linux/drivers/staging/lustre/lustre/include/obd.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.sun.com/software/products/lustre/docs/GPLv2.pdf
  19 *
  20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21 * CA 95054 USA or visit www.sun.com if you need additional information or
  22 * have any questions.
  23 *
  24 * GPL HEADER END
  25 */
  26/*
  27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  28 * Use is subject to license terms.
  29 *
  30 * Copyright (c) 2011, 2012, Intel Corporation.
  31 */
  32/*
  33 * This file is part of Lustre, http://www.lustre.org/
  34 * Lustre is a trademark of Sun Microsystems, Inc.
  35 */
  36
  37#ifndef __OBD_H
  38#define __OBD_H
  39
  40#include "linux/obd.h"
  41
  42#define IOC_OSC_TYPE     'h'
  43#define IOC_OSC_MIN_NR       20
  44#define IOC_OSC_SET_ACTIVE   _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
  45#define IOC_OSC_MAX_NR       50
  46
  47#define IOC_MDC_TYPE     'i'
  48#define IOC_MDC_MIN_NR       20
  49#define IOC_MDC_MAX_NR       50
  50
  51#include "lustre/lustre_idl.h"
  52#include "lustre_lib.h"
  53#include "lu_ref.h"
  54#include "lustre_export.h"
  55#include "lustre_fid.h"
  56#include "lustre_fld.h"
  57
  58#define MAX_OBD_DEVICES 8192
  59
  60struct osc_async_rc {
  61        int     ar_rc;
  62        int     ar_force_sync;
  63        __u64   ar_min_xid;
  64};
  65
  66struct lov_oinfo {               /* per-stripe data structure */
  67        struct ost_id   loi_oi;    /* object ID/Sequence on the target OST */
  68        int loi_ost_idx;           /* OST stripe index in lov_tgt_desc->tgts */
  69        int loi_ost_gen;           /* generation of this loi_ost_idx */
  70
  71        unsigned long loi_kms_valid:1;
  72        __u64 loi_kms;       /* known minimum size */
  73        struct ost_lvb loi_lvb;
  74        struct osc_async_rc     loi_ar;
  75};
  76
  77static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
  78{
  79        oinfo->loi_kms = kms;
  80        oinfo->loi_kms_valid = 1;
  81}
  82
  83static inline void loi_init(struct lov_oinfo *loi)
  84{
  85}
  86
  87struct lov_stripe_md {
  88        atomic_t     lsm_refc;
  89        spinlock_t      lsm_lock;
  90        pid_t       lsm_lock_owner; /* debugging */
  91
  92        /* maximum possible file size, might change as OSTs status changes,
  93         * e.g. disconnected, deactivated */
  94        __u64       lsm_maxbytes;
  95        struct {
  96                /* Public members. */
  97                struct ost_id lw_object_oi; /* lov object id/seq */
  98
  99                /* LOV-private members start here -- only for use in lov/. */
 100                __u32 lw_magic;
 101                __u32 lw_stripe_size;      /* size of the stripe */
 102                __u32 lw_pattern;         /* striping pattern (RAID0, RAID1) */
 103                __u16 lw_stripe_count;  /* number of objects being striped over */
 104                __u16 lw_layout_gen;       /* generation of the layout */
 105                char  lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
 106        } lsm_wire;
 107
 108        struct lov_oinfo *lsm_oinfo[0];
 109};
 110
 111#define lsm_oi           lsm_wire.lw_object_oi
 112#define lsm_magic       lsm_wire.lw_magic
 113#define lsm_layout_gen   lsm_wire.lw_layout_gen
 114#define lsm_stripe_size  lsm_wire.lw_stripe_size
 115#define lsm_pattern      lsm_wire.lw_pattern
 116#define lsm_stripe_count lsm_wire.lw_stripe_count
 117#define lsm_pool_name    lsm_wire.lw_pool_name
 118
 119static inline bool lsm_is_released(struct lov_stripe_md *lsm)
 120{
 121        return !!(lsm->lsm_pattern & LOV_PATTERN_F_RELEASED);
 122}
 123
 124static inline bool lsm_has_objects(struct lov_stripe_md *lsm)
 125{
 126        if (lsm == NULL)
 127                return false;
 128        if (lsm_is_released(lsm))
 129                return false;
 130        return true;
 131}
 132
 133static inline int lov_stripe_md_size(unsigned int stripe_count)
 134{
 135        struct lov_stripe_md lsm;
 136
 137        return sizeof(lsm) + stripe_count * sizeof(lsm.lsm_oinfo[0]);
 138}
 139
 140struct obd_info;
 141
 142typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
 143
 144/* obd info for a particular level (lov, osc). */
 145struct obd_info {
 146        /* Flags used for set request specific flags:
 147           - while lock handling, the flags obtained on the enqueue
 148           request are set here.
 149           - while stats, the flags used for control delay/resend.
 150           - while setattr, the flags used for distinguish punch operation
 151         */
 152        __u64              oi_flags;
 153        /* lsm data specific for every OSC. */
 154        struct lov_stripe_md   *oi_md;
 155        /* obdo data specific for every OSC, if needed at all. */
 156        struct obdo         *oi_oa;
 157        /* statfs data specific for every OSC, if needed at all. */
 158        struct obd_statfs      *oi_osfs;
 159        /* An update callback which is called to update some data on upper
 160         * level. E.g. it is used for update lsm->lsm_oinfo at every received
 161         * request in osc level for enqueue requests. It is also possible to
 162         * update some caller data from LOV layer if needed. */
 163        obd_enqueue_update_f    oi_cb_up;
 164};
 165
 166void lov_stripe_lock(struct lov_stripe_md *md);
 167void lov_stripe_unlock(struct lov_stripe_md *md);
 168
 169struct obd_type {
 170        struct list_head typ_chain;
 171        struct obd_ops *typ_dt_ops;
 172        struct md_ops *typ_md_ops;
 173        struct dentry *typ_debugfs_entry;
 174        char *typ_name;
 175        int  typ_refcnt;
 176        struct lu_device_type *typ_lu;
 177        spinlock_t obd_type_lock;
 178        struct kobject *typ_kobj;
 179};
 180
 181struct brw_page {
 182        u64 off;
 183        struct page *pg;
 184        int count;
 185        u32 flag;
 186};
 187
 188/* llog contexts */
 189enum llog_ctxt_id {
 190        LLOG_CONFIG_ORIG_CTXT  =  0,
 191        LLOG_CONFIG_REPL_CTXT,
 192        LLOG_MDS_OST_ORIG_CTXT,
 193        LLOG_MDS_OST_REPL_CTXT,
 194        LLOG_SIZE_ORIG_CTXT,
 195        LLOG_SIZE_REPL_CTXT,
 196        LLOG_RD1_ORIG_CTXT,
 197        LLOG_RD1_REPL_CTXT,
 198        LLOG_TEST_ORIG_CTXT,
 199        LLOG_TEST_REPL_CTXT,
 200        LLOG_LOVEA_ORIG_CTXT,
 201        LLOG_LOVEA_REPL_CTXT,
 202        LLOG_CHANGELOG_ORIG_CTXT,       /**< changelog generation on mdd */
 203        LLOG_CHANGELOG_REPL_CTXT,       /**< changelog access on clients */
 204        LLOG_CHANGELOG_USER_ORIG_CTXT,  /**< for multiple changelog consumers */
 205        LLOG_AGENT_ORIG_CTXT,           /**< agent requests generation on cdt */
 206        LLOG_MAX_CTXTS
 207};
 208
 209struct timeout_item {
 210        enum timeout_event ti_event;
 211        unsigned long    ti_timeout;
 212        timeout_cb_t       ti_cb;
 213        void          *ti_cb_data;
 214        struct list_head         ti_obd_list;
 215        struct list_head         ti_chain;
 216};
 217
 218#define OSC_MAX_RIF_DEFAULT       8
 219#define MDS_OSC_MAX_RIF_DEFAULT   50
 220#define OSC_MAX_RIF_MAX  256
 221#define OSC_MAX_DIRTY_DEFAULT  (OSC_MAX_RIF_DEFAULT * 4)
 222#define OSC_MAX_DIRTY_MB_MAX   2048     /* arbitrary, but < MAX_LONG bytes */
 223#define OSC_DEFAULT_RESENDS      10
 224
 225/* possible values for fo_sync_lock_cancel */
 226enum {
 227        NEVER_SYNC_ON_CANCEL = 0,
 228        BLOCKING_SYNC_ON_CANCEL = 1,
 229        ALWAYS_SYNC_ON_CANCEL = 2,
 230        NUM_SYNC_ON_CANCEL_STATES
 231};
 232
 233#define MDC_MAX_RIF_DEFAULT       8
 234#define MDC_MAX_RIF_MAX  512
 235
 236struct mdc_rpc_lock;
 237struct obd_import;
 238struct client_obd {
 239        struct rw_semaphore  cl_sem;
 240        struct obd_uuid   cl_target_uuid;
 241        struct obd_import       *cl_import; /* ptlrpc connection state */
 242        int                   cl_conn_count;
 243        /* max_mds_easize is purely a performance thing so we don't have to
 244         * call obd_size_diskmd() all the time. */
 245        int                      cl_default_mds_easize;
 246        int                      cl_max_mds_easize;
 247        int                      cl_default_mds_cookiesize;
 248        int                      cl_max_mds_cookiesize;
 249
 250        enum lustre_sec_part     cl_sp_me;
 251        enum lustre_sec_part     cl_sp_to;
 252        struct sptlrpc_flavor    cl_flvr_mgc;   /* fixed flavor of mgc->mgs */
 253
 254        /* the grant values are protected by loi_list_lock below */
 255        long                 cl_dirty;   /* all _dirty_ in bytes */
 256        long                 cl_dirty_max;     /* allowed w/o rpc */
 257        long                 cl_dirty_transit; /* dirty synchronous */
 258        long                 cl_avail_grant;   /* bytes of credit for ost */
 259        long                 cl_lost_grant;    /* lost credits (trunc) */
 260
 261        /* since we allocate grant by blocks, we don't know how many grant will
 262         * be used to add a page into cache. As a solution, we reserve maximum
 263         * grant before trying to dirty a page and unreserve the rest.
 264         * See osc_{reserve|unreserve}_grant for details. */
 265        long             cl_reserved_grant;
 266        struct list_head cl_cache_waiters; /* waiting for cache/grant */
 267        unsigned long    cl_next_shrink_grant;   /* jiffies */
 268        struct list_head cl_grant_shrink_list;  /* Timeout event list */
 269        int              cl_grant_shrink_interval; /* seconds */
 270
 271        /* A chunk is an optimal size used by osc_extent to determine
 272         * the extent size. A chunk is max(PAGE_CACHE_SIZE, OST block size) */
 273        int               cl_chunkbits;
 274        int               cl_chunk;
 275        int               cl_extent_tax; /* extent overhead, by bytes */
 276
 277        /* keep track of objects that have lois that contain pages which
 278         * have been queued for async brw.  this lock also protects the
 279         * lists of osc_client_pages that hang off of the loi */
 280        /*
 281         * ->cl_loi_list_lock protects consistency of
 282         * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
 283         * ->ap_completion() call-backs are executed under this lock. As we
 284         * cannot guarantee that these call-backs never block on all platforms
 285         * (as a matter of fact they do block on Mac OS X), type of
 286         * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
 287         * and blocking mutex on Mac OS X. (Alternative is to make this lock
 288         * blocking everywhere, but we don't want to slow down fast-path of
 289         * our main platform.)
 290         *
 291         * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
 292         * with client_obd_list_{un,}lock() and
 293         * client_obd_list_lock_{init,done}() functions.
 294         *
 295         * NB by Jinshan: though field names are still _loi_, but actually
 296         * osc_object{}s are in the list.
 297         */
 298        client_obd_lock_t       cl_loi_list_lock;
 299        struct list_head               cl_loi_ready_list;
 300        struct list_head               cl_loi_hp_ready_list;
 301        struct list_head               cl_loi_write_list;
 302        struct list_head               cl_loi_read_list;
 303        int                   cl_r_in_flight;
 304        int                   cl_w_in_flight;
 305        /* just a sum of the loi/lop pending numbers to be exported by /proc */
 306        atomic_t             cl_pending_w_pages;
 307        atomic_t             cl_pending_r_pages;
 308        __u32                    cl_max_pages_per_rpc;
 309        int                   cl_max_rpcs_in_flight;
 310        struct obd_histogram     cl_read_rpc_hist;
 311        struct obd_histogram     cl_write_rpc_hist;
 312        struct obd_histogram     cl_read_page_hist;
 313        struct obd_histogram     cl_write_page_hist;
 314        struct obd_histogram     cl_read_offset_hist;
 315        struct obd_histogram     cl_write_offset_hist;
 316
 317        /* lru for osc caching pages */
 318        struct cl_client_cache  *cl_cache;
 319        struct list_head         cl_lru_osc; /* member of cl_cache->ccc_lru */
 320        atomic_t                *cl_lru_left;
 321        atomic_t                 cl_lru_busy;
 322        atomic_t                 cl_lru_shrinkers;
 323        atomic_t                 cl_lru_in_list;
 324        struct list_head         cl_lru_list; /* lru page list */
 325        client_obd_lock_t        cl_lru_list_lock; /* page list protector */
 326
 327        /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
 328        atomic_t             cl_destroy_in_flight;
 329        wait_queue_head_t             cl_destroy_waitq;
 330
 331        struct mdc_rpc_lock     *cl_rpc_lock;
 332        struct mdc_rpc_lock     *cl_close_lock;
 333
 334        /* mgc datastruct */
 335        atomic_t             cl_mgc_refcount;
 336        struct obd_export       *cl_mgc_mgsexp;
 337
 338        /* checksumming for data sent over the network */
 339        unsigned int         cl_checksum:1; /* 0 = disabled, 1 = enabled */
 340        /* supported checksum types that are worked out at connect time */
 341        __u32               cl_supp_cksum_types;
 342        /* checksum algorithm to be used */
 343        cksum_type_t         cl_cksum_type;
 344
 345        /* also protected by the poorly named _loi_list_lock lock above */
 346        struct osc_async_rc      cl_ar;
 347
 348        /* used by quotacheck when the servers are older than 2.4 */
 349        int                   cl_qchk_stat; /* quotacheck stat of the peer */
 350#define CL_NOT_QUOTACHECKED 1   /* client->cl_qchk_stat init value */
 351#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 7, 50, 0)
 352#warning "please consider removing quotacheck compatibility code"
 353#endif
 354
 355        /* sequence manager */
 356        struct lu_client_seq    *cl_seq;
 357
 358        atomic_t             cl_resends; /* resend count */
 359
 360        /* ptlrpc work for writeback in ptlrpcd context */
 361        void                *cl_writeback_work;
 362        /* hash tables for osc_quota_info */
 363        struct cfs_hash       *cl_quota_hash[MAXQUOTAS];
 364};
 365
 366#define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
 367
 368struct obd_id_info {
 369        __u32   idx;
 370        u64     *data;
 371};
 372
 373struct echo_client_obd {
 374        struct obd_export       *ec_exp;   /* the local connection to osc/lov */
 375        spinlock_t              ec_lock;
 376        struct list_head           ec_objects;
 377        struct list_head           ec_locks;
 378        int               ec_nstripes;
 379        __u64           ec_unique;
 380};
 381
 382/* Generic subset of OSTs */
 383struct ost_pool {
 384        __u32         *op_array;      /* array of index of
 385                                                   lov_obd->lov_tgts */
 386        unsigned int    op_count;      /* number of OSTs in the array */
 387        unsigned int    op_size;       /* allocated size of lp_array */
 388        struct rw_semaphore op_rw_sem;     /* to protect ost_pool use */
 389};
 390
 391/* Round-robin allocator data */
 392struct lov_qos_rr {
 393        __u32          lqr_start_idx;   /* start index of new inode */
 394        __u32          lqr_offset_idx;  /* aliasing for start_idx  */
 395        int              lqr_start_count; /* reseed counter */
 396        struct ost_pool     lqr_pool;   /* round-robin optimized list */
 397        unsigned long       lqr_dirty:1;     /* recalc round-robin list */
 398};
 399
 400/* allow statfs data caching for 1 second */
 401#define OBD_STATFS_CACHE_SECONDS 1
 402
 403struct lov_statfs_data {
 404        struct obd_info   lsd_oi;
 405        struct obd_statfs lsd_statfs;
 406};
 407
 408/* Stripe placement optimization */
 409struct lov_qos {
 410        struct list_head    lq_oss_list; /* list of OSSs that targets use */
 411        struct rw_semaphore lq_rw_sem;
 412        __u32           lq_active_oss_count;
 413        unsigned int    lq_prio_free;   /* priority for free space */
 414        unsigned int    lq_threshold_rr;/* priority for rr */
 415        struct lov_qos_rr   lq_rr;        /* round robin qos data */
 416        unsigned long       lq_dirty:1,     /* recalc qos data */
 417                            lq_same_space:1,/* the ost's all have approx.
 418                                               the same space avail */
 419                            lq_reset:1,     /* zero current penalties */
 420                            lq_statfs_in_progress:1; /* statfs op in
 421                                                        progress */
 422        /* qos statfs data */
 423        struct lov_statfs_data *lq_statfs_data;
 424        wait_queue_head_t        lq_statfs_waitq; /* waitqueue to notify statfs
 425                                              * requests completion */
 426};
 427
 428struct lov_tgt_desc {
 429        struct list_head          ltd_kill;
 430        struct obd_uuid     ltd_uuid;
 431        struct obd_device  *ltd_obd;
 432        struct obd_export  *ltd_exp;
 433        __u32          ltd_gen;
 434        __u32          ltd_index;   /* index in lov_obd->tgts */
 435        unsigned long       ltd_active:1,/* is this target up for requests */
 436                            ltd_activate:1,/* should  target be activated */
 437                            ltd_reap:1;  /* should this target be deleted */
 438};
 439
 440/* Pool metadata */
 441#define pool_tgt_size(_p)   _p->pool_obds.op_size
 442#define pool_tgt_count(_p)  _p->pool_obds.op_count
 443#define pool_tgt_array(_p)  _p->pool_obds.op_array
 444#define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
 445
 446struct pool_desc {
 447        char              pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
 448        struct ost_pool       pool_obds;              /* pool members */
 449        atomic_t          pool_refcount;          /* pool ref. counter */
 450        struct lov_qos_rr     pool_rr;          /* round robin qos */
 451        struct hlist_node      pool_hash;             /* access by poolname */
 452        struct list_head            pool_list;        /* serial access */
 453        struct dentry           *pool_debugfs_entry;    /* file in /proc */
 454        struct obd_device    *pool_lobd;        /* obd of the lov/lod to which
 455                                                *  this pool belongs */
 456};
 457
 458struct lov_obd {
 459        struct lov_desc  desc;
 460        struct lov_tgt_desc   **lov_tgts;             /* sparse array */
 461        struct ost_pool  lov_packed;        /* all OSTs in a packed
 462                                                          array */
 463        struct mutex            lov_lock;
 464        struct obd_connect_data lov_ocd;
 465        atomic_t            lov_refcount;
 466        __u32              lov_tgt_count;        /* how many OBD's */
 467        __u32              lov_active_tgt_count;  /* how many active */
 468        __u32              lov_death_row;/* tgts scheduled to be deleted */
 469        __u32              lov_tgt_size;   /* size of tgts array */
 470        int                  lov_connects;
 471        int                  lov_pool_count;
 472        struct cfs_hash      *lov_pools_hash_body; /* used for key access */
 473        struct list_head        lov_pool_list; /* used for sequential access */
 474        struct dentry           *lov_pool_debugfs_entry;
 475        enum lustre_sec_part    lov_sp_me;
 476
 477        /* Cached LRU pages from upper layer */
 478        void                   *lov_cache;
 479
 480        struct rw_semaphore     lov_notify_lock;
 481
 482        struct kobject          *lov_tgts_kobj;
 483};
 484
 485struct lmv_tgt_desc {
 486        struct obd_uuid         ltd_uuid;
 487        struct obd_export       *ltd_exp;
 488        int                     ltd_idx;
 489        struct mutex            ltd_fid_mutex;
 490        unsigned long           ltd_active:1; /* target up for requests */
 491};
 492
 493enum placement_policy {
 494        PLACEMENT_CHAR_POLICY   = 0,
 495        PLACEMENT_NID_POLICY    = 1,
 496        PLACEMENT_INVAL_POLICY  = 2,
 497        PLACEMENT_MAX_POLICY
 498};
 499
 500struct lmv_obd {
 501        int                     refcount;
 502        struct lu_client_fld    lmv_fld;
 503        spinlock_t              lmv_lock;
 504        enum placement_policy   lmv_placement;
 505        struct lmv_desc         desc;
 506        struct obd_uuid         cluuid;
 507        struct obd_export       *exp;
 508
 509        struct mutex            init_mutex;
 510        int                     connected;
 511        int                     max_easize;
 512        int                     max_def_easize;
 513        int                     max_cookiesize;
 514        int                     max_def_cookiesize;
 515        int                     server_timeout;
 516
 517        int                     tgts_size; /* size of tgts array */
 518        struct lmv_tgt_desc     **tgts;
 519
 520        struct obd_connect_data conn_data;
 521        struct kobject          *lmv_tgts_kobj;
 522};
 523
 524struct niobuf_local {
 525        __u64           lnb_file_offset;
 526        __u32           lnb_page_offset;
 527        __u32           len;
 528        __u32           flags;
 529        struct page     *page;
 530        struct dentry   *dentry;
 531        int             lnb_grant_used;
 532        int             rc;
 533};
 534
 535#define LUSTRE_FLD_NAME  "fld"
 536#define LUSTRE_SEQ_NAME  "seq"
 537
 538#define LUSTRE_MDD_NAME  "mdd"
 539#define LUSTRE_OSD_LDISKFS_NAME "osd-ldiskfs"
 540#define LUSTRE_OSD_ZFS_NAME     "osd-zfs"
 541#define LUSTRE_VVP_NAME  "vvp"
 542#define LUSTRE_LMV_NAME  "lmv"
 543#define LUSTRE_SLP_NAME  "slp"
 544#define LUSTRE_LOD_NAME         "lod"
 545#define LUSTRE_OSP_NAME         "osp"
 546#define LUSTRE_LWP_NAME         "lwp"
 547
 548/* obd device type names */
 549 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
 550#define LUSTRE_MDS_NAME  "mds"
 551#define LUSTRE_MDT_NAME  "mdt"
 552#define LUSTRE_MDC_NAME  "mdc"
 553#define LUSTRE_OSS_NAME  "ost"       /* FIXME change name to oss */
 554#define LUSTRE_OST_NAME  "obdfilter" /* FIXME change name to ost */
 555#define LUSTRE_OSC_NAME  "osc"
 556#define LUSTRE_LOV_NAME  "lov"
 557#define LUSTRE_MGS_NAME  "mgs"
 558#define LUSTRE_MGC_NAME  "mgc"
 559
 560#define LUSTRE_ECHO_NAME        "obdecho"
 561#define LUSTRE_ECHO_CLIENT_NAME "echo_client"
 562#define LUSTRE_QMT_NAME  "qmt"
 563
 564/* Constant obd names (post-rename) */
 565#define LUSTRE_MDS_OBDNAME "MDS"
 566#define LUSTRE_OSS_OBDNAME "OSS"
 567#define LUSTRE_MGS_OBDNAME "MGS"
 568#define LUSTRE_MGC_OBDNAME "MGC"
 569
 570/* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
 571#define N_LOCAL_TEMP_PAGE 0x10000000
 572
 573struct obd_trans_info {
 574        __u64               oti_transno;
 575        __u64               oti_xid;
 576        /* Only used on the server side for tracking acks. */
 577        struct oti_req_ack_lock {
 578                struct lustre_handle lock;
 579                __u32           mode;
 580        }                       oti_ack_locks[4];
 581        void                *oti_handle;
 582        struct llog_cookie       oti_onecookie;
 583        struct llog_cookie      *oti_logcookies;
 584        int                   oti_numcookies;
 585        /** synchronous write is needed */
 586        unsigned long            oti_sync_write:1;
 587
 588        /* initial thread handling transaction */
 589        struct ptlrpc_thread *oti_thread;
 590        __u32               oti_conn_cnt;
 591        /** VBR: versions */
 592        __u64               oti_pre_version;
 593        /** JobID */
 594        char                *oti_jobid;
 595
 596        struct obd_uuid  *oti_ost_uuid;
 597};
 598
 599static inline void oti_init(struct obd_trans_info *oti,
 600                            struct ptlrpc_request *req)
 601{
 602        if (oti == NULL)
 603                return;
 604        memset(oti, 0, sizeof(*oti));
 605
 606        if (req == NULL)
 607                return;
 608
 609        oti->oti_xid = req->rq_xid;
 610        /** VBR: take versions from request */
 611        if (req->rq_reqmsg != NULL &&
 612            lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
 613                __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
 614
 615                oti->oti_pre_version = pre_version ? pre_version[0] : 0;
 616                oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
 617        }
 618
 619        /** called from mds_create_objects */
 620        if (req->rq_repmsg != NULL)
 621                oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
 622        oti->oti_thread = req->rq_svc_thread;
 623        if (req->rq_reqmsg != NULL)
 624                oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
 625}
 626
 627static inline void oti_alloc_cookies(struct obd_trans_info *oti,
 628                                     int num_cookies)
 629{
 630        if (!oti)
 631                return;
 632
 633        if (num_cookies == 1)
 634                oti->oti_logcookies = &oti->oti_onecookie;
 635        else
 636                oti->oti_logcookies = libcfs_kvzalloc(num_cookies * sizeof(oti->oti_onecookie),
 637                                                      GFP_NOFS);
 638
 639        oti->oti_numcookies = num_cookies;
 640}
 641
 642static inline void oti_free_cookies(struct obd_trans_info *oti)
 643{
 644        if (!oti || !oti->oti_logcookies)
 645                return;
 646
 647        if (oti->oti_logcookies == &oti->oti_onecookie)
 648                LASSERT(oti->oti_numcookies == 1);
 649        else
 650                kvfree(oti->oti_logcookies);
 651
 652        oti->oti_logcookies = NULL;
 653        oti->oti_numcookies = 0;
 654}
 655
 656/*
 657 * Events signalled through obd_notify() upcall-chain.
 658 */
 659enum obd_notify_event {
 660        /* target added */
 661        OBD_NOTIFY_CREATE,
 662        /* Device connect start */
 663        OBD_NOTIFY_CONNECT,
 664        /* Device activated */
 665        OBD_NOTIFY_ACTIVE,
 666        /* Device deactivated */
 667        OBD_NOTIFY_INACTIVE,
 668        /* Device disconnected */
 669        OBD_NOTIFY_DISCON,
 670        /* Connect data for import were changed */
 671        OBD_NOTIFY_OCD,
 672        /* Sync request */
 673        OBD_NOTIFY_SYNC_NONBLOCK,
 674        OBD_NOTIFY_SYNC,
 675        /* Configuration event */
 676        OBD_NOTIFY_CONFIG,
 677        /* Administratively deactivate/activate event */
 678        OBD_NOTIFY_DEACTIVATE,
 679        OBD_NOTIFY_ACTIVATE
 680};
 681
 682/*
 683 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
 684 * and liblustre being main examples).
 685 */
 686struct obd_notify_upcall {
 687        int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
 688                          enum obd_notify_event ev, void *owner, void *data);
 689        /* Opaque datum supplied by upper layer listener */
 690        void *onu_owner;
 691};
 692
 693struct target_recovery_data {
 694        svc_handler_t           trd_recovery_handler;
 695        pid_t                   trd_processing_task;
 696        struct completion       trd_starting;
 697        struct completion       trd_finishing;
 698};
 699
 700struct obd_llog_group {
 701        int             olg_seq;
 702        struct llog_ctxt  *olg_ctxts[LLOG_MAX_CTXTS];
 703        wait_queue_head_t       olg_waitq;
 704        spinlock_t         olg_lock;
 705        struct mutex       olg_cat_processing;
 706};
 707
 708/* corresponds to one of the obd's */
 709#define OBD_DEVICE_MAGIC        0XAB5CD6EF
 710#define OBD_DEV_BY_DEVNAME      0xffffd0de
 711
 712struct lvfs_run_ctxt {
 713        struct dt_device *dt;
 714};
 715
 716struct obd_device {
 717        struct obd_type *obd_type;
 718        __u32              obd_magic;
 719
 720        /* common and UUID name of this device */
 721        char                obd_name[MAX_OBD_NAME];
 722        struct obd_uuid  obd_uuid;
 723
 724        struct lu_device       *obd_lu_dev;
 725
 726        int                  obd_minor;
 727        /* bitfield modification is protected by obd_dev_lock */
 728        unsigned long obd_attached:1,      /* finished attach */
 729                      obd_set_up:1,     /* finished setup */
 730                      obd_version_recov:1, /* obd uses version checking */
 731                      obd_replayable:1,    /* recovery is enabled; inform clients */
 732                      obd_no_transno:1,    /* no committed-transno notification */
 733                      obd_no_recov:1,      /* fail instead of retry messages */
 734                      obd_stopping:1,      /* started cleanup */
 735                      obd_starting:1,      /* started setup */
 736                      obd_force:1,       /* cleanup with > 0 obd refcount */
 737                      obd_fail:1,         /* cleanup with failover */
 738                      obd_async_recov:1,   /* allow asynchronous orphan cleanup */
 739                      obd_no_conn:1,       /* deny new connections */
 740                      obd_inactive:1,      /* device active/inactive
 741                                           * (for /proc/status only!!) */
 742                      obd_no_ir:1,       /* no imperative recovery. */
 743                      obd_process_conf:1;  /* device is processing mgs config */
 744        /* use separate field as it is set in interrupt to don't mess with
 745         * protection of other bits using _bh lock */
 746        unsigned long obd_recovery_expired:1;
 747        /* uuid-export hash body */
 748        struct cfs_hash      *obd_uuid_hash;
 749        atomic_t            obd_refcount;
 750        wait_queue_head_t            obd_refcount_waitq;
 751        struct list_head              obd_exports;
 752        struct list_head              obd_unlinked_exports;
 753        struct list_head              obd_delayed_exports;
 754        int                  obd_num_exports;
 755        spinlock_t              obd_nid_lock;
 756        struct ldlm_namespace  *obd_namespace;
 757        struct ptlrpc_client    obd_ldlm_client; /* XXX OST/MDS only */
 758        /* a spinlock is OK for what we do now, may need a semaphore later */
 759        spinlock_t              obd_dev_lock; /* protect OBD bitfield above */
 760        struct mutex            obd_dev_mutex;
 761        __u64                   obd_last_committed;
 762        spinlock_t              obd_osfs_lock;
 763        struct obd_statfs       obd_osfs;       /* locked by obd_osfs_lock */
 764        __u64                   obd_osfs_age;
 765        struct lvfs_run_ctxt    obd_lvfs_ctxt;
 766        struct obd_llog_group   obd_olg;        /* default llog group */
 767        struct obd_device       *obd_observer;
 768        struct rw_semaphore     obd_observer_link_sem;
 769        struct obd_notify_upcall obd_upcall;
 770        struct obd_export       *obd_self_export;
 771
 772        union {
 773                struct client_obd cli;
 774                struct echo_client_obd echo_client;
 775                struct lov_obd lov;
 776                struct lmv_obd lmv;
 777        } u;
 778        /* Fields used by LProcFS */
 779        unsigned int       obd_cntr_base;
 780        struct lprocfs_stats  *obd_stats;
 781
 782        unsigned int       md_cntr_base;
 783        struct lprocfs_stats  *md_stats;
 784
 785        struct dentry           *obd_debugfs_entry;
 786        struct dentry           *obd_svc_debugfs_entry;
 787        struct lprocfs_stats  *obd_svc_stats;
 788        atomic_t           obd_evict_inprogress;
 789        wait_queue_head_t           obd_evict_inprogress_waitq;
 790        struct list_head        obd_evict_list; /* protected with pet_lock */
 791
 792        /**
 793         * Ldlm pool part. Save last calculated SLV and Limit.
 794         */
 795        rwlock_t                obd_pool_lock;
 796        int                 obd_pool_limit;
 797        __u64             obd_pool_slv;
 798
 799        /**
 800         * A list of outstanding class_incref()'s against this obd. For
 801         * debugging.
 802         */
 803        struct lu_ref     obd_reference;
 804
 805        int                    obd_conn_inprogress;
 806
 807        struct kobject          obd_kobj; /* sysfs object */
 808        struct completion       obd_kobj_unregister;
 809};
 810
 811#define OBD_LLOG_FL_SENDNOW     0x0001
 812#define OBD_LLOG_FL_EXIT        0x0002
 813
 814enum obd_cleanup_stage {
 815/* Special case hack for MDS LOVs */
 816        OBD_CLEANUP_EARLY,
 817/* can be directly mapped to .ldto_device_fini() */
 818        OBD_CLEANUP_EXPORTS,
 819};
 820
 821/* get/set_info keys */
 822#define KEY_ASYNC              "async"
 823#define KEY_BLOCKSIZE_BITS      "blocksize_bits"
 824#define KEY_BLOCKSIZE      "blocksize"
 825#define KEY_CHANGELOG_CLEAR     "changelog_clear"
 826#define KEY_FID2PATH        "fid2path"
 827#define KEY_CHECKSUM        "checksum"
 828#define KEY_CLEAR_FS        "clear_fs"
 829#define KEY_CONN_DATA      "conn_data"
 830#define KEY_EVICT_BY_NID        "evict_by_nid"
 831#define KEY_FIEMAP            "fiemap"
 832#define KEY_FLUSH_CTX      "flush_ctx"
 833#define KEY_GRANT_SHRINK        "grant_shrink"
 834#define KEY_HSM_COPYTOOL_SEND   "hsm_send"
 835#define KEY_INIT_RECOV_BACKUP   "init_recov_bk"
 836#define KEY_INIT_RECOV    "initial_recov"
 837#define KEY_INTERMDS        "inter_mds"
 838#define KEY_LAST_ID          "last_id"
 839#define KEY_LAST_FID            "last_fid"
 840#define KEY_LOCK_TO_STRIPE      "lock_to_stripe"
 841#define KEY_LOVDESC          "lovdesc"
 842#define KEY_LOV_IDX          "lov_idx"
 843#define KEY_MAX_EASIZE          "max_easize"
 844#define KEY_DEFAULT_EASIZE      "default_easize"
 845#define KEY_MDS_CONN        "mds_conn"
 846#define KEY_MGSSEC            "mgssec"
 847#define KEY_NEXT_ID          "next_id"
 848#define KEY_READ_ONLY      "read-only"
 849#define KEY_REGISTER_TARGET     "register_target"
 850#define KEY_SET_FS            "set_fs"
 851#define KEY_TGT_COUNT      "tgt_count"
 852/*      KEY_SET_INFO in lustre_idl.h */
 853#define KEY_SPTLRPC_CONF        "sptlrpc_conf"
 854#define KEY_CONNECT_FLAG        "connect_flags"
 855#define KEY_SYNC_LOCK_CANCEL    "sync_lock_cancel"
 856
 857#define KEY_CACHE_SET           "cache_set"
 858#define KEY_CACHE_LRU_SHRINK    "cache_lru_shrink"
 859#define KEY_CHANGELOG_INDEX     "changelog_index"
 860
 861struct lu_context;
 862
 863/* /!\ must be coherent with include/linux/namei.h on patched kernel */
 864#define IT_OPEN     (1 << 0)
 865#define IT_CREAT    (1 << 1)
 866#define IT_READDIR  (1 << 2)
 867#define IT_GETATTR  (1 << 3)
 868#define IT_LOOKUP   (1 << 4)
 869#define IT_UNLINK   (1 << 5)
 870#define IT_TRUNC    (1 << 6)
 871#define IT_GETXATTR (1 << 7)
 872#define IT_EXEC     (1 << 8)
 873#define IT_PIN      (1 << 9)
 874#define IT_LAYOUT   (1 << 10)
 875#define IT_QUOTA_DQACQ (1 << 11)
 876#define IT_QUOTA_CONN  (1 << 12)
 877#define IT_SETXATTR (1 << 13)
 878
 879static inline int it_to_lock_mode(struct lookup_intent *it)
 880{
 881        /* CREAT needs to be tested before open (both could be set) */
 882        if (it->it_op & IT_CREAT)
 883                return LCK_CW;
 884        else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP |
 885                              IT_LAYOUT))
 886                return LCK_CR;
 887        else if (it->it_op &  IT_GETXATTR)
 888                return LCK_PR;
 889        else if (it->it_op &  IT_SETXATTR)
 890                return LCK_PW;
 891
 892        LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
 893        return -EINVAL;
 894}
 895
 896struct md_op_data {
 897        struct lu_fid      op_fid1; /* operation fid1 (usually parent) */
 898        struct lu_fid      op_fid2; /* operation fid2 (usually child) */
 899        struct lu_fid      op_fid3; /* 2 extra fids to find conflicting */
 900        struct lu_fid      op_fid4; /* to the operation locks. */
 901        u32                     op_mds;  /* what mds server open will go to */
 902        struct lustre_handle    op_handle;
 903        s64                     op_mod_time;
 904        const char           *op_name;
 905        int                  op_namelen;
 906        __u32              op_mode;
 907        struct lmv_stripe_md   *op_mea1;
 908        struct lmv_stripe_md   *op_mea2;
 909        __u32              op_suppgids[2];
 910        __u32              op_fsuid;
 911        __u32              op_fsgid;
 912        cfs_cap_t              op_cap;
 913        void               *op_data;
 914
 915        /* iattr fields and blocks. */
 916        struct iattr        op_attr;
 917        unsigned int        op_attr_flags;
 918        __u64              op_valid;
 919        loff_t            op_attr_blocks;
 920
 921        /* Size-on-MDS epoch and flags. */
 922        __u64              op_ioepoch;
 923        __u32              op_flags;
 924
 925        /* Various operation flags. */
 926        enum mds_op_bias        op_bias;
 927
 928        /* Operation type */
 929        __u32              op_opc;
 930
 931        /* Used by readdir */
 932        __u64              op_offset;
 933
 934        /* Used by readdir */
 935        __u32              op_npages;
 936
 937        /* used to transfer info between the stacks of MD client
 938         * see enum op_cli_flags */
 939        __u32                   op_cli_flags;
 940
 941        /* File object data version for HSM release, on client */
 942        __u64                   op_data_version;
 943        struct lustre_handle    op_lease_handle;
 944};
 945
 946enum op_cli_flags {
 947        CLI_SET_MEA     = 1 << 0,
 948        CLI_RM_ENTRY    = 1 << 1,
 949};
 950
 951struct md_enqueue_info;
 952/* metadata stat-ahead */
 953
 954struct md_enqueue_info {
 955        struct md_op_data       mi_data;
 956        struct lookup_intent    mi_it;
 957        struct lustre_handle    mi_lockh;
 958        struct inode       *mi_dir;
 959        int (*mi_cb)(struct ptlrpc_request *req,
 960                struct md_enqueue_info *minfo, int rc);
 961        __u64              mi_cbdata;
 962        unsigned int        mi_generation;
 963};
 964
 965struct obd_ops {
 966        struct module *o_owner;
 967        int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
 968                           void *karg, void *uarg);
 969        int (*o_get_info)(const struct lu_env *env, struct obd_export *,
 970                          __u32 keylen, void *key, __u32 *vallen, void *val,
 971                          struct lov_stripe_md *lsm);
 972        int (*o_set_info_async)(const struct lu_env *, struct obd_export *,
 973                                __u32 keylen, void *key,
 974                                __u32 vallen, void *val,
 975                                struct ptlrpc_request_set *set);
 976        int (*o_attach)(struct obd_device *dev, u32 len, void *data);
 977        int (*o_detach)(struct obd_device *dev);
 978        int (*o_setup)(struct obd_device *dev, struct lustre_cfg *cfg);
 979        int (*o_precleanup)(struct obd_device *dev,
 980                            enum obd_cleanup_stage cleanup_stage);
 981        int (*o_cleanup)(struct obd_device *dev);
 982        int (*o_process_config)(struct obd_device *dev, u32 len, void *data);
 983        int (*o_postrecov)(struct obd_device *dev);
 984        int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
 985                          int priority);
 986        int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
 987        /* connect to the target device with given connection
 988         * data. @ocd->ocd_connect_flags is modified to reflect flags actually
 989         * granted by the target, which are guaranteed to be a subset of flags
 990         * asked for. If @ocd == NULL, use default parameters. */
 991        int (*o_connect)(const struct lu_env *env,
 992                         struct obd_export **exp, struct obd_device *src,
 993                         struct obd_uuid *cluuid, struct obd_connect_data *ocd,
 994                         void *localdata);
 995        int (*o_reconnect)(const struct lu_env *env,
 996                           struct obd_export *exp, struct obd_device *src,
 997                           struct obd_uuid *cluuid,
 998                           struct obd_connect_data *ocd,
 999                           void *localdata);
1000        int (*o_disconnect)(struct obd_export *exp);
1001
1002        /* Initialize/finalize fids infrastructure. */
1003        int (*o_fid_init)(struct obd_device *obd,
1004                          struct obd_export *exp, enum lu_cli_type type);
1005        int (*o_fid_fini)(struct obd_device *obd);
1006
1007        /* Allocate new fid according to passed @hint. */
1008        int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1009                           struct md_op_data *op_data);
1010
1011        /*
1012         * Object with @fid is getting deleted, we may want to do something
1013         * about this.
1014         */
1015        int (*o_statfs)(const struct lu_env *, struct obd_export *exp,
1016                        struct obd_statfs *osfs, __u64 max_age, __u32 flags);
1017        int (*o_statfs_async)(struct obd_export *exp, struct obd_info *oinfo,
1018                              __u64 max_age, struct ptlrpc_request_set *set);
1019        int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1020                        struct lov_stripe_md *mem_src);
1021        int (*o_unpackmd)(struct obd_export *exp,
1022                          struct lov_stripe_md **mem_tgt,
1023                          struct lov_mds_md *disk_src, int disk_len);
1024        int (*o_preallocate)(struct lustre_handle *, u32 *req, u64 *ids);
1025        int (*o_create)(const struct lu_env *env, struct obd_export *exp,
1026                        struct obdo *oa, struct lov_stripe_md **ea,
1027                        struct obd_trans_info *oti);
1028        int (*o_destroy)(const struct lu_env *env, struct obd_export *exp,
1029                         struct obdo *oa, struct lov_stripe_md *ea,
1030                         struct obd_trans_info *oti, struct obd_export *md_exp);
1031        int (*o_setattr)(const struct lu_env *, struct obd_export *exp,
1032                         struct obd_info *oinfo, struct obd_trans_info *oti);
1033        int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1034                               struct obd_trans_info *oti,
1035                               struct ptlrpc_request_set *rqset);
1036        int (*o_getattr)(const struct lu_env *env, struct obd_export *exp,
1037                         struct obd_info *oinfo);
1038        int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1039                               struct ptlrpc_request_set *set);
1040        int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1041                            u64 size, int shrink);
1042        int (*o_preprw)(const struct lu_env *env, int cmd,
1043                        struct obd_export *exp, struct obdo *oa, int objcount,
1044                        struct obd_ioobj *obj, struct niobuf_remote *remote,
1045                        int *nr_pages, struct niobuf_local *local,
1046                        struct obd_trans_info *oti);
1047        int (*o_commitrw)(const struct lu_env *env, int cmd,
1048                          struct obd_export *exp, struct obdo *oa,
1049                          int objcount, struct obd_ioobj *obj,
1050                          struct niobuf_remote *remote, int pages,
1051                          struct niobuf_local *local,
1052                          struct obd_trans_info *oti, int rc);
1053        int (*o_find_cbdata)(struct obd_export *, struct lov_stripe_md *,
1054                             ldlm_iterator_t it, void *data);
1055        int (*o_init_export)(struct obd_export *exp);
1056        int (*o_destroy_export)(struct obd_export *exp);
1057
1058        /* metadata-only methods */
1059        int (*o_import_event)(struct obd_device *, struct obd_import *,
1060                              enum obd_import_event);
1061
1062        int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1063                        enum obd_notify_event ev, void *data);
1064
1065        int (*o_health_check)(const struct lu_env *env, struct obd_device *);
1066        struct obd_uuid *(*o_get_uuid)(struct obd_export *exp);
1067
1068        /* quota methods */
1069        int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1070                            struct obd_quotactl *);
1071        int (*o_quotactl)(struct obd_device *, struct obd_export *,
1072                          struct obd_quotactl *);
1073
1074        /* pools methods */
1075        int (*o_pool_new)(struct obd_device *obd, char *poolname);
1076        int (*o_pool_del)(struct obd_device *obd, char *poolname);
1077        int (*o_pool_add)(struct obd_device *obd, char *poolname,
1078                          char *ostname);
1079        int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1080                          char *ostname);
1081        void (*o_getref)(struct obd_device *obd);
1082        void (*o_putref)(struct obd_device *obd);
1083        /*
1084         * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1085         * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1086         * Also, add a wrapper function in include/linux/obd_class.h. */
1087};
1088
1089enum {
1090        LUSTRE_OPC_MKDIR    = (1 << 0),
1091        LUSTRE_OPC_SYMLINK  = (1 << 1),
1092        LUSTRE_OPC_MKNOD    = (1 << 2),
1093        LUSTRE_OPC_CREATE   = (1 << 3),
1094        LUSTRE_OPC_ANY      = (1 << 4)
1095};
1096
1097/* lmv structures */
1098#define MEA_MAGIC_LAST_CHAR      0xb2221ca1
1099#define MEA_MAGIC_ALL_CHARS      0xb222a11c
1100#define MEA_MAGIC_HASH_SEGMENT   0xb222a11b
1101
1102#define MAX_HASH_SIZE_32         0x7fffffffUL
1103#define MAX_HASH_SIZE       0x7fffffffffffffffULL
1104#define MAX_HASH_HIGHEST_BIT     0x1000000000000000ULL
1105
1106struct lustre_md {
1107        struct mdt_body  *body;
1108        struct lov_stripe_md    *lsm;
1109        struct lmv_stripe_md    *mea;
1110#ifdef CONFIG_FS_POSIX_ACL
1111        struct posix_acl        *posix_acl;
1112#endif
1113        struct mdt_remote_perm  *remote_perm;
1114};
1115
1116struct md_open_data {
1117        struct obd_client_handle *mod_och;
1118        struct ptlrpc_request    *mod_open_req;
1119        struct ptlrpc_request    *mod_close_req;
1120        atomic_t                  mod_refcount;
1121        bool                      mod_is_create;
1122};
1123
1124struct lookup_intent;
1125
1126struct md_ops {
1127        int (*m_getstatus)(struct obd_export *, struct lu_fid *);
1128        int (*m_null_inode)(struct obd_export *, const struct lu_fid *);
1129        int (*m_find_cbdata)(struct obd_export *, const struct lu_fid *,
1130                             ldlm_iterator_t, void *);
1131        int (*m_close)(struct obd_export *, struct md_op_data *,
1132                       struct md_open_data *, struct ptlrpc_request **);
1133        int (*m_create)(struct obd_export *, struct md_op_data *,
1134                        const void *, int, int, __u32, __u32, cfs_cap_t,
1135                        __u64, struct ptlrpc_request **);
1136        int (*m_done_writing)(struct obd_export *, struct md_op_data  *,
1137                              struct md_open_data *);
1138        int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1139                         struct lookup_intent *, struct md_op_data *,
1140                         struct lustre_handle *, void *, int,
1141                         struct ptlrpc_request **, __u64);
1142        int (*m_getattr)(struct obd_export *, struct md_op_data *,
1143                         struct ptlrpc_request **);
1144        int (*m_getattr_name)(struct obd_export *, struct md_op_data *,
1145                              struct ptlrpc_request **);
1146        int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1147                             void *, int, struct lookup_intent *, int,
1148                             struct ptlrpc_request **,
1149                             ldlm_blocking_callback, __u64);
1150        int (*m_link)(struct obd_export *, struct md_op_data *,
1151                      struct ptlrpc_request **);
1152        int (*m_rename)(struct obd_export *, struct md_op_data *,
1153                        const char *, int, const char *, int,
1154                        struct ptlrpc_request **);
1155        int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1156                           const struct lu_fid *,
1157                           struct ptlrpc_request **);
1158        int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1159                         int, void *, int, struct ptlrpc_request **,
1160                         struct md_open_data **mod);
1161        int (*m_sync)(struct obd_export *, const struct lu_fid *,
1162                      struct ptlrpc_request **);
1163        int (*m_readpage)(struct obd_export *, struct md_op_data *,
1164                          struct page **, struct ptlrpc_request **);
1165
1166        int (*m_unlink)(struct obd_export *, struct md_op_data *,
1167                        struct ptlrpc_request **);
1168
1169        int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1170                          u64, const char *, const char *, int, int, int, __u32,
1171                          struct ptlrpc_request **);
1172
1173        int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1174                          u64, const char *, const char *, int, int, int,
1175                          struct ptlrpc_request **);
1176
1177        int (*m_init_ea_size)(struct obd_export *, int, int, int, int);
1178
1179        int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1180                               struct obd_export *, struct obd_export *,
1181                               struct lustre_md *);
1182
1183        int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1184
1185        int (*m_set_open_replay_data)(struct obd_export *,
1186                                      struct obd_client_handle *,
1187                                      struct lookup_intent *);
1188        int (*m_clear_open_replay_data)(struct obd_export *,
1189                                        struct obd_client_handle *);
1190        int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u64 *);
1191
1192        ldlm_mode_t (*m_lock_match)(struct obd_export *, __u64,
1193                                    const struct lu_fid *, ldlm_type_t,
1194                                    ldlm_policy_data_t *, ldlm_mode_t,
1195                                    struct lustre_handle *);
1196
1197        int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1198                               ldlm_policy_data_t *, ldlm_mode_t,
1199                               ldlm_cancel_flags_t flags, void *opaque);
1200
1201        int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1202                                 __u32, struct ptlrpc_request **);
1203
1204        int (*m_intent_getattr_async)(struct obd_export *,
1205                                      struct md_enqueue_info *,
1206                                      struct ldlm_enqueue_info *);
1207
1208        int (*m_revalidate_lock)(struct obd_export *, struct lookup_intent *,
1209                                 struct lu_fid *, __u64 *bits);
1210
1211        /*
1212         * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1213         * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1214         * wrapper function in include/linux/obd_class.h.
1215         */
1216};
1217
1218struct lsm_operations {
1219        void (*lsm_free)(struct lov_stripe_md *);
1220        int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1221                           struct obd_export *md_exp);
1222        void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, u64 *,
1223                                    u64 *);
1224        void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, u64 *,
1225                                     u64 *);
1226        int (*lsm_lmm_verify)(struct lov_mds_md *lmm, int lmm_bytes,
1227                               __u16 *stripe_count);
1228        int (*lsm_unpackmd)(struct lov_obd *lov, struct lov_stripe_md *lsm,
1229                             struct lov_mds_md *lmm);
1230};
1231
1232extern const struct lsm_operations lsm_v1_ops;
1233extern const struct lsm_operations lsm_v3_ops;
1234static inline const struct lsm_operations *lsm_op_find(int magic)
1235{
1236        switch (magic) {
1237        case LOV_MAGIC_V1:
1238               return &lsm_v1_ops;
1239        case LOV_MAGIC_V3:
1240               return &lsm_v3_ops;
1241        default:
1242               CERROR("Cannot recognize lsm_magic %08x\n", magic);
1243               return NULL;
1244        }
1245}
1246
1247/* Requests for obd_extent_calc() */
1248#define OBD_CALC_STRIPE_START   1
1249#define OBD_CALC_STRIPE_END     2
1250
1251static inline struct md_open_data *obd_mod_alloc(void)
1252{
1253        struct md_open_data *mod;
1254
1255        mod = kzalloc(sizeof(*mod), GFP_NOFS);
1256        if (mod == NULL)
1257                return NULL;
1258        atomic_set(&mod->mod_refcount, 1);
1259        return mod;
1260}
1261
1262#define obd_mod_get(mod) atomic_inc(&(mod)->mod_refcount)
1263#define obd_mod_put(mod)                                        \
1264({                                                            \
1265        if (atomic_dec_and_test(&(mod)->mod_refcount)) {          \
1266                if ((mod)->mod_open_req)                          \
1267                        ptlrpc_req_finished((mod)->mod_open_req);   \
1268                kfree(mod);                           \
1269        }                                                      \
1270})
1271
1272void obdo_from_inode(struct obdo *dst, struct inode *src, u32 valid);
1273void obdo_set_parent_fid(struct obdo *dst, const struct lu_fid *parent);
1274
1275/* return 1 if client should be resend request */
1276static inline int client_should_resend(int resend, struct client_obd *cli)
1277{
1278        return atomic_read(&cli->cl_resends) ?
1279               atomic_read(&cli->cl_resends) > resend : 1;
1280}
1281
1282/**
1283 * Return device name for this device
1284 *
1285 * XXX: lu_device is declared before obd_device, while a pointer pointing
1286 * back to obd_device in lu_device, so this helper function defines here
1287 * instead of in lu_object.h
1288 */
1289static inline const char *lu_dev_name(const struct lu_device *lu_dev)
1290{
1291        return lu_dev->ld_obd->obd_name;
1292}
1293
1294static inline bool filename_is_volatile(const char *name, int namelen, int *idx)
1295{
1296        const char      *start;
1297        char            *end;
1298
1299        if (strncmp(name, LUSTRE_VOLATILE_HDR, LUSTRE_VOLATILE_HDR_LEN) != 0)
1300                return false;
1301
1302        /* caller does not care of idx */
1303        if (idx == NULL)
1304                return true;
1305
1306        /* volatile file, the MDT can be set from name */
1307        /* name format is LUSTRE_VOLATILE_HDR:[idx]: */
1308        /* if no MDT is specified, use std way */
1309        if (namelen < LUSTRE_VOLATILE_HDR_LEN + 2)
1310                goto bad_format;
1311        /* test for no MDT idx case */
1312        if ((*(name + LUSTRE_VOLATILE_HDR_LEN) == ':') &&
1313            (*(name + LUSTRE_VOLATILE_HDR_LEN + 1) == ':')) {
1314                *idx = -1;
1315                return true;
1316        }
1317        /* we have an idx, read it */
1318        start = name + LUSTRE_VOLATILE_HDR_LEN + 1;
1319        *idx = simple_strtoul(start, &end, 0);
1320        /* error cases:
1321         * no digit, no trailing :, negative value
1322         */
1323        if (((*idx == 0) && (end == start)) ||
1324            (*end != ':') || (*idx < 0))
1325                goto bad_format;
1326
1327        return true;
1328bad_format:
1329        /* bad format of mdt idx, we cannot return an error
1330         * to caller so we use hash algo */
1331        CERROR("Bad volatile file name format: %s\n",
1332               name + LUSTRE_VOLATILE_HDR_LEN);
1333        return false;
1334}
1335
1336static inline int cli_brw_size(struct obd_device *obd)
1337{
1338        LASSERT(obd != NULL);
1339        return obd->u.cli.cl_max_pages_per_rpc << PAGE_CACHE_SHIFT;
1340}
1341
1342#endif /* __OBD_H */
1343