Whamcloud - gitweb
LU-1194 llog: fix for not sync llcd at thread stop
[fs/lustre-release.git] / lustre / include / obd.h
index 3ff60f0..5fa66e2 100644 (file)
@@ -1,6 +1,4 @@
-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
  * GPL HEADER START
  *
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  * GPL HEADER END
  */
 /*
- * Copyright  2008 Sun Microsystems, Inc. All rights reserved
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2012, Whamcloud, Inc.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
 
 #define IOC_MDC_TYPE         'i'
 #define IOC_MDC_MIN_NR       20
-/* Moved to lustre_user.h
-#define IOC_MDC_LOOKUP       _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
-#define IOC_MDC_GETSTRIPE    _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
 #define IOC_MDC_MAX_NR       50
 
 #include <lustre/lustre_idl.h>
-#include <lu_target.h>
+#ifdef HAVE_SERVER_SUPPORT
+# include <lu_target.h>
+#endif
 #include <lu_ref.h>
 #include <lustre_lib.h>
 #include <lustre_export.h>
 #include <lustre_quota.h>
 #include <lustre_fld.h>
 #include <lustre_capa.h>
-#include <class_hash.h>
 
 #include <libcfs/bitmap.h>
 
 
 #define MAX_OBD_DEVICES 8192
 
-/* this is really local to the OSC */
-struct loi_oap_pages {
-        struct list_head        lop_pending;
-        struct list_head        lop_urgent;
-        struct list_head        lop_pending_group;
-        int                     lop_num_pending;
-};
-
 struct osc_async_rc {
         int     ar_rc;
         int     ar_force_sync;
@@ -89,24 +79,17 @@ struct osc_async_rc {
 };
 
 struct lov_oinfo {                 /* per-stripe data structure */
-        __u64 loi_id;              /* object ID on the target OST */
-        __u64 loi_gr;              /* object group on the target OST */
+        struct ost_id   loi_oi;    /* object ID/Sequence on the target OST */
         int loi_ost_idx;           /* OST stripe index in lov_tgt_desc->tgts */
         int loi_ost_gen;           /* generation of this loi_ost_idx */
 
-        /* used by the osc to keep track of what objects to build into rpcs */
-        struct loi_oap_pages loi_read_lop;
-        struct loi_oap_pages loi_write_lop;
-        struct list_head loi_ready_item;
-        struct list_head loi_hp_ready_item;
-        struct list_head loi_write_item;
-        struct list_head loi_read_item;
-
         unsigned long loi_kms_valid:1;
         __u64 loi_kms;             /* known minimum size */
         struct ost_lvb loi_lvb;
         struct osc_async_rc     loi_ar;
 };
+#define loi_id  loi_oi.oi_id
+#define loi_seq loi_oi.oi_seq
 
 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
 {
@@ -116,59 +99,36 @@ static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
 
 static inline void loi_init(struct lov_oinfo *loi)
 {
-        CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
-        CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
-        CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
-        CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
-        CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
-        CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
-        CFS_INIT_LIST_HEAD(&loi->loi_ready_item);
-        CFS_INIT_LIST_HEAD(&loi->loi_hp_ready_item);
-        CFS_INIT_LIST_HEAD(&loi->loi_write_item);
-        CFS_INIT_LIST_HEAD(&loi->loi_read_item);
 }
 
-/*extent array item for describing the joined file extent info*/
-struct lov_extent {
-        __u64 le_start;            /* extent start */
-        __u64 le_len;              /* extent length */
-        int   le_loi_idx;          /* extent #1 loi's index in lsm loi array */
-        int   le_stripe_count;     /* extent stripe count*/
-};
-
-/*Lov array info for describing joined file array EA info*/
-struct lov_array_info {
-        struct llog_logid    lai_array_id;    /* MDS med llog object id */
-        unsigned             lai_ext_count; /* number of extent count */
-        struct lov_extent    *lai_ext_array; /* extent desc array */
-};
-
 struct lov_stripe_md {
-        spinlock_t       lsm_lock;
+        cfs_spinlock_t   lsm_lock;
         pid_t            lsm_lock_owner; /* debugging */
 
+        /* maximum possible file size, might change as OSTs status changes,
+         * e.g. disconnected, deactivated */
+        __u64            lsm_maxbytes;
         struct {
                 /* Public members. */
                 __u64 lw_object_id;        /* lov object id */
-                __u64 lw_object_gr;        /* lov object group */
-                __u64 lw_maxbytes;         /* maximum possible file size */
+                __u64 lw_object_seq;       /* lov object seq */
 
                 /* LOV-private members start here -- only for use in lov/. */
                 __u32 lw_magic;
                 __u32 lw_stripe_size;      /* size of the stripe */
                 __u32 lw_pattern;          /* striping pattern (RAID0, RAID1) */
-                unsigned lw_stripe_count;  /* number of objects being striped over */
+                __u16 lw_stripe_count;  /* number of objects being striped over */
+                __u16 lw_layout_gen;       /* generation of the layout */
                 char  lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
         } lsm_wire;
 
-        struct lov_array_info *lsm_array; /*Only for joined file array info*/
         struct lov_oinfo *lsm_oinfo[0];
 };
 
 #define lsm_object_id    lsm_wire.lw_object_id
-#define lsm_object_gr    lsm_wire.lw_object_gr
-#define lsm_maxbytes     lsm_wire.lw_maxbytes
+#define lsm_object_seq   lsm_wire.lw_object_seq
 #define lsm_magic        lsm_wire.lw_magic
+#define lsm_layout_gen   lsm_wire.lw_layout_gen
 #define lsm_stripe_size  lsm_wire.lw_stripe_size
 #define lsm_pattern      lsm_wire.lw_pattern
 #define lsm_stripe_count lsm_wire.lw_stripe_count
@@ -188,6 +148,7 @@ struct obd_info {
            - while lock handling, the flags obtained on the enqueue
            request are set here.
            - while stats, the flags used for control delay/resend.
+           - while setattr, the flags used for distinguish punch operation
          */
         int                     oi_flags;
         /* Lock handle specific for every OSC lock. */
@@ -206,6 +167,8 @@ struct obd_info {
         /* oss capability, its type is obd_capa in client to avoid copy.
          * in contrary its type is lustre_capa in OSS. */
         void                   *oi_capa;
+       /* transfer jobid from ost_sync() to filter_sync()... */
+       char                   *oi_jobid;
 };
 
 /* compare all relevant fields. */
@@ -219,18 +182,69 @@ static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
         return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
 }
 
+static inline int lov_lum_lsm_cmp(struct lov_user_md *lum,
+                                  struct lov_stripe_md  *lsm)
+{
+        if (lsm->lsm_magic != lum->lmm_magic)
+                return 1;
+        if ((lsm->lsm_stripe_count != 0) && (lum->lmm_stripe_count != 0) &&
+            (lsm->lsm_stripe_count != lum->lmm_stripe_count))
+                return 2;
+        if ((lsm->lsm_stripe_size != 0) && (lum->lmm_stripe_size != 0) &&
+            (lsm->lsm_stripe_size != lum->lmm_stripe_size))
+                return 3;
+        if ((lsm->lsm_pattern != 0) && (lum->lmm_pattern != 0) &&
+            (lsm->lsm_pattern != lum->lmm_pattern))
+                return 4;
+        if ((lsm->lsm_magic == LOV_MAGIC_V3) &&
+            (strncmp(lsm->lsm_pool_name,
+                     ((struct lov_user_md_v3 *)lum)->lmm_pool_name,
+                     LOV_MAXPOOLNAME) != 0))
+                return 5;
+        return 0;
+}
+
+static inline int lov_lum_swab_if_needed(struct lov_user_md_v3 *lumv3,
+                                         int *lmm_magic,
+                                         struct lov_user_md *lum)
+{
+        if (lum && cfs_copy_from_user(lumv3, lum,sizeof(struct lov_user_md_v1)))
+                return -EFAULT;
+
+        *lmm_magic = lumv3->lmm_magic;
+
+        if (*lmm_magic == __swab32(LOV_USER_MAGIC_V1)) {
+                lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lumv3);
+                *lmm_magic = LOV_USER_MAGIC_V1;
+        } else if (*lmm_magic == LOV_USER_MAGIC_V3) {
+                if (lum && cfs_copy_from_user(lumv3, lum, sizeof(*lumv3)))
+                        return -EFAULT;
+        } else if (*lmm_magic == __swab32(LOV_USER_MAGIC_V3)) {
+                if (lum && cfs_copy_from_user(lumv3, lum, sizeof(*lumv3)))
+                        return -EFAULT;
+                lustre_swab_lov_user_md_v3(lumv3);
+                *lmm_magic = LOV_USER_MAGIC_V3;
+        } else if (*lmm_magic != LOV_USER_MAGIC_V1) {
+                CDEBUG(D_IOCTL,
+                       "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n",
+                       *lmm_magic, LOV_USER_MAGIC_V1, LOV_USER_MAGIC_V3);
+                       return -EINVAL;
+        }
+        return 0;
+}
+
 void lov_stripe_lock(struct lov_stripe_md *md);
 void lov_stripe_unlock(struct lov_stripe_md *md);
 
 struct obd_type {
-        struct list_head typ_chain;
+        cfs_list_t typ_chain;
         struct obd_ops *typ_dt_ops;
         struct md_ops *typ_md_ops;
         cfs_proc_dir_entry_t *typ_procroot;
         char *typ_name;
         int  typ_refcnt;
         struct lu_device_type *typ_lu;
-        spinlock_t obd_type_lock;
+        cfs_spinlock_t obd_type_lock;
 };
 
 struct brw_page {
@@ -244,29 +258,30 @@ struct brw_page {
 
 struct ost_server_data;
 
+struct osd_properties {
+        size_t osd_max_ea_size;
+};
+
+#define OBT_MAGIC       0xBDDECEAE
 /* hold common fields for "target" device */
 struct obd_device_target {
+        __u32                     obt_magic;
+        __u32                     obt_instance;
         struct super_block       *obt_sb;
         /** last_rcvd file */
         struct file              *obt_rcvd_filp;
-        /** server data in last_rcvd file */
-        struct lr_server_data    *obt_lsd;
-        /** Lock protecting client bitmap */
-        spinlock_t                obt_client_bitmap_lock;
-        /** Bitmap of known clients */
-        unsigned long            *obt_client_bitmap;
-        /** Server last transaction number */
-        __u64                     obt_last_transno;
-        /** Lock protecting last transaction number */
-        spinlock_t                obt_translock;
-        /** Number of mounts */
+#ifdef HAVE_SERVER_SUPPORT
+        struct lu_target         *obt_lut;
+#endif
         __u64                     obt_mount_count;
-        struct semaphore          obt_quotachecking;
+        cfs_semaphore_t           obt_quotachecking;
         struct lustre_quota_ctxt  obt_qctxt;
         lustre_quota_version_t    obt_qfmt;
-        struct rw_semaphore       obt_rwsem;
+        cfs_rw_semaphore_t        obt_rwsem;
         struct vfsmount          *obt_vfsmnt;
         struct file              *obt_health_check_filp;
+       struct osd_properties     obt_osd_properties;
+       struct obd_job_stats      obt_jobstats;
 };
 
 /* llog contexts */
@@ -304,24 +319,21 @@ struct filter_ext {
 struct filter_obd {
         /* NB this field MUST be first */
         struct obd_device_target fo_obt;
-        struct lu_target     fo_lut;
         const char          *fo_fstype;
 
         int                  fo_group_count;
         cfs_dentry_t        *fo_dentry_O;
         cfs_dentry_t       **fo_dentry_O_groups;
         struct filter_subdirs   *fo_dentry_O_sub;
-        struct semaphore     fo_init_lock;      /* group initialization lock */
+        cfs_mutex_t          fo_init_lock;      /* group initialization lock */
         int                  fo_committed_group;
 
-#define CLIENT_QUOTA_DEFAULT_RESENDS 10
-
-        spinlock_t           fo_objidlock;      /* protect fo_lastobjid */
+        cfs_spinlock_t       fo_objidlock;      /* protect fo_lastobjid */
 
         unsigned long        fo_destroys_in_progress;
-        struct semaphore     fo_create_locks[FILTER_SUBDIR_COUNT];
+        cfs_mutex_t          fo_create_locks[FILTER_SUBDIR_COUNT];
 
-        struct list_head     fo_export_list;
+        cfs_list_t fo_export_list;
         int                  fo_subdir_count;
 
         obd_size             fo_tot_dirty;      /* protected by obd_osfs_lock */
@@ -330,7 +342,8 @@ struct filter_obd {
         int                  fo_tot_granted_clients;
 
         obd_size             fo_readcache_max_filesize;
-        int                  fo_read_cache:1,   /**< enable read-only cache */
+        cfs_spinlock_t       fo_flags_lock;
+        unsigned int         fo_read_cache:1,   /**< enable read-only cache */
                              fo_writethrough_cache:1,/**< read cache writes */
                              fo_mds_ost_sync:1, /**< MDS-OST orphan recovery*/
                              fo_raid_degraded:1;/**< RAID device degraded */
@@ -342,10 +355,10 @@ struct filter_obd {
         __u64               *fo_last_objids; /* last created objid for groups,
                                               * protected by fo_objidlock */
 
-        struct semaphore     fo_alloc_lock;
+        cfs_mutex_t          fo_alloc_lock;
 
-        atomic_t             fo_r_in_flight;
-        atomic_t             fo_w_in_flight;
+        cfs_atomic_t         fo_r_in_flight;
+        cfs_atomic_t         fo_w_in_flight;
 
         /*
          * per-filter pool of kiobuf's allocated by filter_common_setup() and
@@ -362,43 +375,39 @@ struct filter_obd {
         struct filter_iobuf    **fo_iobuf_pool;
         int                      fo_iobuf_count;
 
-        struct list_head         fo_llog_list;
-        spinlock_t               fo_llog_list_lock;
+        cfs_list_t               fo_llog_list;
+        cfs_spinlock_t           fo_llog_list_lock;
 
         struct brw_stats         fo_filter_stats;
         struct lustre_quota_ctxt fo_quota_ctxt;
-        spinlock_t               fo_quotacheck_lock;
-        atomic_t                 fo_quotachecking;
+        cfs_spinlock_t           fo_quotacheck_lock;
+        cfs_atomic_t             fo_quotachecking;
 
         int                      fo_fmd_max_num; /* per exp filter_mod_data */
         int                      fo_fmd_max_age; /* jiffies to fmd expiry */
+        unsigned long            fo_syncjournal:1, /* sync journal on writes */
+                                 fo_sync_lock_cancel:2;/* sync on lock cancel */
+
 
         /* sptlrpc stuff */
-        rwlock_t                 fo_sptlrpc_lock;
+        cfs_rwlock_t             fo_sptlrpc_lock;
         struct sptlrpc_rule_set  fo_sptlrpc_rset;
 
         /* capability related */
         unsigned int             fo_fl_oss_capa;
-        struct list_head         fo_capa_keys;
-        struct hlist_head       *fo_capa_hash;
+        cfs_list_t               fo_capa_keys;
+        cfs_hlist_head_t        *fo_capa_hash;
         struct llog_commit_master *fo_lcm;
         int                      fo_sec_level;
 };
 
-#define fo_translock            fo_obt.obt_translock
-#define fo_rcvd_filp            fo_obt.obt_rcvd_filp
-#define fo_fsd                  fo_obt.obt_lsd
-#define fo_last_rcvd_slots      fo_obt.obt_client_bitmap
-#define fo_mount_count          fo_obt.obt_mount_count
-#define fo_vfsmnt               fo_obt.obt_vfsmnt
-
 struct timeout_item {
         enum timeout_event ti_event;
         cfs_time_t         ti_timeout;
         timeout_cb_t       ti_cb;
         void              *ti_cb_data;
-        struct list_head   ti_obd_list;
-        struct list_head   ti_chain;
+        cfs_list_t         ti_obd_list;
+        cfs_list_t         ti_chain;
 };
 
 #define OSC_MAX_RIF_DEFAULT       8
@@ -407,13 +416,21 @@ struct timeout_item {
 #define OSC_MAX_DIRTY_MB_MAX   2048     /* arbitrary, but < MAX_LONG bytes */
 #define OSC_DEFAULT_RESENDS      10
 
+/* possible values for fo_sync_lock_cancel */
+enum {
+        NEVER_SYNC_ON_CANCEL = 0,
+        BLOCKING_SYNC_ON_CANCEL = 1,
+        ALWAYS_SYNC_ON_CANCEL = 2,
+        NUM_SYNC_ON_CANCEL_STATES
+};
+
 #define MDC_MAX_RIF_DEFAULT       8
 #define MDC_MAX_RIF_MAX         512
 
 struct mdc_rpc_lock;
 struct obd_import;
 struct client_obd {
-        struct rw_semaphore      cl_sem;
+        cfs_rw_semaphore_t       cl_sem;
         struct obd_uuid          cl_target_uuid;
         struct obd_import       *cl_import; /* ptlrpc connection state */
         int                      cl_conn_count;
@@ -436,11 +453,22 @@ struct client_obd {
         long                     cl_dirty_transit; /* dirty synchronous */
         long                     cl_avail_grant;   /* bytes of credit for ost */
         long                     cl_lost_grant;    /* lost credits (trunc) */
-        struct list_head         cl_cache_waiters; /* waiting for cache/grant */
-        cfs_time_t               cl_next_shrink_grant;   /* jiffies */
-        struct list_head         cl_grant_shrink_list;  /* Timeout event list */
-        struct semaphore         cl_grant_sem;   /*grant shrink list semaphore*/
-        int                      cl_grant_shrink_interval; /* seconds */
+
+       /* since we allocate grant by blocks, we don't know how many grant will
+        * be used to add a page into cache. As a solution, we reserve maximum
+        * grant before trying to dirty a page and unreserve the rest.
+        * See osc_{reserve|unreserve}_grant for details. */
+       long                 cl_reserved_grant;
+       cfs_list_t           cl_cache_waiters; /* waiting for cache/grant */
+       cfs_time_t           cl_next_shrink_grant;   /* jiffies */
+       cfs_list_t           cl_grant_shrink_list;  /* Timeout event list */
+       int                  cl_grant_shrink_interval; /* seconds */
+
+       /* A chunk is an optimal size used by osc_extent to determine
+        * the extent size. A chunk is max(CFS_PAGE_SIZE, OST block size) */
+       int                  cl_chunkbits;
+       int                  cl_chunk;
+       int                  cl_extent_tax; /* extent overhead, by bytes */
 
         /* keep track of objects that have lois that contain pages which
          * have been queued for async brw.  this lock also protects the
@@ -459,18 +487,21 @@ struct client_obd {
          * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
          * with client_obd_list_{un,}lock() and
          * client_obd_list_lock_{init,done}() functions.
-         */
+        *
+        * NB by Jinshan: though field names are still _loi_, but actually
+        * osc_object{}s are in the list.
+        */
         client_obd_lock_t        cl_loi_list_lock;
-        struct list_head         cl_loi_ready_list;
-        struct list_head         cl_loi_hp_ready_list;
-        struct list_head         cl_loi_write_list;
-        struct list_head         cl_loi_read_list;
+        cfs_list_t               cl_loi_ready_list;
+        cfs_list_t               cl_loi_hp_ready_list;
+        cfs_list_t               cl_loi_write_list;
+        cfs_list_t               cl_loi_read_list;
         int                      cl_r_in_flight;
         int                      cl_w_in_flight;
         /* just a sum of the loi/lop pending numbers to be exported by /proc */
-        int                      cl_pending_w_pages;
-        int                      cl_pending_r_pages;
-        int                      cl_max_pages_per_rpc;
+       cfs_atomic_t             cl_pending_w_pages;
+       cfs_atomic_t             cl_pending_r_pages;
+       int                      cl_max_pages_per_rpc;
         int                      cl_max_rpcs_in_flight;
         struct obd_histogram     cl_read_rpc_hist;
         struct obd_histogram     cl_write_rpc_hist;
@@ -480,19 +511,18 @@ struct client_obd {
         struct obd_histogram     cl_write_offset_hist;
 
         /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
-        atomic_t                 cl_destroy_in_flight;
+        cfs_atomic_t             cl_destroy_in_flight;
         cfs_waitq_t              cl_destroy_waitq;
 
         struct mdc_rpc_lock     *cl_rpc_lock;
-        struct mdc_rpc_lock     *cl_setattr_lock;
         struct mdc_rpc_lock     *cl_close_lock;
         struct osc_creator       cl_oscc;
 
         /* mgc datastruct */
-        struct semaphore         cl_mgc_sem;
+        cfs_semaphore_t          cl_mgc_sem;
         struct vfsmount         *cl_mgc_vfsmnt;
         struct dentry           *cl_mgc_configs_dir;
-        atomic_t                 cl_mgc_refcount;
+        cfs_atomic_t             cl_mgc_refcount;
         struct obd_export       *cl_mgc_mgsexp;
 
         /* checksumming for data sent over the network */
@@ -511,22 +541,25 @@ struct client_obd {
         /* sequence manager */
         struct lu_client_seq    *cl_seq;
 
-        atomic_t                 cl_resends; /* resend count */
-        atomic_t                 cl_quota_resends; /* quota related resend count */
+        cfs_atomic_t             cl_resends; /* resend count */
+
+        /* ptlrpc work for writeback in ptlrpcd context */
+        void                    *cl_writeback_work;
 };
 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
 
 #define CL_NOT_QUOTACHECKED 1   /* client->cl_qchk_stat init value */
 
 struct mgs_obd {
+        struct obd_device_target         mgs_obt;
         struct ptlrpc_service           *mgs_service;
         struct vfsmount                 *mgs_vfsmnt;
         struct super_block              *mgs_sb;
         struct dentry                   *mgs_configs_dir;
-        struct dentry                   *mgs_fid_de;
-        struct list_head                 mgs_fs_db_list;
-        struct semaphore                 mgs_sem;
+        cfs_list_t                       mgs_fs_db_list;
+        cfs_mutex_t                      mgs_mutex;
         cfs_proc_dir_entry_t            *mgs_proc_live;
+        cfs_time_t                       mgs_start_time;
 };
 
 struct mds_obd {
@@ -535,26 +568,24 @@ struct mds_obd {
         struct ptlrpc_service           *mds_service;
         struct ptlrpc_service           *mds_setattr_service;
         struct ptlrpc_service           *mds_readpage_service;
-        cfs_dentry_t                    *mds_fid_de;
         int                              mds_max_mdsize;
         int                              mds_max_cookiesize;
         __u64                            mds_io_epoch;
         unsigned long                    mds_atime_diff;
-        struct semaphore                 mds_epoch_sem;
         struct ll_fid                    mds_rootfid;
         cfs_dentry_t                    *mds_pending_dir;
         cfs_dentry_t                    *mds_logs_dir;
         cfs_dentry_t                    *mds_objects_dir;
         struct llog_handle              *mds_cfg_llh;
-        struct obd_device               *mds_osc_obd; /* XXX lov_obd */
+        struct obd_device               *mds_lov_obd;
         struct obd_uuid                  mds_lov_uuid;
         char                            *mds_profile;
-        struct obd_export               *mds_osc_exp; /* XXX lov_exp */
+        struct obd_export               *mds_lov_exp;
         struct lov_desc                  mds_lov_desc;
         __u32                            mds_id;
 
         /* mark pages dirty for write. */
-        bitmap_t                        *mds_lov_page_dirty;
+        cfs_bitmap_t                    *mds_lov_page_dirty;
         /* array for store pages with obd_id */
         void                           **mds_lov_page_array;
         /* file for store objid */
@@ -566,8 +597,7 @@ struct mds_obd {
 
 
         struct lustre_quota_info         mds_quota_info;
-        struct semaphore                 mds_qonoff_sem;
-        struct semaphore                 mds_health_sem;
+        cfs_rw_semaphore_t               mds_qonoff_sem;
         unsigned long                    mds_fl_user_xattr:1,
                                          mds_fl_acl:1,
                                          mds_evict_ost_nids:1,
@@ -581,17 +611,9 @@ struct mds_obd {
 
         /* for capability keys update */
         struct lustre_capa_key          *mds_capa_keys;
-        struct rw_semaphore              mds_notify_lock;
+        cfs_rw_semaphore_t               mds_notify_lock;
 };
 
-#define mds_transno_lock         mds_obt.obt_translock
-#define mds_rcvd_filp            mds_obt.obt_rcvd_filp
-#define mds_server_data          mds_obt.obt_lsd
-#define mds_client_bitmap        mds_obt.obt_client_bitmap
-#define mds_mount_count          mds_obt.obt_mount_count
-#define mds_last_transno         mds_obt.obt_last_transno
-#define mds_vfsmnt               mds_obt.obt_vfsmnt
-
 /* lov objid */
 extern __u32 mds_max_ost_index;
 
@@ -612,35 +634,36 @@ struct obd_id_info {
 /* */
 
 struct echo_obd {
-        struct obdo          eo_oa;
-        spinlock_t           eo_lock;
-        __u64                eo_lastino;
-        struct lustre_handle eo_nl_lock;
-        atomic_t             eo_prep;
+        struct obd_device_target eo_obt;
+        struct obdo              eo_oa;
+        cfs_spinlock_t           eo_lock;
+        __u64                    eo_lastino;
+        struct lustre_handle     eo_nl_lock;
+        cfs_atomic_t             eo_prep;
 };
 
 struct ost_obd {
         struct ptlrpc_service *ost_service;
         struct ptlrpc_service *ost_create_service;
         struct ptlrpc_service *ost_io_service;
-        struct semaphore       ost_health_sem;
+        cfs_mutex_t            ost_health_mutex;
 };
 
 struct echo_client_obd {
         struct obd_export   *ec_exp;   /* the local connection to osc/lov */
-        spinlock_t           ec_lock;
-        struct list_head     ec_objects;
-        struct list_head     ec_locks;
+        cfs_spinlock_t       ec_lock;
+        cfs_list_t           ec_objects;
+        cfs_list_t           ec_locks;
         int                  ec_nstripes;
         __u64                ec_unique;
 };
 
 struct lov_qos_oss {
         struct obd_uuid     lqo_uuid;       /* ptlrpc's c_remote_uuid */
-        struct list_head    lqo_oss_list;   /* link to lov_qos */
+        cfs_list_t          lqo_oss_list;   /* link to lov_qos */
         __u64               lqo_bavail;     /* total bytes avail on OSS */
         __u64               lqo_penalty;    /* current penalty */
-        __u64               lqo_penalty_per_obj; /* penalty decrease every obj*/
+        __u64               lqo_penalty_per_obj;/* penalty decrease every obj*/
         time_t              lqo_used;       /* last used time, seconds */
         __u32               lqo_ost_count;  /* number of osts on this oss */
 };
@@ -656,11 +679,11 @@ struct ltd_qos {
 
 /* Generic subset of OSTs */
 struct ost_pool {
-        __u32              *op_array;        /* array of index of
-                                                lov_obd->lov_tgts */
-        unsigned int        op_count;        /* number of OSTs in the array */
-        unsigned int        op_size;         /* allocated size of lp_array */
-        struct rw_semaphore op_rw_sem;       /* to protect ost_pool use */
+        __u32              *op_array;      /* array of index of
+                                                   lov_obd->lov_tgts */
+        unsigned int        op_count;      /* number of OSTs in the array */
+        unsigned int        op_size;       /* allocated size of lp_array */
+        cfs_rw_semaphore_t  op_rw_sem;     /* to protect ost_pool use */
 };
 
 /* Round-robin allocator data */
@@ -672,14 +695,17 @@ struct lov_qos_rr {
         unsigned long       lqr_dirty:1;     /* recalc round-robin list */
 };
 
+/* allow statfs data caching for 1 second */
+#define OBD_STATFS_CACHE_SECONDS 1
+
 struct lov_statfs_data {
         struct obd_info   lsd_oi;
         struct obd_statfs lsd_statfs;
 };
 /* Stripe placement optimization */
 struct lov_qos {
-        struct list_head    lq_oss_list;    /* list of OSSs that targets use */
-        struct rw_semaphore lq_rw_sem;
+        cfs_list_t          lq_oss_list; /* list of OSSs that targets use */
+        cfs_rw_semaphore_t  lq_rw_sem;
         __u32               lq_active_oss_count;
         unsigned int        lq_prio_free;   /* priority for free space */
         unsigned int        lq_threshold_rr;/* priority for rr */
@@ -688,7 +714,8 @@ struct lov_qos {
                             lq_same_space:1,/* the ost's all have approx.
                                                the same space avail */
                             lq_reset:1,     /* zero current penalties */
-                            lq_statfs_in_progress:1; /* statfs op in progress */
+                            lq_statfs_in_progress:1; /* statfs op in
+                                                        progress */
         /* qos statfs data */
         struct lov_statfs_data *lq_statfs_data;
         cfs_waitq_t         lq_statfs_waitq; /* waitqueue to notify statfs
@@ -696,7 +723,7 @@ struct lov_qos {
 };
 
 struct lov_tgt_desc {
-        struct list_head    ltd_kill;
+        cfs_list_t          ltd_kill;
         struct obd_uuid     ltd_uuid;
         struct obd_device  *ltd_obd;
         struct obd_export  *ltd_exp;
@@ -704,7 +731,7 @@ struct lov_tgt_desc {
         __u32               ltd_gen;
         __u32               ltd_index;   /* index in lov_obd->tgts */
         unsigned long       ltd_active:1,/* is this target up for requests */
-                            ltd_activate:1,/* should this target be activated */
+                            ltd_activate:1,/* should  target be activated */
                             ltd_reap:1;  /* should this target be deleted */
 };
 
@@ -718,10 +745,10 @@ struct lov_tgt_desc {
 struct pool_desc {
         char                  pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
         struct ost_pool       pool_obds;              /* pool members */
-        atomic_t              pool_refcount;          /* pool ref. counter */
+        cfs_atomic_t          pool_refcount;          /* pool ref. counter */
         struct lov_qos_rr     pool_rr;                /* round robin qos */
-        struct hlist_node     pool_hash;              /* access by poolname */
-        struct list_head      pool_list;              /* serial access */
+        cfs_hlist_node_t      pool_hash;              /* access by poolname */
+        cfs_list_t            pool_list;              /* serial access */
         cfs_proc_dir_entry_t *pool_proc_entry;        /* file in /proc */
         struct lov_obd       *pool_lov;               /* lov obd to which this
                                                          pool belong */
@@ -732,18 +759,18 @@ struct lov_obd {
         struct lov_tgt_desc   **lov_tgts;              /* sparse array */
         struct ost_pool         lov_packed;            /* all OSTs in a packed
                                                           array */
-        struct semaphore        lov_lock;
+        cfs_mutex_t             lov_lock;
         struct obd_connect_data lov_ocd;
         struct lov_qos          lov_qos;               /* qos info per lov */
-        atomic_t                lov_refcount;
+        cfs_atomic_t            lov_refcount;
         __u32                   lov_tgt_count;         /* how many OBD's */
         __u32                   lov_active_tgt_count;  /* how many active */
         __u32                   lov_death_row;/* tgts scheduled to be deleted */
         __u32                   lov_tgt_size;   /* size of tgts array */
         int                     lov_connects;
         int                     lov_pool_count;
-        lustre_hash_t          *lov_pools_hash_body; /* used for key access */
-        struct list_head        lov_pool_list; /* used for sequential access */
+        cfs_hash_t             *lov_pools_hash_body; /* used for key access */
+        cfs_list_t              lov_pool_list; /* used for sequential access */
         cfs_proc_dir_entry_t   *lov_pool_proc_entry;
         enum lustre_sec_part    lov_sp_me;
 };
@@ -751,9 +778,9 @@ struct lov_obd {
 struct lmv_tgt_desc {
         struct obd_uuid         ltd_uuid;
         struct obd_export      *ltd_exp;
-        int                     ltd_active;   /* is this target up for requests */
+        int                     ltd_active; /* is this target up for requests */
         int                     ltd_idx;
-        struct semaphore        ltd_fid_sem;
+        cfs_mutex_t             ltd_fid_mutex;
 };
 
 enum placement_policy {
@@ -768,7 +795,7 @@ typedef enum placement_policy placement_policy_t;
 struct lmv_obd {
         int                     refcount;
         struct lu_client_fld    lmv_fld;
-        spinlock_t              lmv_lock;
+        cfs_spinlock_t          lmv_lock;
         placement_policy_t      lmv_placement;
         struct lmv_desc         desc;
         struct obd_uuid         cluuid;
@@ -779,7 +806,7 @@ struct lmv_obd {
         int                     max_def_easize;
         int                     max_cookiesize;
         int                     server_timeout;
-        struct semaphore        init_sem;
+        cfs_mutex_t             init_mutex;
 
         struct lmv_tgt_desc     *tgts;
         int                     tgts_size;
@@ -791,6 +818,7 @@ struct lmv_obd {
 };
 
 struct niobuf_local {
+       __u64 lnb_file_offset;
         __u64 offset;
         __u32 len;
         __u32 flags;
@@ -805,7 +833,8 @@ struct niobuf_local {
 
 #define LUSTRE_CMM_NAME         "cmm"
 #define LUSTRE_MDD_NAME         "mdd"
-#define LUSTRE_OSD_NAME         "osd"
+#define LUSTRE_OSD_NAME         "osd-ldiskfs"
+#define LUSTRE_OSD_ZFS_NAME     "osd-zfs"
 #define LUSTRE_VVP_NAME         "vvp"
 #define LUSTRE_LMV_NAME         "lmv"
 #define LUSTRE_CMM_MDC_NAME     "cmm-mdc"
@@ -848,12 +877,16 @@ struct obd_trans_info {
         struct llog_cookie       oti_onecookie;
         struct llog_cookie      *oti_logcookies;
         int                      oti_numcookies;
+       /** synchronous write is needed */
+       long                     oti_sync_write:1;
 
         /* initial thread handling transaction */
         struct ptlrpc_thread *   oti_thread;
         __u32                    oti_conn_cnt;
         /** VBR: versions */
         __u64                    oti_pre_version;
+       /** JobID */
+       char                    *oti_jobid;
 
         struct obd_uuid         *oti_ost_uuid;
 };
@@ -893,8 +926,8 @@ static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
         if (num_cookies == 1)
                 oti->oti_logcookies = &oti->oti_onecookie;
         else
-                OBD_ALLOC(oti->oti_logcookies,
-                          num_cookies * sizeof(oti->oti_onecookie));
+                OBD_ALLOC_LARGE(oti->oti_logcookies,
+                                num_cookies * sizeof(oti->oti_onecookie));
 
         oti->oti_numcookies = num_cookies;
 }
@@ -907,8 +940,8 @@ static inline void oti_free_cookies(struct obd_trans_info *oti)
         if (oti->oti_logcookies == &oti->oti_onecookie)
                 LASSERT(oti->oti_numcookies == 1);
         else
-                OBD_FREE(oti->oti_logcookies,
-                         oti->oti_numcookies * sizeof(oti->oti_onecookie));
+                OBD_FREE_LARGE(oti->oti_logcookies,
+                               oti->oti_numcookies*sizeof(oti->oti_onecookie));
         oti->oti_logcookies = NULL;
         oti->oti_numcookies = 0;
 }
@@ -917,6 +950,8 @@ static inline void oti_free_cookies(struct obd_trans_info *oti)
  * Events signalled through obd_notify() upcall-chain.
  */
 enum obd_notify_event {
+        /* target added */
+        OBD_NOTIFY_CREATE,
         /* Device connect start */
         OBD_NOTIFY_CONNECT,
         /* Device activated */
@@ -933,7 +968,10 @@ enum obd_notify_event {
         /* Configuration event */
         OBD_NOTIFY_CONFIG,
         /* Trigger quota recovery */
-        OBD_NOTIFY_QUOTA
+        OBD_NOTIFY_QUOTA,
+        /* Administratively deactivate/activate event */
+        OBD_NOTIFY_DEACTIVATE,
+        OBD_NOTIFY_ACTIVATE
 };
 
 /* bit-mask flags for config events */
@@ -957,15 +995,8 @@ struct obd_notify_upcall {
 struct target_recovery_data {
         svc_handler_t     trd_recovery_handler;
         pid_t             trd_processing_task;
-        struct completion trd_starting;
-        struct completion trd_finishing;
-};
-
-enum filter_groups {
-        FILTER_GROUP_MDS0 = 0,
-        FILTER_GROUP_LLOG = 1,
-        FILTER_GROUP_ECHO = 2 ,
-        FILTER_GROUP_MDS1_N_BASE = 3
+        cfs_completion_t  trd_starting;
+        cfs_completion_t  trd_finishing;
 };
 
 /**
@@ -978,60 +1009,42 @@ enum filter_groups {
   * 2. The group number indexing starts from 0 instead of 3
   */
 
-static inline int filter_group_is_mds(obd_gr group)
-{
-        return (group == FILTER_GROUP_MDS0 ||
-                group >= FILTER_GROUP_MDS1_N_BASE);
-}
-
-#define LASSERT_MDS_GROUP(group) LASSERT(filter_group_is_mds(group))
+#define LASSERT_SEQ_IS_MDT(seq) LASSERT(fid_seq_is_mdt(seq))
 
-static inline __u64 objgrp_to_mdsno(obd_gr group)
+static inline __u64 objseq_to_mdsno(obd_seq seq)
 {
-        LASSERT(filter_group_is_mds(group));
-        if (group == FILTER_GROUP_MDS0)
+        LASSERT_SEQ_IS_MDT(seq);
+        if (seq == FID_SEQ_OST_MDT0)
                 return 0;
-        return group - FILTER_GROUP_MDS1_N_BASE + 1;
+        return seq - FID_SEQ_OST_MDT1 + 1;
 }
 
-static inline int mdt_to_obd_objgrp(int mdtid)
+static inline int mdt_to_obd_objseq(int mdtid)
 {
         /**
-         * MDS0 uses group 0 always, other MDSes will use groups from
-         * FILTER_GROUP_MDS1_N_BASE
+         * MDS0 uses seq 0 pre FID-on-OST, other MDSes will use seq from
+         * FID_SEQ_OST_MDT1
          */
         if (mdtid)
-                return FILTER_GROUP_MDS1_N_BASE + mdtid - 1;
+                return FID_SEQ_OST_MDT1 + mdtid - 1;
         return 0;
 }
 
-static inline __u64 obdo_mdsno(struct obdo *oa)
-{
-        LASSERT((oa->o_valid & OBD_MD_FLGROUP));
-        return objgrp_to_mdsno(oa->o_gr);
-}
-
-static inline int obdo_is_mds(struct obdo *oa)
-{
-        LASSERT(oa->o_valid & OBD_MD_FLGROUP);
-        return filter_group_is_mds(oa->o_gr);
-}
-
 struct obd_llog_group {
-        struct list_head   olg_list;
-        int                olg_group;
+        cfs_list_t         olg_list;
+        int                olg_seq;
         struct llog_ctxt  *olg_ctxts[LLOG_MAX_CTXTS];
         cfs_waitq_t        olg_waitq;
-        spinlock_t         olg_lock;
+        cfs_spinlock_t     olg_lock;
         struct obd_export *olg_exp;
         int                olg_initializing;
-        struct semaphore   olg_cat_processing;
+        cfs_mutex_t        olg_cat_processing;
 };
 
 /* corresponds to one of the obd's */
-#define MAX_OBD_NAME 128
 #define OBD_DEVICE_MAGIC        0XAB5CD6EF
 #define OBD_DEV_BY_DEVNAME      0xffffd0de
+
 struct obd_device {
         struct obd_type        *obd_type;
         __u32                   obd_magic;
@@ -1043,16 +1056,15 @@ struct obd_device {
         struct lu_device       *obd_lu_dev;
 
         int                     obd_minor;
+        /* bitfield modification is protected by obd_dev_lock */
         unsigned long obd_attached:1,      /* finished attach */
                       obd_set_up:1,        /* finished setup */
                       obd_recovering:1,    /* there are recoverable clients */
                       obd_abort_recovery:1,/* recovery expired */
                       obd_version_recov:1, /* obd uses version checking */
-                      obd_recovery_expired:1,
                       obd_replayable:1,    /* recovery is enabled; inform clients */
                       obd_no_transno:1,    /* no committed-transno notification */
                       obd_no_recov:1,      /* fail instead of retry messages */
-                      obd_req_replaying:1, /* replaying requests */
                       obd_stopping:1,      /* started cleanup */
                       obd_starting:1,      /* started setup */
                       obd_force:1,         /* cleanup with > 0 obd refcount */
@@ -1061,63 +1073,73 @@ struct obd_device {
                       obd_no_conn:1,       /* deny new connections */
                       obd_inactive:1,      /* device active/inactive
                                            * (for /proc/status only!!) */
+                      obd_no_ir:1,         /* no imperative recovery. */
                       obd_process_conf:1;  /* device is processing mgs config */
+        /* use separate field as it is set in interrupt to don't mess with
+         * protection of other bits using _bh lock */
+        unsigned long obd_recovery_expired:1;
         /* uuid-export hash body */
-        struct lustre_hash     *obd_uuid_hash;
+        cfs_hash_t             *obd_uuid_hash;
         /* nid-export hash body */
-        struct lustre_hash     *obd_nid_hash;
+        cfs_hash_t             *obd_nid_hash;
         /* nid stats body */
-        struct lustre_hash     *obd_nid_stats_hash;
-        struct list_head        obd_nid_stats;
-        atomic_t                obd_refcount;
+        cfs_hash_t             *obd_nid_stats_hash;
+        cfs_list_t              obd_nid_stats;
+        cfs_atomic_t            obd_refcount;
         cfs_waitq_t             obd_refcount_waitq;
-        struct list_head        obd_exports;
-        struct list_head        obd_unlinked_exports;
-        struct list_head        obd_delayed_exports;
+        cfs_list_t              obd_exports;
+        cfs_list_t              obd_unlinked_exports;
+        cfs_list_t              obd_delayed_exports;
         int                     obd_num_exports;
-        spinlock_t              obd_nid_lock;
+        cfs_spinlock_t          obd_nid_lock;
         struct ldlm_namespace  *obd_namespace;
         struct ptlrpc_client    obd_ldlm_client; /* XXX OST/MDS only */
         /* a spinlock is OK for what we do now, may need a semaphore later */
-        spinlock_t              obd_dev_lock;
-        struct semaphore        obd_dev_sem;
+        cfs_spinlock_t          obd_dev_lock; /* protects obd bitfield above */
+        cfs_mutex_t             obd_dev_mutex;
         __u64                   obd_last_committed;
         struct fsfilt_operations *obd_fsops;
-        spinlock_t              obd_osfs_lock;
+        cfs_spinlock_t          obd_osfs_lock;
         struct obd_statfs       obd_osfs;       /* locked by obd_osfs_lock */
         __u64                   obd_osfs_age;
         struct lvfs_run_ctxt    obd_lvfs_ctxt;
         struct obd_llog_group   obd_olg; /* default llog group */
-        struct obd_device       *obd_observer;
+        struct obd_device      *obd_observer;
+        cfs_rw_semaphore_t      obd_observer_link_sem;
         struct obd_notify_upcall obd_upcall;
         struct obd_export       *obd_self_export;
         /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
-        struct list_head        obd_exports_timed;
+        cfs_list_t              obd_exports_timed;
         time_t                  obd_eviction_timer; /* for ping evictor */
 
         int                              obd_max_recoverable_clients;
-        int                              obd_connected_clients;
+        cfs_atomic_t                     obd_connected_clients;
         int                              obd_stale_clients;
         int                              obd_delayed_clients;
-        spinlock_t                       obd_processing_task_lock; /* BH lock (timer) */
+        /* this lock protects all recovery list_heads, timer and
+         * obd_next_recovery_transno value */
+        cfs_spinlock_t                   obd_recovery_task_lock;
         __u64                            obd_next_recovery_transno;
         int                              obd_replayed_requests;
         int                              obd_requests_queued_for_recovery;
         cfs_waitq_t                      obd_next_transno_waitq;
+        /* protected by obd_recovery_task_lock */
         cfs_timer_t                      obd_recovery_timer;
         time_t                           obd_recovery_start; /* seconds */
         time_t                           obd_recovery_end; /* seconds, for lprocfs_status */
-        time_t                           obd_recovery_max_time; /* seconds, bz13079 */
+        int                              obd_recovery_time_hard;
         int                              obd_recovery_timeout;
+        int                              obd_recovery_ir_factor;
 
         /* new recovery stuff from CMD2 */
         struct target_recovery_data      obd_recovery_data;
         int                              obd_replayed_locks;
-        atomic_t                         obd_req_replay_clients;
-        atomic_t                         obd_lock_replay_clients;
-        struct list_head                 obd_req_replay_queue;
-        struct list_head                 obd_lock_replay_queue;
-        struct list_head                 obd_final_req_queue;
+        cfs_atomic_t                     obd_req_replay_clients;
+        cfs_atomic_t                     obd_lock_replay_clients;
+        /* all lists are protected by obd_recovery_task_lock */
+        cfs_list_t                       obd_req_replay_queue;
+        cfs_list_t                       obd_lock_replay_queue;
+        cfs_list_t                       obd_final_req_queue;
         int                              obd_recovery_stage;
 
         union {
@@ -1143,13 +1165,14 @@ struct obd_device {
         cfs_proc_dir_entry_t  *obd_proc_exports_entry;
         cfs_proc_dir_entry_t  *obd_svc_procroot;
         struct lprocfs_stats  *obd_svc_stats;
-        atomic_t               obd_evict_inprogress;
+        cfs_atomic_t           obd_evict_inprogress;
         cfs_waitq_t            obd_evict_inprogress_waitq;
+        cfs_list_t             obd_evict_list; /* protected with pet_lock */
 
         /**
          * Ldlm pool part. Save last calculated SLV and Limit.
          */
-        rwlock_t               obd_pool_lock;
+        cfs_rwlock_t           obd_pool_lock;
         int                    obd_pool_limit;
         __u64                  obd_pool_slv;
 
@@ -1161,6 +1184,7 @@ struct obd_device {
 };
 
 #define OBD_LLOG_FL_SENDNOW     0x0001
+#define OBD_LLOG_FL_EXIT       0x0002
 
 enum obd_cleanup_stage {
 /* Special case hack for MDS LOVs */
@@ -1199,19 +1223,34 @@ enum obd_cleanup_stage {
 #define KEY_REGISTER_TARGET     "register_target"
 #define KEY_REVIMP_UPD          "revimp_update"
 #define KEY_SET_FS              "set_fs"
+#define KEY_TGT_COUNT           "tgt_count"
 /*      KEY_SET_INFO in lustre_idl.h */
 #define KEY_SPTLRPC_CONF        "sptlrpc_conf"
-#define KEY_UNLINKED            "unlinked"
-
+#define KEY_CONNECT_FLAG        "connect_flags"
+#define KEY_SYNC_LOCK_CANCEL    "sync_lock_cancel"
 
 struct lu_context;
 
+/* /!\ must be coherent with include/linux/namei.h on patched kernel */
+#define IT_OPEN     (1 << 0)
+#define IT_CREAT    (1 << 1)
+#define IT_READDIR  (1 << 2)
+#define IT_GETATTR  (1 << 3)
+#define IT_LOOKUP   (1 << 4)
+#define IT_UNLINK   (1 << 5)
+#define IT_TRUNC    (1 << 6)
+#define IT_GETXATTR (1 << 7)
+#define IT_EXEC     (1 << 8)
+#define IT_PIN      (1 << 9)
+#define IT_LAYOUT   (1 << 10)
+
 static inline int it_to_lock_mode(struct lookup_intent *it)
 {
         /* CREAT needs to be tested before open (both could be set) */
         if (it->it_op & IT_CREAT)
                 return LCK_CW;
-        else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
+        else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP |
+                              IT_LAYOUT))
                 return LCK_CR;
 
         LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
@@ -1225,7 +1264,7 @@ struct md_op_data {
         struct lu_fid           op_fid4; /* to the operation locks. */
         mdsno_t                 op_mds;  /* what mds server open will go to */
         struct lustre_handle    op_handle;
-        __u64                   op_mod_time;
+        obd_time                op_mod_time;
         const char             *op_name;
         int                     op_namelen;
         __u32                   op_mode;
@@ -1244,6 +1283,7 @@ struct md_op_data {
         unsigned int            op_attr_flags;
 #endif
 #endif
+        __u64                   op_valid;
         loff_t                  op_attr_blocks;
 
         /* Size-on-MDS epoch and flags. */
@@ -1259,6 +1299,10 @@ struct md_op_data {
 
         /* Operation type */
         __u32                   op_opc;
+
+        /* Used by readdir */
+        __u32                   op_npages;
+        __u64                   op_offset;
 };
 
 struct md_enqueue_info;
@@ -1271,20 +1315,21 @@ struct md_enqueue_info {
         struct md_op_data       mi_data;
         struct lookup_intent    mi_it;
         struct lustre_handle    mi_lockh;
-        struct dentry          *mi_dentry;
         struct inode           *mi_dir;
         md_enqueue_cb_t         mi_cb;
+        __u64                   mi_cbdata;
         unsigned int            mi_generation;
-        void                   *mi_cbdata;
 };
 
 struct obd_ops {
-        struct module *o_owner;
+        cfs_module_t *o_owner;
         int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
                            void *karg, void *uarg);
-        int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
-                          __u32 *vallen, void *val, struct lov_stripe_md *lsm);
-        int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
+        int (*o_get_info)(const struct lu_env *env, struct obd_export *,
+                          __u32 keylen, void *key, __u32 *vallen, void *val,
+                          struct lov_stripe_md *lsm);
+        int (*o_set_info_async)(const struct lu_env *, struct obd_export *,
+                                __u32 keylen, void *key,
                                 __u32 vallen, void *val,
                                 struct ptlrpc_request_set *set);
         int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
@@ -1328,34 +1373,35 @@ struct obd_ops {
          */
         int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
 
-        int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
-                        __u64 max_age, __u32 flags);
-        int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
+        int (*o_statfs)(const struct lu_env *, struct obd_export *exp,
+                        struct obd_statfs *osfs, __u64 max_age, __u32 flags);
+        int (*o_statfs_async)(struct obd_export *exp, struct obd_info *oinfo,
                               __u64 max_age, struct ptlrpc_request_set *set);
         int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
                         struct lov_stripe_md *mem_src);
         int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
                           struct lov_mds_md *disk_src, int disk_len);
-        int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
-                         struct lov_stripe_md *mem_tgt);
         int (*o_preallocate)(struct lustre_handle *, obd_count *req,
                              obd_id *ids);
         /* FIXME: add fid capability support for create & destroy! */
         int (*o_precreate)(struct obd_export *exp);
-        int (*o_create)(struct obd_export *exp,  struct obdo *oa,
-                        struct lov_stripe_md **ea, struct obd_trans_info *oti);
+        int (*o_create)(const struct lu_env *env, struct obd_export *exp,
+                        struct obdo *oa, struct lov_stripe_md **ea,
+                        struct obd_trans_info *oti);
         int (*o_create_async)(struct obd_export *exp,  struct obd_info *oinfo,
                               struct lov_stripe_md **ea,
                               struct obd_trans_info *oti);
-        int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
-                         struct lov_stripe_md *ea, struct obd_trans_info *oti,
-                         struct obd_export *md_exp, void *capa);
-        int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
-                         struct obd_trans_info *oti);
+        int (*o_destroy)(const struct lu_env *env, struct obd_export *exp,
+                         struct obdo *oa, struct lov_stripe_md *ea,
+                         struct obd_trans_info *oti, struct obd_export *md_exp,
+                         void *capa);
+        int (*o_setattr)(const struct lu_env *, struct obd_export *exp,
+                         struct obd_info *oinfo, struct obd_trans_info *oti);
         int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
                                struct obd_trans_info *oti,
                                struct ptlrpc_request_set *rqset);
-        int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
+        int (*o_getattr)(const struct lu_env *env, struct obd_export *exp,
+                         struct obd_info *oinfo);
         int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
                                struct ptlrpc_request_set *set);
         int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
@@ -1365,12 +1411,12 @@ struct obd_ops {
                            struct ost_lvb *lvb, int kms_only);
         int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
                             obd_off size, int shrink);
-        int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
-                       struct obd_trans_info *oti,
+        int (*o_punch)(const struct lu_env *, struct obd_export *exp,
+                       struct obd_info *oinfo, struct obd_trans_info *oti,
                        struct ptlrpc_request_set *rqset);
-        int (*o_sync)(struct obd_export *exp, struct obdo *oa,
-                      struct lov_stripe_md *ea, obd_size start, obd_size end,
-                      void *capa);
+        int (*o_sync)(const struct lu_env *env, struct obd_export *exp,
+                      struct obd_info *oinfo, obd_size start, obd_size end,
+                      struct ptlrpc_request_set *set);
         int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
                          struct lov_stripe_md *src, obd_size start,
                          obd_size end, struct obd_trans_info *oti);
@@ -1378,15 +1424,15 @@ struct obd_ops {
                       struct lustre_handle *srconn, struct lov_stripe_md *src,
                       obd_size start, obd_size end, struct obd_trans_info *);
         int (*o_iterate)(struct lustre_handle *conn,
-                         int (*)(obd_id, obd_gr, void *),
-                         obd_id *startid, obd_gr group, void *data);
-        int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
-                        int objcount, struct obd_ioobj *obj,
-                        struct niobuf_remote *remote, int *nr_pages,
-                        struct niobuf_local *local,
-                        struct obd_trans_info *oti,
-                        struct lustre_capa *capa);
-        int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
+                         int (*)(obd_id, obd_seq, void *),
+                         obd_id *startid, obd_seq seq, void *data);
+        int (*o_preprw)(const struct lu_env *env, int cmd,
+                        struct obd_export *exp, struct obdo *oa, int objcount,
+                        struct obd_ioobj *obj, struct niobuf_remote *remote,
+                        int *nr_pages, struct niobuf_local *local,
+                        struct obd_trans_info *oti, struct lustre_capa *capa);
+        int (*o_commitrw)(const struct lu_env *env, int cmd,
+                          struct obd_export *exp, struct obdo *oa,
                           int objcount, struct obd_ioobj *obj,
                           struct niobuf_remote *remote, int pages,
                           struct niobuf_local *local,
@@ -1396,10 +1442,12 @@ struct obd_ops {
                          struct ptlrpc_request_set *rqset);
         int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
                                ldlm_iterator_t it, void *data);
+        int (*o_find_cbdata)(struct obd_export *, struct lov_stripe_md *,
+                             ldlm_iterator_t it, void *data);
         int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
                         __u32 mode, struct lustre_handle *);
         int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
-                               int flags, void *opaque);
+                               ldlm_cancel_flags_t flags, void *opaque);
         int (*o_init_export)(struct obd_export *exp);
         int (*o_destroy_export)(struct obd_export *exp);
         int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
@@ -1422,7 +1470,7 @@ struct obd_ops {
         int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
                         enum obd_notify_event ev, void *data);
 
-        int (*o_health_check)(struct obd_device *);
+        int (*o_health_check)(const struct lu_env *env, struct obd_device *);
         struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
 
         /* quota methods */
@@ -1432,10 +1480,11 @@ struct obd_ops {
                           struct obd_quotactl *);
         int (*o_quota_adjust_qunit)(struct obd_export *exp,
                                     struct quota_adjust_qunit *oqaq,
-                                    struct lustre_quota_ctxt *qctxt);
+                                    struct lustre_quota_ctxt *qctxt,
+                                    struct ptlrpc_request_set *rqset);
 
 
-        int (*o_ping)(struct obd_export *exp);
+        int (*o_ping)(const struct lu_env *, struct obd_export *exp);
 
         /* pools methods */
         int (*o_pool_new)(struct obd_device *obd, char *poolname);
@@ -1485,6 +1534,7 @@ struct md_open_data {
         struct obd_client_handle *mod_och;
         struct ptlrpc_request    *mod_open_req;
         struct ptlrpc_request    *mod_close_req;
+        cfs_atomic_t              mod_refcount;
 };
 
 struct lookup_intent;
@@ -1494,6 +1544,8 @@ struct md_ops {
                            struct obd_capa **);
         int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
                                ldlm_iterator_t, void *);
+        int (*m_find_cbdata)(struct obd_export *, const struct lu_fid *,
+                             ldlm_iterator_t, void *);
         int (*m_close)(struct obd_export *, struct md_op_data *,
                        struct md_open_data *, struct ptlrpc_request **);
         int (*m_create)(struct obd_export *, struct md_op_data *,
@@ -1505,12 +1557,10 @@ struct md_ops {
                          struct lookup_intent *, struct md_op_data *,
                          struct lustre_handle *, void *, int,
                          struct ptlrpc_request **, int);
-        int (*m_getattr)(struct obd_export *, const struct lu_fid *,
-                         struct obd_capa *, obd_valid, int,
+        int (*m_getattr)(struct obd_export *, struct md_op_data *,
                          struct ptlrpc_request **);
-        int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
-                              struct obd_capa *, const char *, int, obd_valid,
-                              int, __u32, struct ptlrpc_request **);
+        int (*m_getattr_name)(struct obd_export *, struct md_op_data *,
+                              struct ptlrpc_request **);
         int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
                              void *, int, struct lookup_intent *, int,
                              struct ptlrpc_request **,
@@ -1528,9 +1578,8 @@ struct md_ops {
                          struct md_open_data **mod);
         int (*m_sync)(struct obd_export *, const struct lu_fid *,
                       struct obd_capa *, struct ptlrpc_request **);
-        int (*m_readpage)(struct obd_export *, const struct lu_fid *,
-                          struct obd_capa *, __u64, struct page *,
-                          struct ptlrpc_request **);
+        int (*m_readpage)(struct obd_export *, struct md_op_data *,
+                          struct page **, struct ptlrpc_request **);
 
         int (*m_unlink)(struct obd_export *, struct md_op_data *,
                         struct ptlrpc_request **);
@@ -1558,7 +1607,7 @@ struct md_ops {
                                       struct ptlrpc_request *);
         int (*m_clear_open_replay_data)(struct obd_export *,
                                         struct obd_client_handle *);
-        int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u32 *);
+        int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u64 *);
 
         ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
                                     const struct lu_fid *, ldlm_type_t,
@@ -1566,8 +1615,8 @@ struct md_ops {
                                     struct lustre_handle *);
 
         int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
-                               ldlm_policy_data_t *, ldlm_mode_t, int flags,
-                               void *opaque);
+                               ldlm_policy_data_t *, ldlm_mode_t,
+                               ldlm_cancel_flags_t flags, void *opaque);
         int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
                             renew_capa_cb_t cb);
         int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
@@ -1581,9 +1630,8 @@ struct md_ops {
                                       struct md_enqueue_info *,
                                       struct ldlm_enqueue_info *);
 
-        int (*m_revalidate_lock)(struct obd_export *,
-                                 struct lookup_intent *,
-                                 struct lu_fid *);
+        int (*m_revalidate_lock)(struct obd_export *, struct lookup_intent *,
+                                 struct lu_fid *, __u64 *bits);
 
         /*
          * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
@@ -1600,26 +1648,19 @@ struct lsm_operations {
                                     obd_off *);
         void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
                                      obd_off *);
-        obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
-        obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
-        int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
-        int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
         int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
-                               int *stripe_count);
+                               __u16 *stripe_count);
         int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
                              struct lov_mds_md *lmm);
 };
 
 extern const struct lsm_operations lsm_v1_ops;
-extern const struct lsm_operations lsm_join_ops;
 extern const struct lsm_operations lsm_v3_ops;
 static inline const struct lsm_operations *lsm_op_find(int magic)
 {
         switch(magic) {
         case LOV_MAGIC_V1:
                return &lsm_v1_ops;
-        case LOV_MAGIC_JOIN:
-               return &lsm_join_ops;
         case LOV_MAGIC_V3:
                return &lsm_v3_ops;
         default:
@@ -1672,4 +1713,46 @@ static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
         return oinfo->oi_capa;
 }
 
+static inline struct md_open_data *obd_mod_alloc(void)
+{
+        struct md_open_data *mod;
+        OBD_ALLOC_PTR(mod);
+        if (mod == NULL)
+                return NULL;
+        cfs_atomic_set(&mod->mod_refcount, 1);
+        return mod;
+}
+
+#define obd_mod_get(mod) cfs_atomic_inc(&(mod)->mod_refcount)
+#define obd_mod_put(mod)                                        \
+({                                                              \
+        if (cfs_atomic_dec_and_test(&(mod)->mod_refcount)) {          \
+                if ((mod)->mod_open_req)                          \
+                        ptlrpc_req_finished((mod)->mod_open_req);   \
+                OBD_FREE_PTR(mod);                              \
+        }                                                       \
+})
+
+void obdo_from_inode(struct obdo *dst, struct inode *src, obd_flag valid);
+void obdo_set_parent_fid(struct obdo *dst, const struct lu_fid *parent);
+
+/* return 1 if client should be resend request */
+static inline int client_should_resend(int resend, struct client_obd *cli)
+{
+        return cfs_atomic_read(&cli->cl_resends) ?
+               cfs_atomic_read(&cli->cl_resends) > resend : 1;
+}
+
+/**
+ * Return device name for this device
+ *
+ * XXX: lu_device is declared before obd_device, while a pointer pointing
+ * back to obd_device in lu_device, so this helper function defines here
+ * instead of in lu_object.h
+ */
+static inline const char *lu_dev_name(const struct lu_device *lu_dev)
+{
+        return lu_dev->ld_obd->obd_name;
+}
+
 #endif /* __OBD_H */