize support */
 384#define LL_SBI_LAZYSTATFS       0x800 /* lazystatfs mount option */
 385/*      LL_SBI_SOM_PREVIEW     0x1000    SOM preview mount option, obsolete */
 386#define LL_SBI_32BIT_API       0x2000 /* generate 32 bit inodes. */
 387#define LL_SBI_64BIT_HASH      0x4000 /* support 64-bits dir hash/offset */
 388#define LL_SBI_AGL_ENABLED     0x8000 /* enable agl */
 389#define LL_SBI_VERBOSE  0x10000 /* verbose mount/umount */
 390#define LL_SBI_LAYOUT_LOCK    0x20000 /* layout lock support */
 391#define LL_SBI_USER_FID2PATH  0x40000 /* allow fid2path by unprivileged users */
 392#define LL_SBI_XATTR_CACHE    0x80000 /* support for xattr cache */
 393#define LL_SBI_NOROOTSQUASH     0x100000 /* do not apply root squash */
 394#define LL_SBI_ALWAYS_PING      0x200000 /* always ping even if server
 395                                          * suppress_pings */
 396
 397#define LL_SBI_FLAGS {  \
 398        "nolck",        \
 399        "checksum",     \
 400        "flock",        \
 401        "user_xattr",   \
 402        "acl",          \
 403        "???",          \
 404        "???",          \
 405        "mds_capa",     \
 406        "oss_capa",     \
 407        "flock",        \
 408        "lru_resize",   \
 409        "lazy_statfs",  \
 410        "som",          \
 411        "32bit_api",    \
 412        "64bit_hash",   \
 413        "agl",          \
 414        "verbose",      \
 415        "layout",       \
 416        "user_fid2path",\
 417        "xattr_cache",  \
 418        "norootsquash", \
 419        "always_ping",  \
 420}
 421
 422/*
 423 * This is embedded into llite super-blocks to keep track of connect
 424 * flags (capabilities) supported by all imports given mount is
 425 * connected to.
 426 */
 427struct lustre_client_ocd {
 428        /*
 429         * This is conjunction of connect_flags across all imports
 430         * (LOVs) this mount is connected to. This field is updated by
 431         * cl_ocd_update() under ->lco_lock.
 432         */
 433        __u64                    lco_flags;
 434        struct mutex             lco_lock;
 435        struct obd_export       *lco_md_exp;
 436        struct obd_export       *lco_dt_exp;
 437};
 438
 439struct ll_sb_info {
 440        /* this protects pglist and ra_info.  It isn't safe to
 441         * grab from interrupt contexts
 442         */
 443        spinlock_t                ll_lock;
 444        spinlock_t                ll_pp_extent_lock; /* pp_extent entry*/
 445        spinlock_t                ll_process_lock; /* ll_rw_process_info */
 446        struct obd_uuid    ll_sb_uuid;
 447        struct obd_export       *ll_md_exp;
 448        struct obd_export       *ll_dt_exp;
 449        struct dentry           *ll_debugfs_entry;
 450        struct lu_fid        ll_root_fid; /* root object fid */
 451
 452        int                    ll_flags;
 453        unsigned int              ll_umounting:1,
 454                                  ll_xattr_cache_enabled:1,
 455                                  ll_client_common_fill_super_succeeded:1;
 456
 457        struct lustre_client_ocd  ll_lco;
 458
 459        struct lprocfs_stats     *ll_stats; /* lprocfs stats counter */
 460
 461        /*
 462         * Used to track "unstable" pages on a client, and maintain a
 463         * LRU list of clean pages. An "unstable" page is defined as
 464         * any page which is sent to a server as part of a bulk request,
 465         * but is uncommitted to stable storage.
 466         */
 467        struct cl_client_cache    *ll_cache;
 468
 469        struct lprocfs_stats     *ll_ra_stats;
 470
 471        struct ll_ra_info        ll_ra_info;
 472        unsigned int          ll_namelen;
 473        const struct file_operations    *ll_fop;
 474
 475        unsigned int              ll_md_brw_pages; /* readdir pages per RPC */
 476
 477        struct lu_site     *ll_site;
 478        struct cl_device         *ll_cl;
 479        /* Statistics */
 480        struct ll_rw_extents_info ll_rw_extents_info;
 481        int                    ll_extent_process_count;
 482        struct ll_rw_process_info ll_rw_process_info[LL_PROCESS_HIST_MAX];
 483        unsigned int          ll_offset_process_count;
 484        struct ll_rw_process_info ll_rw_offset_info[LL_OFFSET_HIST_MAX];
 485        unsigned int          ll_rw_offset_entry_count;
 486        int                    ll_stats_track_id;
 487        enum stats_track_type     ll_stats_track_type;
 488        int                    ll_rw_stats_on;
 489
 490        /* metadata stat-ahead */
 491        unsigned int          ll_sa_max;     /* max statahead RPCs */
 492        atomic_t                  ll_sa_total;   /* statahead thread started
 493                                                  * count
 494                                                  */
 495        atomic_t                  ll_sa_wrong;   /* statahead thread stopped for
 496                                                  * low hit ratio
 497                                                  */
 498        atomic_t                ll_sa_running;  /* running statahead thread
 499                                                 * count
 500                                                 */
 501        atomic_t                  ll_agl_total;  /* AGL thread started count */
 502
 503        dev_t                     ll_sdev_orig; /* save s_dev before assign for
 504                                                 * clustered nfs
 505                                                 */
 506        /* root squash */
 507        struct root_squash_info   ll_squash;
 508        struct path              ll_mnt;
 509
 510        __kernel_fsid_t           ll_fsid;
 511        struct kobject           ll_kobj; /* sysfs object */
 512        struct super_block      *ll_sb; /* struct super_block (for sysfs code)*/
 513        struct completion        ll_kobj_unregister;
 514};
 515
 516/*
 517 * per file-descriptor read-ahead data.
 518 */
 519struct ll_readahead_state {
 520        spinlock_t  ras_lock;
 521        /*
 522         * index of the last page that read(2) needed and that wasn't in the
 523         * cache. Used by ras_update() to detect seeks.
 524         *
 525         * XXX nikita: if access seeks into cached region, Lustre doesn't see
 526         * this.
 527         */
 528        unsigned long   ras_last_readpage;
 529        /*
 530         * number of pages read after last read-ahead window reset. As window
 531         * is reset on each seek, this is effectively a number of consecutive
 532         * accesses. Maybe ->ras_accessed_in_window is better name.
 533         *
 534         * XXX nikita: window is also reset (by ras_update()) when Lustre
 535         * believes that memory pressure evicts read-ahead pages. In that
 536         * case, it probably doesn't make sense to expand window to
 537         * PTLRPC_MAX_BRW_PAGES on the third access.
 538         */
 539        unsigned long   ras_consecutive_pages;
 540        /*
 541         * number of read requests after the last read-ahead window reset
 542         * As window is reset on each seek, this is effectively the number
 543         * on consecutive read request and is used to trigger read-ahead.
 544         */
 545        unsigned long   ras_consecutive_requests;
 546        /*
 547         * Parameters of current read-ahead window. Handled by
 548         * ras_update(). On the initial access to the file or after a seek,
 549         * window is reset to 0. After 3 consecutive accesses, window is
 550         * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
 551         * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
 552         */
 553        unsigned long   ras_window_start, ras_window_len;
 554        /*
 555         * Optimal RPC size. It decides how many pages will be sent
 556         * for each read-ahead.
 557         */
 558        unsigned long   ras_rpc_size;
 559        /*
 560         * Where next read-ahead should start at. This lies within read-ahead
 561         * window. Read-ahead window is read in pieces rather than at once
 562         * because: 1. lustre limits total number of pages under read-ahead by
 563         * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
 564         * not covered by DLM lock.
 565         */
 566        unsigned long   ras_next_readahead;
 567        /*
 568         * Total number of ll_file_read requests issued, reads originating
 569         * due to mmap are not counted in this total.  This value is used to
 570         * trigger full file read-ahead after multiple reads to a small file.
 571         */
 572        unsigned long   ras_requests;
 573        /*
 574         * Page index with respect to the current request, these value
 575         * will not be accurate when dealing with reads issued via mmap.
 576         */
 577        unsigned long   ras_request_index;
 578        /*
 579         * The following 3 items are used for detecting the stride I/O
 580         * mode.
 581         * In stride I/O mode,
 582         * ...............|-----data-----|****gap*****|--------|******|....
 583         *    offset      |-stride_pages-|-stride_gap-|
 584         * ras_stride_offset = offset;
 585         * ras_stride_length = stride_pages + stride_gap;
 586         * ras_stride_pages = stride_pages;
 587         * Note: all these three items are counted by pages.
 588         */
 589        unsigned long   ras_stride_length;
 590        unsigned long   ras_stride_pages;
 591        pgoff_t  ras_stride_offset;
 592        /*
 593         * number of consecutive stride request count, and it is similar as
 594         * ras_consecutive_requests, but used for stride I/O mode.
 595         * Note: only more than 2 consecutive stride request are detected,
 596         * stride read-ahead will be enable
 597         */
 598        unsigned long   ras_consecutive_stride_requests;
 599};
 600
 601extern struct kmem_cache *ll_file_data_slab;
 602struct lustre_handle;
 603struct ll_file_data {
 604        struct ll_readahead_state fd_ras;
 605        struct ll_grouplock fd_grouplock;
 606        __u64 lfd_pos;
 607        __u32 fd_flags;
 608        fmode_t fd_omode;
 609        /* openhandle if lease exists for this file.
 610         * Borrow lli->lli_och_mutex to protect assignment
 611         */
 612        struct obd_client_handle *fd_lease_och;
 613        struct obd_client_handle *fd_och;
 614        struct file *fd_file;
 615        /* Indicate whether need to report failure when close.
 616         * true: failure is known, not report again.
 617         * false: unknown failure, should report.
 618         */
 619        bool fd_write_failed;
 620        rwlock_t fd_lock; /* protect lcc list */
 621        struct list_head fd_lccs; /* list of ll_cl_context */
 622};
 623
 624extern struct dentry *llite_root;
 625extern struct kset *llite_kset;
 626
 627static inline struct inode *ll_info2i(struct ll_inode_info *lli)
 628{
 629        return &lli->lli_vfs_inode;
 630}
 631
 632__u32 ll_i2suppgid(struct inode *i);
 633void ll_i2gids(__u32 *suppgids, struct inode *i1, struct inode *i2);
 634
 635static inline int ll_need_32bit_api(struct ll_sb_info *sbi)
 636{
 637#if BITS_PER_LONG == 32
 638        return 1;
 639#elif defined(CONFIG_COMPAT)
 640        return unlikely(in_compat_syscall() || (sbi->ll_flags & LL_SBI_32BIT_API));
 641#else
 642        return unlikely(sbi->ll_flags & LL_SBI_32BIT_API);
 643#endif
 644}
 645
 646void ll_ras_enter(struct file *f);
 647
 648/* llite/lcommon_misc.c */
 649int cl_init_ea_size(struct obd_export *md_exp, struct obd_export *dt_exp);
 650int cl_ocd_update(struct obd_device *host,
 651                  struct obd_device *watched,
 652                  enum obd_notify_event ev, void *owner, void *data);
 653int cl_get_grouplock(struct cl_object *obj, unsigned long gid, int nonblock,
 654                     struct ll_grouplock *cg);
 655void cl_put_grouplock(struct ll_grouplock *cg);
 656
 657/* llite/lproc_llite.c */
 658int ldebugfs_register_mountpoint(struct dentry *parent,
 659                                 struct super_block *sb, char *osc, char *mdc);
 660void ldebugfs_unregister_mountpoint(struct ll_sb_info *sbi);
 661void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count);
 662void lprocfs_llite_init_vars(struct lprocfs_static_vars *lvars);
 663void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
 664                       struct ll_file_data *file, loff_t pos,
 665                       size_t count, int rw);
 666
 667enum {
 668        LPROC_LL_DIRTY_HITS,
 669        LPROC_LL_DIRTY_MISSES,
 670        LPROC_LL_READ_BYTES,
 671        LPROC_LL_WRITE_BYTES,
 672        LPROC_LL_BRW_READ,
 673        LPROC_LL_BRW_WRITE,
 674        LPROC_LL_IOCTL,
 675        LPROC_LL_OPEN,
 676        LPROC_LL_RELEASE,
 677        LPROC_LL_MAP,
 678        LPROC_LL_LLSEEK,
 679        LPROC_LL_FSYNC,
 680        LPROC_LL_READDIR,
 681        LPROC_LL_SETATTR,
 682        LPROC_LL_TRUNC,
 683        LPROC_LL_FLOCK,
 684        LPROC_LL_GETATTR,
 685        LPROC_LL_CREATE,
 686        LPROC_LL_LINK,
 687        LPROC_LL_UNLINK,
 688        LPROC_LL_SYMLINK,
 689        LPROC_LL_MKDIR,
 690        LPROC_LL_RMDIR,
 691        LPROC_LL_MKNOD,
 692        LPROC_LL_RENAME,
 693        LPROC_LL_STAFS,
 694        LPROC_LL_ALLOC_INODE,
 695        LPROC_LL_SETXATTR,
 696        LPROC_LL_GETXATTR,
 697        LPROC_LL_GETXATTR_HITS,
 698        LPROC_LL_LISTXATTR,
 699        LPROC_LL_REMOVEXATTR,
 700        LPROC_LL_INODE_PERM,
 701        LPROC_LL_FILE_OPCODES
 702};
 703
 704/* llite/dir.c */
 705extern const struct file_operations ll_dir_operations;
 706extern const struct inode_operations ll_dir_inode_operations;
 707int ll_dir_read(struct inode *inode, __u64 *ppos, struct md_op_data *op_data,
 708                struct dir_context *ctx);
 709int ll_get_mdt_idx(struct inode *inode);
 710int ll_get_mdt_idx_by_fid(struct ll_sb_info *sbi, const struct lu_fid *fid);
 711struct page *ll_get_dir_page(struct inode *dir, struct md_op_data *op_data,
 712                             __u64 offset);
 713void ll_release_page(struct inode *inode, struct page *page, bool remove);
 714
 715/* llite/namei.c */
 716extern const struct inode_operations ll_special_inode_operations;
 717
 718struct inode *ll_iget(struct super_block *sb, ino_t hash,
 719                      struct lustre_md *lic);
 720int ll_test_inode_by_fid(struct inode *inode, void *opaque);
 721int ll_md_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
 722                       void *data, int flag);
 723struct dentry *ll_splice_alias(struct inode *inode, struct dentry *de);
 724void ll_update_times(struct ptlrpc_request *request, struct inode *inode);
 725
 726/* llite/rw.c */
 727int ll_writepage(struct page *page, struct writeback_control *wbc);
 728int ll_writepages(struct address_space *mapping, struct writeback_control *wbc);
 729int ll_readpage(struct file *file, struct page *page);
 730void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
 731int vvp_io_write_commit(const struct lu_env *env, struct cl_io *io);
 732struct ll_cl_context *ll_cl_find(struct file *file);
 733void ll_cl_add(struct file *file, const struct lu_env *env, struct cl_io *io);
 734void ll_cl_remove(struct file *file, const struct lu_env *env);
 735
 736extern const struct address_space_operations ll_aops;
 737
 738/* llite/file.c */
 739extern const struct file_operations ll_file_operations;
 740extern const struct file_operations ll_file_operations_flock;
 741extern const struct file_operations ll_file_operations_noflock;
 742extern const struct inode_operations ll_file_inode_operations;
 743int ll_have_md_lock(struct inode *inode, __u64 *bits,
 744                    enum ldlm_mode l_req_mode);
 745enum ldlm_mode ll_take_md_lock(struct inode *inode, __u64 bits,
 746                               struct lustre_handle *lockh, __u64 flags,
 747                               enum ldlm_mode mode);
 748int ll_file_open(struct inode *inode, struct file *file);
 749int ll_file_release(struct inode *inode, struct file *file);
 750int ll_release_openhandle(struct inode *inode, struct lookup_intent *it);
 751int ll_md_real_close(struct inode *inode, fmode_t fmode);
 752int ll_getattr(const struct path *path, struct kstat *stat,
 753               u32 request_mask, unsigned int flags);
 754struct posix_acl *ll_get_acl(struct inode *inode, int type);
 755int ll_migrate(struct inode *parent, struct file *file, int mdtidx,
 756               const char *name, int namelen);
 757int ll_get_fid_by_name(struct inode *parent, const char *name,
 758                       int namelen, struct lu_fid *fid, struct inode **inode);
 759int ll_inode_permission(struct inode *inode, int mask);
 760
 761int ll_lov_setstripe_ea_info(struct inode *inode, struct dentry *dentry,
 762                             __u64 flags, struct lov_user_md *lum,
 763                             int lum_size);
 764int ll_lov_getstripe_ea_info(struct inode *inode, const char *filename,
 765                             struct lov_mds_md **lmm, int *lmm_size,
 766                             struct ptlrpc_request **request);
 767int ll_dir_setstripe(struct inode *inode, struct lov_user_md *lump,
 768                     int set_default);
 769int ll_dir_getstripe(struct inode *inode, void **lmmp, int *lmm_size,
 770                     struct ptlrpc_request **request, u64 valid);
 771int ll_fsync(struct file *file, loff_t start, loff_t end, int data);
 772int ll_merge_attr(const struct lu_env *env, struct inode *inode);
 773int ll_fid2path(struct inode *inode, void __user *arg);
 774int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
 775int ll_hsm_release(struct inode *inode);
 776int ll_hsm_state_set(struct inode *inode, struct hsm_state_set *hss);
 777
 778/* llite/dcache.c */
 779
 780extern const struct dentry_operations ll_d_ops;
 781void ll_intent_drop_lock(struct lookup_intent *it);
 782void ll_intent_release(struct lookup_intent *it);
 783void ll_invalidate_aliases(struct inode *inode);
 784void ll_lookup_finish_locks(struct lookup_intent *it, struct inode *inode);
 785int ll_revalidate_it_finish(struct ptlrpc_request *request,
 786                            struct lookup_intent *it, struct inode *inode);
 787
 788/* llite/llite_lib.c */
 789extern struct super_operations lustre_super_operations;
 790
 791void ll_lli_init(struct ll_inode_info *lli);
 792int ll_fill_super(struct super_block *sb, struct vfsmount *mnt);
 793void ll_put_super(struct super_block *sb);
 794void ll_kill_super(struct super_block *sb);
 795struct inode *ll_inode_from_resource_lock(struct ldlm_lock *lock);
 796void ll_dir_clear_lsm_md(struct inode *inode);
 797void ll_clear_inode(struct inode *inode);
 798int ll_setattr_raw(struct dentry *dentry, struct iattr *attr, bool hsm_import);
 799int ll_setattr(struct dentry *de, struct iattr *attr);
 800int ll_statfs(struct dentry *de, struct kstatfs *sfs);
 801int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
 802                       __u64 max_age, __u32 flags);
 803int ll_update_inode(struct inode *inode, struct lustre_md *md);
 804int ll_read_inode2(struct inode *inode, void *opaque);
 805void ll_delete_inode(struct inode *inode);
 806int ll_iocontrol(struct inode *inode, struct file *file,
 807                 unsigned int cmd, unsigned long arg);
 808int ll_flush_ctx(struct inode *inode);
 809void ll_umount_begin(struct super_block *sb);
 810int ll_remount_fs(struct super_block *sb, int *flags, char *data);
 811int ll_show_options(struct seq_file *seq, struct dentry *dentry);
 812void ll_dirty_page_discard_warn(struct page *page, int ioret);
 813int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
 814                  struct super_block *sb, struct lookup_intent *it);
 815int ll_obd_statfs(struct inode *inode, void __user *arg);
 816int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
 817int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
 818int ll_set_default_mdsize(struct ll_sb_info *sbi, int default_mdsize);
 819int ll_process_config(struct lustre_cfg *lcfg);
 820
 821enum {
 822        LUSTRE_OPC_MKDIR        = 0,
 823        LUSTRE_OPC_SYMLINK      = 1,
 824        LUSTRE_OPC_MKNOD        = 2,
 825        LUSTRE_OPC_CREATE       = 3,
 826        LUSTRE_OPC_ANY          = 5,
 827};
 828
 829struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
 830                                      struct inode *i1, struct inode *i2,
 831                                      const char *name, size_t namelen,
 832                                      u32 mode, __u32 opc, void *data);
 833void ll_finish_md_op_data(struct md_op_data *op_data);
 834int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
 835char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
 836void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
 837void ll_open_cleanup(struct super_block *sb, struct ptlrpc_request *open_req);
 838ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
 839                        struct lov_user_md **kbuf);
 840
 841/* Compute expected user md size when passing in a md from user space */
 842static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
 843{
 844        switch (lum->lmm_magic) {
 845        case LOV_USER_MAGIC_V1:
 846                return sizeof(struct lov_user_md_v1);
 847        case LOV_USER_MAGIC_V3:
 848                return sizeof(struct lov_user_md_v3);
 849        case LOV_USER_MAGIC_SPECIFIC:
 850                if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
 851                        return -EINVAL;
 852
 853                return lov_user_md_size(lum->lmm_stripe_count,
 854                                        LOV_USER_MAGIC_SPECIFIC);
 855        }
 856        return -EINVAL;
 857}
 858
 859/* llite/llite_nfs.c */
 860extern const struct export_operations lustre_export_operations;
 861__u32 get_uuid2int(const char *name, int len);
 862void get_uuid2fsid(const char *name, int len, __kernel_fsid_t *fsid);
 863struct inode *search_inode_for_lustre(struct super_block *sb,
 864                                      const struct lu_fid *fid);
 865int ll_dir_get_parent_fid(struct inode *dir, struct lu_fid *parent_fid);
 866
 867/* llite/symlink.c */
 868extern const struct inode_operations ll_fast_symlink_inode_operations;
 869
 870/**
 871 * IO arguments for various VFS I/O interfaces.
 872 */
 873struct vvp_io_args {
 874        /** normal/splice */
 875        union {
 876                struct {
 877                        struct kiocb      *via_iocb;
 878                        struct iov_iter   *via_iter;
 879                } normal;
 880        } u;
 881};
 882
 883struct ll_cl_context {
 884        struct list_head         lcc_list;
 885        void       *lcc_cookie;
 886        const struct lu_env     *lcc_env;
 887        struct cl_io   *lcc_io;
 888        struct cl_page *lcc_page;
 889};
 890
 891struct ll_thread_info {
 892        struct vvp_io_args   lti_args;
 893        struct ra_io_arg     lti_ria;
 894        struct ll_cl_context lti_io_ctx;
 895};
 896
 897extern struct lu_context_key ll_thread_key;
 898static inline struct ll_thread_info *ll_env_info(const struct lu_env *env)
 899{
 900        struct ll_thread_info *lti;
 901
 902        lti = lu_context_key_get(&env->le_ctx, &ll_thread_key);
 903        LASSERT(lti);
 904        return lti;
 905}
 906
 907static inline struct vvp_io_args *ll_env_args(const struct lu_env *env)
 908{
 909        return &ll_env_info(env)->lti_args;
 910}
 911
 912/* llite/llite_mmap.c */
 913
 914int ll_teardown_mmaps(struct address_space *mapping, __u64 first, __u64 last);
 915int ll_file_mmap(struct file *file, struct vm_area_struct *vma);
 916void policy_from_vma(union ldlm_policy_data *policy, struct vm_area_struct *vma,
 917                     unsigned long addr, size_t count);
 918struct vm_area_struct *our_vma(struct mm_struct *mm, unsigned long addr,
 919                               size_t count);
 920
 921static inline void ll_invalidate_page(struct page *vmpage)
 922{
 923        struct address_space *mapping = vmpage->mapping;
 924        loff_t offset = vmpage->index << PAGE_SHIFT;
 925
 926        LASSERT(PageLocked(vmpage));
 927        if (!mapping)
 928                return;
 929
 930        /*
 931         * truncate_complete_page() calls
 932         * a_ops->invalidatepage()->cl_page_delete()->vvp_page_delete().
 933         */
 934        ll_teardown_mmaps(mapping, offset, offset + PAGE_SIZE);
 935        truncate_complete_page(mapping, vmpage);
 936}
 937
 938#define    ll_s2sbi(sb) (s2lsi(sb)->lsi_llsbi)
 939
 940/* don't need an addref as the sb_info should be holding one */
 941static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
 942{
 943        return ll_s2sbi(sb)->ll_dt_exp;
 944}
 945
 946/* don't need an addref as the sb_info should be holding one */
 947static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
 948{
 949        return ll_s2sbi(sb)->ll_md_exp;
 950}
 951
 952static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi)
 953{
 954        struct obd_device *obd = sbi->ll_md_exp->exp_obd;
 955
 956        if (!obd)
 957                LBUG();
 958        return &obd->u.cli;
 959}
 960
 961/* FIXME: replace the name of this with LL_SB to conform to kernel stuff */
 962static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
 963{
 964        return ll_s2sbi(inode->i_sb);
 965}
 966
 967static inline struct obd_export *ll_i2dtexp(struct inode *inode)
 968{
 969        return ll_s2dtexp(inode->i_sb);
 970}
 971
 972static inline struct obd_export *ll_i2mdexp(struct inode *inode)
 973{
 974        return ll_s2mdexp(inode->i_sb);
 975}
 976
 977static inline struct lu_fid *ll_inode2fid(struct inode *inode)
 978{
 979        struct lu_fid *fid;
 980
 981        LASSERT(inode);
 982        fid = &ll_i2info(inode)->lli_fid;
 983
 984        return fid;
 985}
 986
 987static inline loff_t ll_file_maxbytes(struct inode *inode)
 988{
 989        struct cl_object *obj = ll_i2info(inode)->lli_clob;
 990
 991        if (!obj)
 992                return MAX_LFS_FILESIZE;
 993
 994        return min_t(loff_t, cl_object_maxbytes(obj), MAX_LFS_FILESIZE);
 995}
 996
 997/* llite/xattr.c */
 998extern const struct xattr_handler *ll_xattr_handlers[];
 999
1000#define XATTR_USER_T            1
1001#define XATTR_TRUSTED_T         2
1002#define XATTR_SECURITY_T        3
1003#define XATTR_ACL_ACCESS_T      4
1004#define XATTR_ACL_DEFAULT_T     5
1005#define XATTR_LUSTRE_T          6
1006#define XATTR_OTHER_T           7
1007
1008ssize_t ll_listxattr(struct dentry *dentry, char *buffer, size_t size);
1009int ll_xattr_list(struct inode *inode, const char *name, int type,
1010                  void *buffer, size_t size, __u64 valid);
1011const struct xattr_handler *get_xattr_type(const char *name);
1012
1013/**
1014 * Common IO arguments for various VFS I/O interfaces.
1015 */
1016int cl_sb_init(struct super_block *sb);
1017int cl_sb_fini(struct super_block *sb);
1018
1019enum ras_update_flags {
1020        LL_RAS_HIT  = 0x1,
1021        LL_RAS_MMAP = 0x2
1022};
1023void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len);
1024void ll_ra_stats_inc(struct inode *inode, enum ra_stat which);
1025
1026/* statahead.c */
1027#define LL_SA_RPC_MIN      2
1028#define LL_SA_RPC_DEF      32
1029#define LL_SA_RPC_MAX      8192
1030
1031#define LL_SA_CACHE_BIT  5
1032#define LL_SA_CACHE_SIZE        (1 << LL_SA_CACHE_BIT)
1033#define LL_SA_CACHE_MASK        (LL_SA_CACHE_SIZE - 1)
1034
1035/* per inode struct, for dir only */
1036struct ll_statahead_info {
1037        struct dentry      *sai_dentry;
1038        atomic_t            sai_refcount;   /* when access this struct, hold
1039                                             * refcount
1040                                             */
1041        unsigned int        sai_max;    /* max ahead of lookup */
1042        __u64              sai_sent;       /* stat requests sent count */
1043        __u64              sai_replied;    /* stat requests which received
1044                                            * reply
1045                                            */
1046        __u64              sai_index;      /* index of statahead entry */
1047        __u64              sai_index_wait; /* index of entry which is the
1048                                            * caller is waiting for
1049                                            */
1050        __u64              sai_hit;     /* hit count */
1051        __u64              sai_miss;       /* miss count:
1052                                            * for "ls -al" case, it includes
1053                                            * hidden dentry miss;
1054                                            * for "ls -l" case, it does not
1055                                            * include hidden dentry miss.
1056                                            * "sai_miss_hidden" is used for
1057                                            * the later case.
1058                                            */
1059        unsigned int        sai_consecutive_miss; /* consecutive miss */
1060        unsigned int        sai_miss_hidden;/* "ls -al", but first dentry
1061                                             * is not a hidden one
1062                                             */
1063        unsigned int        sai_skip_hidden;/* skipped hidden dentry count */
1064        unsigned int        sai_ls_all:1,   /* "ls -al", do stat-ahead for
1065                                             * hidden entries
1066                                             */
1067                                sai_agl_valid:1,/* AGL is valid for the dir */
1068                                sai_in_readpage:1;/* statahead is in readdir() */
1069        wait_queue_head_t       sai_waitq;      /* stat-ahead wait queue */
1070        struct ptlrpc_thread    sai_thread;     /* stat-ahead thread */
1071        struct ptlrpc_thread    sai_agl_thread; /* AGL thread */
1072        struct list_head        sai_interim_entries; /* entries which got async
1073                                                      * stat reply, but not
1074                                                      * instantiated
1075                                                      */
1076        struct list_head        sai_entries;    /* completed entries */
1077        struct list_head        sai_agls;       /* AGLs to be sent */
1078        struct list_head        sai_cache[LL_SA_CACHE_SIZE];
1079        spinlock_t              sai_cache_lock[LL_SA_CACHE_SIZE];
1080        atomic_t                sai_cache_count; /* entry count in cache */
1081};
1082
1083int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
1084void ll_authorize_statahead(struct inode *dir, void *key);
1085void ll_deauthorize_statahead(struct inode *dir, void *key);
1086
1087blkcnt_t dirty_cnt(struct inode *inode);
1088
1089int cl_glimpse_size0(struct inode *inode, int agl);
1090int cl_glimpse_lock(const struct lu_env *env, struct cl_io *io,
1091                    struct inode *inode, struct cl_object *clob, int agl);
1092
1093static inline int cl_glimpse_size(struct inode *inode)
1094{
1095        return cl_glimpse_size0(inode, 0);
1096}
1097
1098static inline int cl_agl(struct inode *inode)
1099{
1100        return cl_glimpse_size0(inode, 1);
1101}
1102
1103static inline int ll_glimpse_size(struct inode *inode)
1104{
1105        struct ll_inode_info *lli = ll_i2info(inode);
1106        int rc;
1107
1108        down_read(&lli->lli_glimpse_sem);
1109        rc = cl_glimpse_size(inode);
1110        lli->lli_glimpse_time = cfs_time_current();
1111        up_read(&lli->lli_glimpse_sem);
1112        return rc;
1113}
1114
1115/*
1116 * dentry may statahead when statahead is enabled and current process has opened
1117 * parent directory, and this dentry hasn't accessed statahead cache before
1118 */
1119static inline bool
1120dentry_may_statahead(struct inode *dir, struct dentry *dentry)
1121{
1122        struct ll_inode_info  *lli;
1123        struct ll_dentry_data *ldd;
1124
1125        if (ll_i2sbi(dir)->ll_sa_max == 0)
1126                return false;
1127
1128        lli = ll_i2info(dir);
1129
1130        /*
1131         * statahead is not allowed for this dir, there may be three causes:
1132         * 1. dir is not opened.
1133         * 2. statahead hit ratio is too low.
1134         * 3. previous stat started statahead thread failed.
1135         */
1136        if (!lli->lli_sa_enabled)
1137                return false;
1138
1139        /* not the same process, don't statahead */
1140        if (lli->lli_opendir_pid != current_pid())
1141                return false;
1142
1143        /*
1144         * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
1145         * multiple times, eg. for 'getattr', 'getxattr' and etc.
1146         * For patchless client, lookup intent is not accurate, which may
1147         * misguide statahead. For example:
1148         * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
1149         * have the same intent -- IT_GETATTR, while one dentry should access
1150         * statahead cache once, otherwise statahead windows is messed up.
1151         * The solution is as following:
1152         * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
1153         * IT_GETATTR for the first time, and subsequent IT_GETATTR will
1154         * bypass interacting with statahead cache by checking
1155         * 'lld_sa_generation == lli->lli_sa_generation'.
1156         */
1157        ldd = ll_d2d(dentry);
1158        if (ldd->lld_sa_generation == lli->lli_sa_generation)
1159                return false;
1160
1161        return true;
1162}
1163
1164/* llite ioctl register support routine */
1165enum llioc_iter {
1166        LLIOC_CONT = 0,
1167        LLIOC_STOP
1168};
1169
1170#define LLIOC_MAX_CMD      256
1171
1172/*
1173 * Rules to write a callback function:
1174 *
1175 * Parameters:
1176 *  @magic: Dynamic ioctl call routine will feed this value with the pointer
1177 *      returned to ll_iocontrol_register.  Callback functions should use this
1178 *      data to check the potential collasion of ioctl cmd. If collasion is
1179 *      found, callback function should return LLIOC_CONT.
1180 *  @rcp: The result of ioctl command.
1181 *
1182 *  Return values:
1183 *      If @magic matches the pointer returned by ll_iocontrol_data, the
1184 *      callback should return LLIOC_STOP; return LLIOC_STOP otherwise.
1185 */
1186typedef enum llioc_iter (*llioc_callback_t)(struct inode *inode,
1187                struct file *file, unsigned int cmd, unsigned long arg,
1188                void *magic, int *rcp);
1189
1190/* export functions */
1191/* Register ioctl block dynamatically for a regular file.
1192 *
1193 * @cmd: the array of ioctl command set
1194 * @count: number of commands in the @cmd
1195 * @cb: callback function, it will be called if an ioctl command is found to
1196 *      belong to the command list @cmd.
1197 *
1198 * Return value:
1199 *      A magic pointer will be returned if success;
1200 *      otherwise, NULL will be returned.
1201 * */
1202void *ll_iocontrol_register(llioc_callback_t cb, int count, unsigned int *cmd);
1203void ll_iocontrol_unregister(void *magic);
1204
1205int cl_sync_file_range(struct inode *inode, loff_t start, loff_t end,
1206                       enum cl_fsync_mode mode, int ignore_layout);
1207
1208/** direct write pages */
1209struct ll_dio_pages {
1210        /** page array to be written. we don't support
1211         * partial pages except the last one.
1212         */
1213        struct page **ldp_pages;
1214        /* offset of each page */
1215        loff_t       *ldp_offsets;
1216        /** if ldp_offsets is NULL, it means a sequential
1217         * pages to be written, then this is the file offset
1218         * of the first page.
1219         */
1220        loff_t  ldp_start_offset;
1221        /** how many bytes are to be written. */
1222        size_t  ldp_size;
1223        /** # of pages in the array. */
1224        int        ldp_nr;
1225};
1226
1227ssize_t ll_direct_rw_pages(const struct lu_env *env, struct cl_io *io,
1228                           int rw, struct inode *inode,
1229                           struct ll_dio_pages *pv);
1230
1231static inline int ll_file_nolock(const struct file *file)
1232{
1233        struct ll_file_data *fd = LUSTRE_FPRIVATE(file);
1234        struct inode *inode = file_inode(file);
1235
1236        return ((fd->fd_flags & LL_FILE_IGNORE_LOCK) ||
1237                (ll_i2sbi(inode)->ll_flags & LL_SBI_NOLCK));
1238}
1239
1240static inline void ll_set_lock_data(struct obd_export *exp, struct inode *inode,
1241                                    struct lookup_intent *it, __u64 *bits)
1242{
1243        if (!it->it_lock_set) {
1244                struct lustre_handle handle;
1245
1246                /* If this inode is a remote object, it will get two
1247                 * separate locks in different namespaces, Master MDT,
1248                 * where the name entry is, will grant LOOKUP lock,
1249                 * remote MDT, where the object is, will grant
1250                 * UPDATE|PERM lock. The inode will be attached to both
1251                 * LOOKUP and PERM locks, so revoking either locks will
1252                 * case the dcache being cleared
1253                 */
1254                if (it->it_remote_lock_mode) {
1255                        handle.cookie = it->it_remote_lock_handle;
1256                        CDEBUG(D_DLMTRACE, "setting l_data to inode " DFID "%p for remote lock %#llx\n",
1257                               PFID(ll_inode2fid(inode)), inode,
1258                               handle.cookie);
1259                        md_set_lock_data(exp, &handle, inode, NULL);
1260                }
1261
1262                handle.cookie = it->it_lock_handle;
1263
1264                CDEBUG(D_DLMTRACE, "setting l_data to inode " DFID "%p for lock %#llx\n",
1265                       PFID(ll_inode2fid(inode)), inode, handle.cookie);
1266
1267                md_set_lock_data(exp, &handle, inode, &it->it_lock_bits);
1268                it->it_lock_set = 1;
1269        }
1270
1271        if (bits)
1272                *bits = it->it_lock_bits;
1273}
1274
1275static inline int d_lustre_invalid(const struct dentry *dentry)
1276{
1277        return ll_d2d(dentry)->lld_invalid;
1278}
1279
1280/*
1281 * Mark dentry INVALID, if dentry refcount is zero (this is normally case for
1282 * ll_md_blocking_ast), unhash this dentry, and let dcache to reclaim it later;
1283 * else dput() of the last refcount will unhash this dentry and kill it.
1284 */
1285static inline void d_lustre_invalidate(struct dentry *dentry, int nested)
1286{
1287        CDEBUG(D_DENTRY, "invalidate dentry %pd (%p) parent %p inode %p refc %d\n",
1288               dentry, dentry,
1289               dentry->d_parent, d_inode(dentry), d_count(dentry));
1290
1291        spin_lock_nested(&dentry->d_lock,
1292                         nested ? DENTRY_D_LOCK_NESTED : DENTRY_D_LOCK_NORMAL);
1293        ll_d2d(dentry)->lld_invalid = 1;
1294        /*
1295         * We should be careful about dentries created by d_obtain_alias().
1296         * These dentries are not put in the dentry tree, instead they are
1297         * linked to sb->s_anon through dentry->d_hash.
1298         * shrink_dcache_for_umount() shrinks the tree and sb->s_anon list.
1299         * If we unhashed such a dentry, unmount would not be able to find
1300         * it and busy inodes would be reported.
1301         */
1302        if (d_count(dentry) == 0 && !(dentry->d_flags & DCACHE_DISCONNECTED))
1303                __d_drop(dentry);
1304        spin_unlock(&dentry->d_lock);
1305}
1306
1307static inline void d_lustre_revalidate(struct dentry *dentry)
1308{
1309        spin_lock(&dentry->d_lock);
1310        LASSERT(ll_d2d(dentry));
1311        ll_d2d(dentry)->lld_invalid = 0;
1312        spin_unlock(&dentry->d_lock);
1313}
1314
1315int ll_layout_conf(struct inode *inode, const struct cl_object_conf *conf);
1316int ll_layout_refresh(struct inode *inode, __u32 *gen);
1317int ll_layout_restore(struct inode *inode, loff_t start, __u64 length);
1318
1319int ll_xattr_init(void);
1320void ll_xattr_fini(void);
1321
1322int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
1323                    struct cl_page *page, enum cl_req_type crt);
1324
1325int ll_getparent(struct file *file, struct getparent __user *arg);
1326
1327/* lcommon_cl.c */
1328int cl_setattr_ost(struct cl_object *obj, const struct iattr *attr,
1329                   unsigned int attr_flags);
1330
1331extern struct lu_env *cl_inode_fini_env;
1332extern u16 cl_inode_fini_refcheck;
1333
1334int cl_file_inode_init(struct inode *inode, struct lustre_md *md);
1335void cl_inode_fini(struct inode *inode);
1336
1337__u64 cl_fid_build_ino(const struct lu_fid *fid, int api32);
1338__u32 cl_fid_build_gen(const struct lu_fid *fid);
1339
1340#endif /* LLITE_INTERNAL_H */
1341