Whamcloud - gitweb
b=22176 Add .sync_fs super block handler
[fs/lustre-release.git] / lustre / include / obd.h
index 12daa90..46f8bb7 100644 (file)
@@ -26,7 +26,7 @@
  * 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.
  */
 /*
 
 #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_object.h>
+#include <lu_target.h>
 #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>
 
@@ -76,9 +72,9 @@
 
 /* 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;
+        cfs_list_t              lop_pending;
+        cfs_list_t              lop_urgent;
+        cfs_list_t              lop_pending_group;
         int                     lop_num_pending;
 };
 
@@ -89,24 +85,25 @@ 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;
-        /* _cli_ is poorly named, it should be _ready_ */
-        struct list_head loi_cli_item;
-        struct list_head loi_write_item;
-        struct list_head loi_read_item;
-
+        cfs_list_t loi_ready_item;
+        cfs_list_t loi_hp_ready_item;
+        cfs_list_t loi_write_item;
+        cfs_list_t loi_read_item;
+        cfs_list_t loi_sync_fs_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)
 {
@@ -122,34 +119,21 @@ static inline void loi_init(struct lov_oinfo *loi)
         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_cli_item);
+        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);
+        CFS_INIT_LIST_HEAD(&loi->loi_sync_fs_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 */
 
         struct {
                 /* Public members. */
                 __u64 lw_object_id;        /* lov object id */
-                __u64 lw_object_gr;        /* lov object group */
+                __u64 lw_object_seq;       /* lov object seq */
                 __u64 lw_maxbytes;         /* maximum possible file size */
 
                 /* LOV-private members start here -- only for use in lov/. */
@@ -160,12 +144,11 @@ struct lov_stripe_md {
                 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_object_seq   lsm_wire.lw_object_seq
 #define lsm_maxbytes     lsm_wire.lw_maxbytes
 #define lsm_magic        lsm_wire.lw_magic
 #define lsm_stripe_size  lsm_wire.lw_stripe_size
@@ -177,6 +160,12 @@ struct obd_info;
 
 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
 
+struct osc_sync_fs_wait {
+        struct obd_info      *sfw_oi;
+        obd_enqueue_update_f  sfw_upcall;
+        int started;
+};
+
 /* obd info for a particular level (lov, osc). */
 struct obd_info {
         /* Lock policy. It keeps an extent which is specific for a particular
@@ -222,14 +211,14 @@ 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 {
@@ -243,13 +232,21 @@ struct brw_page {
 
 struct ost_server_data;
 
+#define OBT_MAGIC       0xBDDECEAE
 /* hold common fields for "target" device */
 struct obd_device_target {
+        __u32                     obt_magic;
         struct super_block       *obt_sb;
-        atomic_t                  obt_quotachecking;
+        /** last_rcvd file */
+        struct file              *obt_rcvd_filp;
+        struct lu_target         *obt_lut;
+        __u64                     obt_mount_count;
+        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;
 };
 
 /* llog contexts */
@@ -266,7 +263,9 @@ enum llog_ctxt_id {
         LLOG_TEST_REPL_CTXT,
         LLOG_LOVEA_ORIG_CTXT,
         LLOG_LOVEA_REPL_CTXT,
-        LLOG_CHANGELOG_ORIG_CTXT,   /**< changelog context */
+        LLOG_CHANGELOG_ORIG_CTXT,      /**< changelog generation on mdd */
+        LLOG_CHANGELOG_REPL_CTXT,      /**< changelog access on clients */
+        LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */
         LLOG_MAX_CTXTS
 };
 
@@ -286,37 +285,33 @@ struct filter_obd {
         /* NB this field MUST be first */
         struct obd_device_target fo_obt;
         const char          *fo_fstype;
-        struct vfsmount     *fo_vfsmnt;
 
         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_semaphore_t      fo_init_lock;      /* group initialization lock */
         int                  fo_committed_group;
 
-
-        spinlock_t           fo_objidlock;      /* protect fo_lastobjid */
-        spinlock_t           fo_translock;      /* protect fsd_last_transno */
-        struct file         *fo_rcvd_filp;
-        struct file         *fo_health_check_filp;
-        struct lr_server_data *fo_fsd;
-        unsigned long       *fo_last_rcvd_slots;
-        __u64                fo_mount_count;
+        cfs_spinlock_t       fo_objidlock;      /* protect fo_lastobjid */
 
         unsigned long        fo_destroys_in_progress;
-        struct semaphore     fo_create_locks[FILTER_SUBDIR_COUNT];
+        cfs_semaphore_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 */
         obd_size             fo_tot_granted;    /* all values in bytes */
         obd_size             fo_tot_pending;
+        int                  fo_tot_granted_clients;
 
         obd_size             fo_readcache_max_filesize;
-        int                  fo_read_cache;
-        int                  fo_writethrough_cache;
+        cfs_spinlock_t       fo_flags_lock;
+        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 */
 
         struct obd_import   *fo_mdc_imp;
         struct obd_uuid      fo_mdc_uuid;
@@ -325,10 +320,10 @@ struct filter_obd {
         __u64               *fo_last_objids; /* last created objid for groups,
                                               * protected by fo_objidlock */
 
-        struct semaphore     fo_alloc_lock;
+        cfs_semaphore_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
@@ -345,42 +340,62 @@ 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;
 };
 
+struct timeout_item {
+        enum timeout_event ti_event;
+        cfs_time_t         ti_timeout;
+        timeout_cb_t       ti_cb;
+        void              *ti_cb_data;
+        cfs_list_t         ti_obd_list;
+        cfs_list_t         ti_chain;
+};
+
 #define OSC_MAX_RIF_DEFAULT       8
 #define OSC_MAX_RIF_MAX         256
 #define OSC_MAX_DIRTY_DEFAULT  (OSC_MAX_RIF_DEFAULT * 4)
 #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;
@@ -403,7 +418,11 @@ 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_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 */
+        cfs_semaphore_t          cl_grant_sem;   /*grant shrink list cfs_semaphore*/
+        int                      cl_grant_shrink_interval; /* seconds */
 
         /* keep track of objects that have lois that contain pages which
          * have been queued for async brw.  this lock also protects the
@@ -424,9 +443,11 @@ struct client_obd {
          * client_obd_list_lock_{init,done}() functions.
          */
         client_obd_lock_t        cl_loi_list_lock;
-        struct list_head         cl_loi_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;
+        cfs_list_t               cl_loi_sync_fs_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 */
@@ -442,7 +463,7 @@ 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;
@@ -451,10 +472,10 @@ struct client_obd {
         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 */
@@ -473,20 +494,23 @@ struct client_obd {
         /* sequence manager */
         struct lu_client_seq    *cl_seq;
 
-        atomic_t                 cl_resends; /* resend count */
+        cfs_atomic_t             cl_resends; /* resend count */
+        struct osc_sync_fs_wait  cl_sf_wait;
 };
+
 #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_semaphore_t                  mgs_sem;
         cfs_proc_dir_entry_t            *mgs_proc_live;
 };
 
@@ -496,32 +520,26 @@ struct mds_obd {
         struct ptlrpc_service           *mds_service;
         struct ptlrpc_service           *mds_setattr_service;
         struct ptlrpc_service           *mds_readpage_service;
-        struct vfsmount                 *mds_vfsmnt;
         cfs_dentry_t                    *mds_fid_de;
         int                              mds_max_mdsize;
         int                              mds_max_cookiesize;
-        struct file                     *mds_rcvd_filp;
-        spinlock_t                       mds_transno_lock;
-        __u64                            mds_last_transno;
-        __u64                            mds_mount_count;
         __u64                            mds_io_epoch;
         unsigned long                    mds_atime_diff;
-        struct semaphore                 mds_epoch_sem;
+        cfs_semaphore_t                  mds_epoch_sem;
         struct ll_fid                    mds_rootfid;
-        struct lr_server_data           *mds_server_data;
         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 */
@@ -531,13 +549,10 @@ struct mds_obd {
         __u32                            mds_lov_objid_lastpage;
         __u32                            mds_lov_objid_lastidx;
 
-        struct file                     *mds_health_check_filp;
-        unsigned long                   *mds_client_bitmap;
-//        struct upcall_cache             *mds_group_hash;
 
         struct lustre_quota_info         mds_quota_info;
-        struct semaphore                 mds_qonoff_sem;
-        struct semaphore                 mds_health_sem;
+        cfs_rw_semaphore_t               mds_qonoff_sem;
+        cfs_semaphore_t                  mds_health_sem;
         unsigned long                    mds_fl_user_xattr:1,
                                          mds_fl_acl:1,
                                          mds_evict_ost_nids:1,
@@ -551,7 +566,7 @@ 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;
 };
 
 /* lov objid */
@@ -575,34 +590,34 @@ struct obd_id_info {
 
 struct echo_obd {
         struct obdo          eo_oa;
-        spinlock_t           eo_lock;
+        cfs_spinlock_t       eo_lock;
         __u64                eo_lastino;
         struct lustre_handle eo_nl_lock;
-        atomic_t             eo_prep;
+        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_semaphore_t        ost_health_sem;
 };
 
 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 */
 };
@@ -618,11 +633,11 @@ 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 */
@@ -634,27 +649,43 @@ 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 */
         struct lov_qos_rr   lq_rr;          /* round robin qos data */
         unsigned long       lq_dirty:1,     /* recalc qos data */
                             lq_same_space:1,/* the ost's all have approx.
                                                the same space avail */
-                            lq_reset:1;     /* zero current penalties */
+                            lq_reset:1,     /* zero current penalties */
+                            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
+                                              * requests completion */
 };
 
 struct lov_tgt_desc {
+        cfs_list_t          ltd_kill;
         struct obd_uuid     ltd_uuid;
+        struct obd_device  *ltd_obd;
         struct obd_export  *ltd_exp;
         struct ltd_qos      ltd_qos;     /* qos info per target */
         __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 */
 };
 
@@ -668,10 +699,10 @@ 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 */
@@ -682,18 +713,18 @@ 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_semaphore_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;
 };
@@ -701,9 +732,9 @@ 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_semaphore_t         ltd_fid_sem;
 };
 
 enum placement_policy {
@@ -718,7 +749,7 @@ 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;
@@ -729,7 +760,7 @@ struct lmv_obd {
         int                     max_def_easize;
         int                     max_cookiesize;
         int                     server_timeout;
-        struct semaphore        init_sem;
+        cfs_semaphore_t         init_sem;
 
         struct lmv_tgt_desc     *tgts;
         int                     tgts_size;
@@ -755,7 +786,7 @@ 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_VVP_NAME         "vvp"
 #define LUSTRE_LMV_NAME         "lmv"
 #define LUSTRE_CMM_MDC_NAME     "cmm-mdc"
@@ -802,6 +833,8 @@ struct obd_trans_info {
         /* initial thread handling transaction */
         struct ptlrpc_thread *   oti_thread;
         __u32                    oti_conn_cnt;
+        /** VBR: versions */
+        __u64                    oti_pre_version;
 
         struct obd_uuid         *oti_ost_uuid;
 };
@@ -817,7 +850,15 @@ static inline void oti_init(struct obd_trans_info *oti,
                 return;
 
         oti->oti_xid = req->rq_xid;
+        /** VBR: take versions from request */
+        if (req->rq_reqmsg != NULL &&
+            lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
+                __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
+                oti->oti_pre_version = pre_version ? pre_version[0] : 0;
+                oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
+        }
 
+        /** called from mds_create_objects */
         if (req->rq_repmsg != NULL)
                 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
         oti->oti_thread = req->rq_svc_thread;
@@ -857,6 +898,10 @@ 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 */
         OBD_NOTIFY_ACTIVE,
         /* Device deactivated */
@@ -869,7 +914,9 @@ enum obd_notify_event {
         OBD_NOTIFY_SYNC_NONBLOCK,
         OBD_NOTIFY_SYNC,
         /* Configuration event */
-        OBD_NOTIFY_CONFIG
+        OBD_NOTIFY_CONFIG,
+        /* Trigger quota recovery */
+        OBD_NOTIFY_QUOTA
 };
 
 /* bit-mask flags for config events */
@@ -885,7 +932,7 @@ enum config_flags {
  */
 struct obd_notify_upcall {
         int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
-                          enum obd_notify_event ev, void *owner);
+                          enum obd_notify_event ev, void *owner, void *data);
         /* Opaque datum supplied by upper layer listener */
         void *onu_owner;
 };
@@ -893,31 +940,10 @@ 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;
 };
 
-static inline __u64 obdo_mdsno(struct obdo *oa)
-{
-        if (oa->o_gr)
-                return oa->o_gr - FILTER_GROUP_MDS1_N_BASE;
-        return 0;
-}
-
-static inline int mdt_to_obd_objgrp(int mdtid)
-{
-        if (mdtid)
-                return FILTER_GROUP_MDS1_N_BASE + mdtid;
-        return 0;
-}
-
 /**
   * In HEAD for CMD, the object is created in group number which is 3>=
   * or indexing starts from 3. To test this assertions are added to disallow
@@ -928,25 +954,43 @@ static inline int mdt_to_obd_objgrp(int mdtid)
   * 2. The group number indexing starts from 0 instead of 3
   */
 
-#define CHECK_MDS_GROUP(group)          (group == FILTER_GROUP_MDS0 || \
-                                         group > FILTER_GROUP_MDS1_N_BASE)
-#define LASSERT_MDS_GROUP(group)        LASSERT(CHECK_MDS_GROUP(group))
+#define LASSERT_SEQ_IS_MDT(seq) LASSERT(fid_seq_is_mdt(seq))
+
+static inline __u64 objseq_to_mdsno(obd_seq seq)
+{
+        LASSERT_SEQ_IS_MDT(seq);
+        if (seq == FID_SEQ_OST_MDT0)
+                return 0;
+        return seq - FID_SEQ_OST_MDT1 + 1;
+}
+
+static inline int mdt_to_obd_objseq(int mdtid)
+{
+        /**
+         * MDS0 uses seq 0 pre FID-on-OST, other MDSes will use seq from
+         * FID_SEQ_OST_MDT1
+         */
+        if (mdtid)
+                return FID_SEQ_OST_MDT1 + mdtid - 1;
+        return 0;
+}
 
 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_semaphore_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;
@@ -958,81 +1002,88 @@ 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,/* somebody ioctl'ed us to abort */
+                      obd_abort_recovery:1,/* recovery expired */
+                      obd_version_recov:1, /* obd uses version checking */
                       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 */
                       obd_fail:1,          /* cleanup with failover */
-                      obd_async_recov:1,   /* allow asyncronous orphan cleanup */
+                      obd_async_recov:1,   /* allow asynchronous orphan cleanup */
                       obd_no_conn:1,       /* deny new connections */
-                      obd_inactive:1;      /* device active/inactive
+                      obd_inactive:1,      /* device active/inactive
                                            * (for /proc/status only!!) */
+                      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;
+        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_semaphore_t         obd_dev_sem;
         __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 */
 
-        /* XXX encapsulate all this recovery data into one struct */
-        svc_handler_t                    obd_recovery_handler;
-        pid_t                            obd_processing_task;
-
         int                              obd_max_recoverable_clients;
         int                              obd_connected_clients;
-        int                              obd_recoverable_clients;
-        spinlock_t                       obd_processing_task_lock; /* BH lock (timer) */
+        int                              obd_stale_clients;
+        int                              obd_delayed_clients;
+        /* 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;
-        struct list_head                 obd_uncommitted_replies;
-        spinlock_t                       obd_uncommitted_replies_lock;
+        /* 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 */
+        time_t                           obd_recovery_time_hard;
         int                              obd_recovery_timeout;
 
         /* 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 {
@@ -1058,13 +1109,14 @@ 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;
 
@@ -1075,9 +1127,6 @@ struct obd_device {
         struct lu_ref          obd_reference;
 };
 
-#define OBD_OPT_FORCE           0x0001
-#define OBD_OPT_FAILOVER        0x0002
-
 #define OBD_LLOG_FL_SENDNOW     0x0001
 
 enum obd_cleanup_stage {
@@ -1088,35 +1137,40 @@ enum obd_cleanup_stage {
 };
 
 /* get/set_info keys */
-#define KEY_READ_ONLY           "read-only"
-#define KEY_MDS_CONN            "mds_conn"
-#define KEY_NEXT_ID             "next_id"
-#define KEY_LOVDESC             "lovdesc"
-#define KEY_INIT_RECOV          "initial_recov"
-#define KEY_INIT_RECOV_BACKUP   "init_recov_bk"
-#define KEY_FLUSH_CTX           "flush_ctx"
+#define KEY_ASYNC               "async"
+#define KEY_BLOCKSIZE_BITS      "blocksize_bits"
+#define KEY_BLOCKSIZE           "blocksize"
 #define KEY_CAPA_KEY            "capa_key"
+#define KEY_CHANGELOG_CLEAR     "changelog_clear"
+#define KEY_FID2PATH            "fid2path"
+#define KEY_CHECKSUM            "checksum"
+#define KEY_CLEAR_FS            "clear_fs"
 #define KEY_CONN_DATA           "conn_data"
-#define KEY_MAX_EASIZE          "max_easize"
-#define KEY_REVIMP_UPD          "revimp_update"
-#define KEY_LOV_IDX             "lov_idx"
+#define KEY_EVICT_BY_NID        "evict_by_nid"
+#define KEY_FIEMAP              "fiemap"
+#define KEY_FLUSH_CTX           "flush_ctx"
+#define KEY_GRANT_SHRINK        "grant_shrink"
+#define KEY_HSM_COPYTOOL_SEND   "hsm_send"
+#define KEY_INIT_RECOV_BACKUP   "init_recov_bk"
+#define KEY_INIT_RECOV          "initial_recov"
+#define KEY_INTERMDS            "inter_mds"
 #define KEY_LAST_ID             "last_id"
-#define KEY_READONLY            "read-only"
 #define KEY_LOCK_TO_STRIPE      "lock_to_stripe"
-#define KEY_CHECKSUM            "checksum"
-#define KEY_UNLINKED            "unlinked"
-#define KEY_EVICT_BY_NID        "evict_by_nid"
+#define KEY_LOVDESC             "lovdesc"
+#define KEY_LOV_IDX             "lov_idx"
+#define KEY_MAX_EASIZE          "max_easize"
+#define KEY_MDS_CONN            "mds_conn"
+#define KEY_MGSSEC              "mgssec"
+#define KEY_NEXT_ID             "next_id"
+#define KEY_READ_ONLY           "read-only"
 #define KEY_REGISTER_TARGET     "register_target"
+#define KEY_REVIMP_UPD          "revimp_update"
 #define KEY_SET_FS              "set_fs"
-#define KEY_CLEAR_FS            "clear_fs"
-#define KEY_BLOCKSIZE           "blocksize"
-#define KEY_BLOCKSIZE_BITS      "blocksize_bits"
-#define KEY_FIEMAP              "FIEMAP"
+/*      KEY_SET_INFO in lustre_idl.h */
 #define KEY_SPTLRPC_CONF        "sptlrpc_conf"
-#define KEY_MGSSEC              "mgssec"
-/* XXX unused ?*/
-#define KEY_INTERMDS            "inter_mds"
-#define KEY_ASYNC               "async"
+#define KEY_CONNECT_FLAG        "connect_flags"
+#define KEY_SYNC_LOCK_CANCEL    "sync_lock_cancel"
+
 
 struct lu_context;
 
@@ -1139,7 +1193,7 @@ 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;
@@ -1158,6 +1212,7 @@ 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. */
@@ -1186,13 +1241,14 @@ struct md_enqueue_info {
         struct lookup_intent    mi_it;
         struct lustre_handle    mi_lockh;
         struct dentry          *mi_dentry;
+        struct inode           *mi_dir;
         md_enqueue_cb_t         mi_cb;
         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,
@@ -1217,7 +1273,7 @@ struct obd_ops {
          * granted by the target, which are guaranteed to be a subset of flags
          * asked for. If @ocd == NULL, use default parameters. */
         int (*o_connect)(const struct lu_env *env,
-                         struct lustre_handle *conn, struct obd_device *src,
+                         struct obd_export **exp, struct obd_device *src,
                          struct obd_uuid *cluuid, struct obd_connect_data *ocd,
                          void *localdata);
         int (*o_reconnect)(const struct lu_env *env,
@@ -1249,14 +1305,15 @@ struct obd_ops {
                         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_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);
@@ -1288,8 +1345,8 @@ 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 (*)(obd_id, obd_seq, void *),
+                         obd_id *startid, obd_seq seq, 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,
@@ -1306,10 +1363,12 @@ 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 *,
@@ -1317,8 +1376,7 @@ struct obd_ops {
 
         /* llog related obd_methods */
         int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
-                           struct obd_device *disk_obd, int count,
-                           struct llog_catid *logid, struct obd_uuid *uuid);
+                           struct obd_device *disk_obd, int *idx);
         int (*o_llog_finish)(struct obd_device *obd, int count);
         int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
 
@@ -1343,7 +1401,8 @@ 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);
@@ -1355,21 +1414,16 @@ struct obd_ops {
                           char *ostname);
         int (*o_pool_rem)(struct obd_device *obd, char *poolname,
                           char *ostname);
+        void (*o_getref)(struct obd_device *obd);
+        void (*o_putref)(struct obd_device *obd);
+        int (*o_sync_fs)(struct obd_device *obd, struct obd_info *oinfo,
+                         int wait);
         /*
          * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
          * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
          * Also, add a wrapper function in include/linux/obd_class.h. */
 };
 
-/* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
-struct lmv_stripe_md {
-        __u32         mea_magic;
-        __u32         mea_count;
-        __u32         mea_master;
-        __u32         mea_padding;
-        struct lu_fid mea_ids[0];
-};
-
 enum {
         LUSTRE_OPC_MKDIR    = (1 << 0),
         LUSTRE_OPC_SYMLINK  = (1 << 1),
@@ -1401,7 +1455,9 @@ struct lustre_md {
 
 struct md_open_data {
         struct obd_client_handle *mod_och;
-        struct list_head          mod_replay_list;
+        struct ptlrpc_request    *mod_open_req;
+        struct ptlrpc_request    *mod_close_req;
+        cfs_atomic_t              mod_refcount;
 };
 
 struct lookup_intent;
@@ -1411,6 +1467,8 @@ 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 *,
@@ -1422,12 +1480,10 @@ 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 **,
@@ -1475,7 +1531,7 @@ 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 *);
+        int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u32 *);
 
         ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
                                     const struct lu_fid *, ldlm_type_t,
@@ -1483,8 +1539,8 @@ 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 *,
@@ -1498,9 +1554,8 @@ 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 *, __u32 *);
 
         /*
          * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
@@ -1514,13 +1569,9 @@ struct lsm_operations {
         int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
                            struct obd_export *md_exp);
         void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
-                                     unsigned long *);
+                                    obd_off *);
         void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
-                                     unsigned long *);
-        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);
+                                     obd_off *);
         int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
                                int *stripe_count);
         int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
@@ -1528,19 +1579,16 @@ struct lsm_operations {
 };
 
 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:
-               CERROR("Cannot recognize lsm_magic %d\n", magic);
+               CERROR("Cannot recognize lsm_magic %08x\n", magic);
                return NULL;
         }
 }
@@ -1552,22 +1600,24 @@ int lvfs_check_io_health(struct obd_device *obd, struct file *file);
 #define OBD_CALC_STRIPE_END     2
 
 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
-                                         int error)
+                                         struct obd_export *exp, int error)
 {
         if (error) {
                 CERROR("%s: transno "LPU64" commit error: %d\n",
                        obd->obd_name, transno, error);
                 return;
         }
-        if (transno > obd->obd_last_committed) {
-                CDEBUG(D_HA, "%s: transno "LPD64" committed\n",
+        if (exp && transno > exp->exp_last_committed) {
+                CDEBUG(D_HA, "%s: transno "LPU64" committed\n",
                        obd->obd_name, transno);
-                obd->obd_last_committed = transno;
-                ptlrpc_commit_replies (obd);
+                exp->exp_last_committed = transno;
+                ptlrpc_commit_replies(exp);
         } else {
-                CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
+                CDEBUG(D_INFO, "%s: transno "LPU64" committed\n",
                        obd->obd_name, transno);
         }
+        if (transno > obd->obd_last_committed)
+                obd->obd_last_committed = transno;
 }
 
 static inline void init_obd_quota_ops(quota_interface_t *interface,
@@ -1582,14 +1632,32 @@ static inline void init_obd_quota_ops(quota_interface_t *interface,
         obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit);
 }
 
-static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
+static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
 {
-        return obdo_mdsno(oinfo->oi_oa);
+        return oinfo->oi_capa;
 }
 
-static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
+static inline struct md_open_data *obd_mod_alloc(void)
 {
-        return oinfo->oi_capa;
+        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);                              \
+        }                                                       \
+})
+
+extern void obdo_from_inode(struct obdo *dst, struct inode *src,
+                            struct lu_fid *parent, obd_flag valid);
+
 #endif /* __OBD_H */