Whamcloud - gitweb
- changes with names on exports and another fields in llite and mds. lov_exp is
[fs/lustre-release.git] / lustre / lmv / lmv_intent.c
index 14640c4..1badf65 100644 (file)
 #include "lmv_internal.h"
 
 
-int lmv_handle_remote_inode(struct obd_export *exp, struct ll_uctxt *uctxt,
-                            void *lmm, int lmmsize, 
-                            struct lookup_intent *it, int flags,
-                            struct ptlrpc_request **reqp,
+static inline void lmv_drop_intent_lock(struct lookup_intent *it)
+{
+        if (it->d.lustre.it_lock_mode != 0)
+                ldlm_lock_decref((void *)&it->d.lustre.it_lock_handle,
+                                 it->d.lustre.it_lock_mode);
+}
+
+int lmv_handle_remote_inode(struct obd_export *exp, void *lmm,
+                            int lmmsize, struct lookup_intent *it,
+                            int flags, struct ptlrpc_request **reqp,
                             ldlm_blocking_callback cb_blocking)
 {
         struct obd_device *obd = exp->exp_obd;
@@ -65,16 +71,20 @@ int lmv_handle_remote_inode(struct obd_export *exp, struct ll_uctxt *uctxt,
         LASSERT(body != NULL);
 
         if (body->valid & OBD_MD_MDS) {
-                /* oh, MDS reports that this is remote inode case
-                 * i.e. we have to ask for real attrs on another MDS */
-                struct ptlrpc_request *req;
-                struct ll_fid nfid;
+                /*
+                 * oh, MDS reports that this is remote inode case i.e. we have
+                 * to ask for real attrs on another MDS.
+                 */
+                struct ptlrpc_request *req = NULL;
                 struct lustre_handle plock;
+                struct lustre_id nid;
                 int pmode;
 
-                if (it->it_op == IT_LOOKUP) {
-                        /* unfortunately, we have to lie to MDC/MDS to
-                         * retrieve attributes llite needs */
+                if (it->it_op == IT_LOOKUP || it->it_op == IT_CHDIR) {
+                        /*
+                         * unfortunately, we have to lie to MDC/MDS to retrieve
+                         * attributes llite needs.
+                         */
                         it->it_op = IT_GETATTR;
                 }
 
@@ -82,25 +92,25 @@ int lmv_handle_remote_inode(struct obd_export *exp, struct ll_uctxt *uctxt,
                 pmode = it->d.lustre.it_lock_mode;
                 if (pmode) {
                         memcpy(&plock, &it->d.lustre.it_lock_handle,
-                                        sizeof(plock));
+                               sizeof(plock));
                         it->d.lustre.it_lock_mode = 0;
                 }
 
-                nfid = body->fid1;
+                nid = body->id1;
                 it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
-                rc = md_intent_lock(lmv->tgts[nfid.mds].exp, uctxt, &nfid,
+                rc = md_intent_lock(lmv->tgts[id_group(&nid)].ltd_exp, &nid,
                                     NULL, 0, lmm, lmmsize, NULL, it, flags,
                                     &req, cb_blocking);
 
-                /* llite needs LOOKUP lock to track dentry revocation in 
-                 * order to maintain dcache consistency. thus drop UPDATE
-                 * lock here and put LOOKUP in request */
+                /*
+                 * llite needs LOOKUP lock to track dentry revocation in order
+                 * to maintain dcache consistency. Thus drop UPDATE lock here
+                 * and put LOOKUP in request.
+                 */
                 if (rc == 0) {
-                        LASSERT(it->d.lustre.it_lock_mode != 0);
-                        ldlm_lock_decref((void *)&it->d.lustre.it_lock_handle,
-                                         it->d.lustre.it_lock_mode);
+                        lmv_drop_intent_lock(it);
                         memcpy(&it->d.lustre.it_lock_handle, &plock,
-                                        sizeof(plock));
+                               sizeof(plock));
                         it->d.lustre.it_lock_mode = pmode;
                         
                 } else if (pmode)
@@ -112,45 +122,47 @@ int lmv_handle_remote_inode(struct obd_export *exp, struct ll_uctxt *uctxt,
         RETURN(rc);
 }
 
-int lmv_intent_open(struct obd_export *exp, struct ll_uctxt *uctxt,
-                    struct ll_fid *pfid, const char *name, int len,
-                    void *lmm, int lmmsize, struct ll_fid *cfid,
-                    struct lookup_intent *it, int flags,
-                    struct ptlrpc_request **reqp,
+int lmv_intent_open(struct obd_export *exp, struct lustre_id *pid,
+                    const char *name, int len, void *lmm, int lmmsize,
+                    struct lustre_id *cid, struct lookup_intent *it,
+                    int flags, struct ptlrpc_request **reqp,
                     ldlm_blocking_callback cb_blocking)
 {
         struct obd_device *obd = exp->exp_obd;
         struct lmv_obd *lmv = &obd->u.lmv;
         struct mds_body *body = NULL;
-        struct ll_fid rpfid = *pfid;
+        struct lustre_id rpid = *pid;
+        int rc, mds, loop = 0;
         struct lmv_obj *obj;
         struct mea *mea;
-        int rc, mds;
         ENTRY;
 
-        /* IT_OPEN is intended to open (and create, possible) an object.
-         * parent (pfid) may be splitted dir */
+        /* IT_OPEN is intended to open (and create, possible) an object. Parent
+         * (pid) may be splitted dir */
 
 repeat:
-        mds = rpfid.mds;
-        obj = lmv_grab_obj(obd, &rpfid, 0);
+        LASSERT(++loop <= 2);
+        mds = id_group(&rpid);
+        obj = lmv_grab_obj(obd, &rpid);
         if (obj) {
                 /* directory is already splitted, so we have to forward
                  * request to the right MDS */
-                mds = raw_name2idx(obj->objcount, (char *)name, len);
-                rpfid = obj->objs[mds].fid;
-                CDEBUG(D_OTHER, "forward to MDS #%u\n", mds);
+                mds = raw_name2idx(obj->hashtype, obj->objcount, 
+                                   (char *)name, len);
+                
+                CDEBUG(D_OTHER, "forward to MDS #%u ("DLID4")\n",
+                       mds, OLID4(&rpid));
+                rpid = obj->objs[mds].id;
+                lmv_put_obj(obj);
         }
 
-        rc = md_intent_lock(lmv->tgts[mds].exp, uctxt, &rpfid, name,
-                            len, lmm, lmmsize, cfid, it, flags, reqp,
-                            cb_blocking);
-        lmv_put_obj(obj);
+        rc = md_intent_lock(lmv->tgts[id_group(&rpid)].ltd_exp, &rpid, name,
+                            len, lmm, lmmsize, cid, it, flags, reqp, cb_blocking);
         if (rc == -ERESTART) {
-                /* directory got splitted. time to update local object
-                 * and repeat the request with proper MDS */
-                LASSERT(fid_equal(pfid, &rpfid));
-                rc = lmv_get_mea_and_update_object(exp, &rpfid);
+                /* directory got splitted. time to update local object and
+                 * repeat the request with proper MDS */
+                LASSERT(id_equal_fid(pid, &rpid));
+                rc = lmv_get_mea_and_update_object(exp, &rpid);
                 if (rc == 0) {
                         ptlrpc_req_finished(*reqp);
                         goto repeat;
@@ -159,148 +171,165 @@ repeat:
         if (rc != 0)
                 RETURN(rc);
 
-        /* okay, MDS has returned success. probably name has been
-         * resolved in remote inode */
-        rc = lmv_handle_remote_inode(exp, uctxt, lmm, lmmsize, it, flags,
-                                     reqp, cb_blocking);
+        /* okay, MDS has returned success. Probably name has been resolved in
+         * remote inode */
+        rc = lmv_handle_remote_inode(exp, lmm, lmmsize, it,
+                                     flags, reqp, cb_blocking);
         if (rc != 0) {
                 LASSERT(rc < 0);
+
+                /* 
+                 * this is possible, that some userspace application will try to
+                 * open file as directory and we will have error -20 here. As
+                 * this is "usual" situation, we should not print error here,
+                 * only debug info.
+                 */
+                CDEBUG(D_OTHER, "can't handle remote %s: dir "DLID4"("DLID4"):"
+                       "%*s: %d\n", LL_IT2STR(it), OLID4(pid), OLID4(&rpid),
+                       len, name, rc);
                 RETURN(rc);
         }
 
-        /* caller may use attrs MDS returns on IT_OPEN lock request
-         * so, we have to update them for splitted dir */
+        /* caller may use attrs MDS returns on IT_OPEN lock request so, we have
+         * to update them for splitted dir */
         body = lustre_msg_buf((*reqp)->rq_repmsg, 1, sizeof(*body));
         LASSERT(body != NULL);
-        cfid = &body->fid1;
-        obj = lmv_grab_obj(obd, cfid, 0);
-        if (rc == 0 && !obj && (mea = is_body_of_splitted_dir(*reqp, 1))) {
+
+        cid = &body->id1;
+        obj = lmv_grab_obj(obd, cid);
+        if (!obj && (mea = lmv_splitted_dir_body(*reqp, 1))) {
                 /* wow! this is splitted dir, we'd like to handle it */
-                rc = lmv_create_obj_from_attrs(exp, &body->fid1, mea);
+                obj = lmv_create_obj(exp, &body->id1, mea);
+                if (IS_ERR(obj))
+                        RETURN(PTR_ERR(obj));
         }
-        obj = lmv_grab_obj(obd, cfid, 0);
+
         if (obj) {
                 /* this is splitted dir and we'd want to get attrs */
-                CDEBUG(D_OTHER, "attrs from slaves for %lu/%lu/%lu\n",
-                                (unsigned long) cfid->mds,
-                                (unsigned long) cfid->id,
-                                (unsigned long) cfid->generation);
-                rc = lmv_revalidate_slaves(exp, reqp, cfid,
-                                it, 1, cb_blocking);
+                CDEBUG(D_OTHER, "attrs from slaves for "DLID4"\n",
+                       OLID4(cid));
+                
+                rc = lmv_revalidate_slaves(exp, reqp, cid, it, 1,
+                                           cb_blocking);
         } else if (S_ISDIR(body->mode)) {
-                /*CWARN("hmmm, %lu/%lu/%lu has not lmv obj?!\n",
-                                (unsigned long) cfid->mds,
-                                (unsigned long) cfid->id,
-                                (unsigned long) cfid->generation);*/
+                CDEBUG(D_OTHER, "object "DLID4" has not lmv obj?\n",
+                       OLID4(cid));
         }
-        lmv_put_obj(obj);
+        
+        if (obj)
+                lmv_put_obj(obj);
+        
         RETURN(rc);
 }
 
-int lmv_intent_getattr(struct obd_export *exp, struct ll_uctxt *uctxt,
-                       struct ll_fid *pfid, const char *name, int len,
-                       void *lmm, int lmmsize, struct ll_fid *cfid,
-                       struct lookup_intent *it, int flags,
-                       struct ptlrpc_request **reqp,
+int lmv_intent_getattr(struct obd_export *exp, struct lustre_id *pid,
+                       const char *name, int len, void *lmm, int lmmsize,
+                       struct lustre_id *cid, struct lookup_intent *it,
+                       int flags, struct ptlrpc_request **reqp,
                        ldlm_blocking_callback cb_blocking)
 {
         struct obd_device *obd = exp->exp_obd;
         struct lmv_obd *lmv = &obd->u.lmv;
         struct mds_body *body = NULL;
-        struct ll_fid rpfid = *pfid;
-        struct lmv_obj *obj, *obj2;
+        struct lustre_id rpid = *pid;
+        struct lmv_obj *obj = NULL, *obj2 = NULL;
         struct mea *mea;
         int rc = 0, mds;
         ENTRY;
 
-        if (cfid) {
-                /* caller wants to revalidate attrs of obj
-                 * we have to revalidate slaves if requested
-                 * object is splitted directory */
-                CDEBUG(D_OTHER, "revalidate attrs for %lu/%lu/%lu\n",
-                       (unsigned long) cfid->mds,
-                       (unsigned long) cfid->id,
-                       (unsigned long) cfid->generation);
-                mds = cfid->mds;
-                obj = lmv_grab_obj(obd, cfid, 0);
+        if (cid) {
+                /* caller wants to revalidate attrs of obj we have to revalidate
+                 * slaves if requested object is splitted directory */
+                CDEBUG(D_OTHER, "revalidate attrs for "DLID4"\n", OLID4(cid));
+                mds = id_group(cid);
+#if 0
+                obj = lmv_grab_obj(obd, cid);
                 if (obj) {
-                        /* in fact, we need not this with current
-                         * _intent_lock(), but it may change some day */
-                        rpfid = obj->objs[mds].fid;
+                        /* in fact, we need not this with current intent_lock(),
+                         * but it may change some day */
+                        if (!id_equal_fid(pid, cid)){
+                                rpid = obj->objs[mds].id;
+                                mds = id_group(&rpid);
+                        }
+                        lmv_put_obj(obj);
                 }
-                rc = md_intent_lock(lmv->tgts[mds].exp, uctxt, &rpfid, name,
-                                    len, lmm, lmmsize, cfid, it, flags, reqp,
-                                    cb_blocking);
-                if (obj && rc >= 0) {
-                        /* this is splitted dir. in order to optimize things
-                         * a bit, we consider obj valid updating missing
-                         * parts. FIXME: do we need to return any lock here?
-                         * it would be fine if we don't. this means that
-                         * nobody should use UPDATE lock to notify about
-                         * object removal */
-                        CDEBUG(D_OTHER,
-                               "revalidate slaves for %lu/%lu/%lu, rc %d\n",
-                               (unsigned long) cfid->mds,
-                               (unsigned long) cfid->id,
-                               (unsigned long) cfid->generation, rc);
-                        rc = lmv_revalidate_slaves(exp, reqp, cfid, it, rc,
-                                                   cb_blocking);
+#endif
+        } else {
+                CDEBUG(D_OTHER, "INTENT getattr for %*s on "DLID4"\n",
+                       len, name, OLID4(pid));
+                mds = id_group(pid);
+                obj = lmv_grab_obj(obd, pid);
+                if (obj && len) {
+                        /* directory is already splitted. calculate mds */
+                        mds = raw_name2idx(obj->hashtype, obj->objcount, 
+                                           (char *)name, len);
+                        rpid = obj->objs[mds].id;
+                        mds = id_group(&rpid);
+                        lmv_put_obj(obj);
+
+                        CDEBUG(D_OTHER, "forward to MDS #%u (slave "DLID4")\n",
+                               mds, OLID4(&rpid));
                 }
-                RETURN(rc);                
         }
 
-        CDEBUG(D_OTHER, "INTENT getattr for %*s on %lu/%lu/%lu\n",
-               len, name, (unsigned long) pfid->mds,
-               (unsigned long) pfid->id,
-               (unsigned long) pfid->generation);
-
-        mds = pfid->mds;
-        obj = lmv_grab_obj(obd, pfid, 0);
-        if (obj && len) {
-                /* directory is already splitted. calculate mds */
-                mds = raw_name2idx(obj->objcount, (char *) name, len);
-                rpfid = obj->objs[mds].fid;
-                CDEBUG(D_OTHER, "forward to MDS #%u (slave %lu/%lu/%lu)\n",
-                       mds, (unsigned long) rpfid.mds,
-                       (unsigned long) rpfid.id,
-                       (unsigned long) rpfid.generation);
-        }
-        rc = md_intent_lock(lmv->tgts[mds].exp, uctxt, &rpfid, name,
-                            len, lmm, lmmsize, NULL, it, flags, reqp,
-                            cb_blocking);
+        /* the same about fid returning. */
+        rc = md_intent_lock(lmv->tgts[mds].ltd_exp, &rpid, name,
+                            len, lmm, lmmsize, cid, it, flags,
+                            reqp, cb_blocking);
         if (rc < 0)
                 RETURN(rc);
-        LASSERT(rc == 0);
+       
+        if (obj && rc > 0) {
+                /* this is splitted dir. In order to optimize things a
+                 * bit, we consider obj valid updating missing parts.
+
+                 * FIXME: do we need to return any lock here? It would
+                 * be fine if we don't. this means that nobody should
+                 * use UPDATE lock to notify about object * removal */
+                CDEBUG(D_OTHER,
+                       "revalidate slaves for "DLID4", rc %d\n",
+                       OLID4(cid), rc);
+                
+                LASSERT(cid != 0);
+                rc = lmv_revalidate_slaves(exp, reqp, cid, it, rc,
+                                           cb_blocking);
+                RETURN(rc);
+        }
 
+        if (*reqp == NULL)
+                RETURN(rc);
         /* okay, MDS has returned success. probably name has been
          * resolved in remote inode */
-        rc = lmv_handle_remote_inode(exp, uctxt, lmm, lmmsize, it, flags,
-                                     reqp, cb_blocking);
+        rc = lmv_handle_remote_inode(exp, lmm, lmmsize, it,
+                                     flags, reqp, cb_blocking);
         if (rc < 0)
                 RETURN(rc);
 
         body = lustre_msg_buf((*reqp)->rq_repmsg, 1, sizeof(*body));
         LASSERT(body != NULL);
-        cfid = &body->fid1;
-        obj2 = lmv_grab_obj(obd, cfid, 0);
+        
+        cid = &body->id1;
+        obj2 = lmv_grab_obj(obd, cid);
 
-        if (rc == 0 && !obj2 && (mea = is_body_of_splitted_dir(*reqp, 1))) {
-                /* wow! this is splitted dir, we'd like to handle it */
+        if (!obj2 && (mea = lmv_splitted_dir_body(*reqp, 1))) {
+                /* wow! this is splitted dir, we'd like to handle it. */
                 body = lustre_msg_buf((*reqp)->rq_repmsg, 1, sizeof(*body));
                 LASSERT(body != NULL);
-                rc = lmv_create_obj_from_attrs(exp, &body->fid1, mea);
-                obj2 = lmv_grab_obj(obd, cfid, 0);
+
+                obj2 = lmv_create_obj(exp, &body->id1, mea);
+                if (IS_ERR(obj2))
+                        RETURN(PTR_ERR(obj2));
         }
 
         if (obj2) {
                 /* this is splitted dir and we'd want to get attrs */
-                CDEBUG(D_OTHER,
-                                "attrs from slaves for %lu/%lu/%lu, rc %d\n",
-                                (unsigned long) cfid->mds,
-                                (unsigned long) cfid->id,
-                                (unsigned long) cfid->generation, rc);
-                rc = lmv_revalidate_slaves(exp, reqp, cfid,
-                                it, 1, cb_blocking);
+                CDEBUG(D_OTHER, "attrs from slaves for "DLID4", rc %d\n",
+                       OLID4(cid), rc);
+                
+                rc = lmv_revalidate_slaves(exp, reqp, cid, it, 1,
+                                           cb_blocking);
+                lmv_put_obj(obj2);
         }
         RETURN(rc);
 }
@@ -309,11 +338,6 @@ void lmv_update_body_from_obj(struct mds_body *body, struct lmv_inode *obj)
 {
         /* update size */
         body->size += obj->size;
-
-        /* update atime */
-        /* update ctime */
-        /* update mtime */
-        /* update nlink */
 }
 
 int lmv_lookup_slaves(struct obd_export *exp, struct ptlrpc_request **reqp)
@@ -324,7 +348,6 @@ int lmv_lookup_slaves(struct obd_export *exp, struct ptlrpc_request **reqp)
         struct lustre_handle *lockh;
         struct ldlm_lock *lock;
         struct mds_body *body2;
-        struct ll_uctxt uctxt;
         struct lmv_obj *obj;
         int i, rc = 0;
         ENTRY;
@@ -332,88 +355,69 @@ int lmv_lookup_slaves(struct obd_export *exp, struct ptlrpc_request **reqp)
         LASSERT(reqp);
         LASSERT(*reqp);
 
-        /* master is locked. we'd like to take locks on slaves
-         * and update attributes to be returned from the slaves
-         * it's important that lookup is called in two cases:
-         *  - for first time (dcache has no such a resolving yet
-         *  - ->d_revalidate() returned false
-         * last case possible only if all the objs (master and
-         * all slaves aren't valid */
+        /* master is locked. we'd like to take locks on slaves and update
+         * attributes to be returned from the slaves it's important that lookup
+         * is called in two cases:
+         
+         *  - for first time (dcache has no such a resolving yet).
+         *  - ->d_revalidate() returned false.
+         
+         * last case possible only if all the objs (master and all slaves aren't
+         * valid */
 
         body = lustre_msg_buf((*reqp)->rq_repmsg, 1, sizeof(*body));
         LASSERT(body != NULL);
 
-        obj = lmv_grab_obj(obd, &body->fid1, 0);
-        LASSERT(obj);
+        obj = lmv_grab_obj(obd, &body->id1);
+        LASSERT(obj != NULL);
 
-        CDEBUG(D_OTHER, "lookup slaves for %lu/%lu/%lu\n",
-               (unsigned long) body->fid1.mds,
-               (unsigned long) body->fid1.id,
-               (unsigned long) body->fid1.generation);
+        CDEBUG(D_OTHER, "lookup slaves for "DLID4"\n", 
+               OLID4(&body->id1));
 
-        uctxt.gid1 = 0;
-        uctxt.gid2 = 0;
+        lmv_lock_obj(obj);
+        
         for (i = 0; i < obj->objcount; i++) {
-                struct ll_fid fid = obj->objs[i].fid;
+                struct lustre_id id = obj->objs[i].id;
                 struct ptlrpc_request *req = NULL;
                 struct lookup_intent it;
 
-                if (fid_equal(&fid, &obj->fid)) {
+                if (id_equal_fid(&id, &obj->id))
                         /* skip master obj */
                         continue;
-                }
 
-                CDEBUG(D_OTHER, "lookup slave %lu/%lu/%lu\n",
-                       (unsigned long) fid.mds,
-                       (unsigned long) fid.id,
-                       (unsigned long) fid.generation);
+                CDEBUG(D_OTHER, "lookup slave "DLID4"\n", OLID4(&id));
 
                 /* is obj valid? */
                 memset(&it, 0, sizeof(it));
                 it.it_op = IT_GETATTR;
-                rc = md_intent_lock(lmv->tgts[fid.mds].exp, &uctxt, &fid,
-                                    NULL, 0, NULL, 0, &fid, &it, 0, &req,
+                rc = md_intent_lock(lmv->tgts[id_group(&id)].ltd_exp, &id,
+                                    NULL, 0, NULL, 0, &id, &it, 0, &req,
                                     lmv_dirobj_blocking_ast);
-                lockh = (struct lustre_handle *) &it.d.lustre.it_lock_handle;
-                if (rc > 0) {
+                
+                lockh = (struct lustre_handle *)&it.d.lustre.it_lock_handle;
+                if (rc > 0 && req == NULL) {
                         /* nice, this slave is valid */
                         LASSERT(req == NULL);
                         CDEBUG(D_OTHER, "cached\n");
                         goto release_lock;
                 }
 
-                if (rc < 0) {
-                        /* error during revalidation */
-                        GOTO(cleanup, rc);
-                }
-
-                /* rc == 0, this means we have no such a lock and can't
-                 * think obj is still valid. lookup it again */
-                LASSERT(req == NULL);
-                req = NULL;
-                memset(&it, 0, sizeof(it));
-                it.it_op = IT_GETATTR;
-                rc = md_intent_lock(lmv->tgts[fid.mds].exp, &uctxt, &fid,
-                                    NULL, 0, NULL, 0, NULL, &it, 0, &req,
-                                    lmv_dirobj_blocking_ast);
-                lockh = (struct lustre_handle *) &it.d.lustre.it_lock_handle;
-                LASSERT(rc <= 0);
-                if (rc < 0) {
+                if (rc < 0)
                         /* error during lookup */
                         GOTO(cleanup, rc);
-                }
                 
                 lock = ldlm_handle2lock(lockh);
                 LASSERT(lock);
-                lock->l_ast_data = obj;
-                atomic_inc(&obj->count);
+
+                lock->l_ast_data = lmv_get_obj(obj);
 
                 body2 = lustre_msg_buf(req->rq_repmsg, 1, sizeof(*body2));
                 LASSERT(body2);
 
                 obj->objs[i].size = body2->size;
+                
                 CDEBUG(D_OTHER, "fresh: %lu\n",
-                                (unsigned long) obj->objs[i].size);
+                       (unsigned long)obj->objs[i].size);
 
                 LDLM_LOCK_PUT(lock);
 
@@ -421,157 +425,172 @@ int lmv_lookup_slaves(struct obd_export *exp, struct ptlrpc_request **reqp)
                         ptlrpc_req_finished(req);
 release_lock:
                 lmv_update_body_from_obj(body, obj->objs + i);
+
                 if (it.d.lustre.it_lock_mode)
                         ldlm_lock_decref(lockh, it.d.lustre.it_lock_mode);
         }
+
+        EXIT;
 cleanup:
-        RETURN(rc);
+        lmv_unlock_obj(obj);
+        lmv_put_obj(obj);
+        return rc;
 }
 
-int lmv_intent_lookup(struct obd_export *exp, struct ll_uctxt *uctxt,
-                struct ll_fid *pfid, const char *name, int len,
-                void *lmm, int lmmsize, struct ll_fid *cfid,
-                struct lookup_intent *it, int flags,
-                struct ptlrpc_request **reqp,
-                ldlm_blocking_callback cb_blocking)
+int lmv_intent_lookup(struct obd_export *exp, struct lustre_id *pid,
+                      const char *name, int len, void *lmm, int lmmsize,
+                      struct lustre_id *cid, struct lookup_intent *it,
+                      int flags, struct ptlrpc_request **reqp,
+                      ldlm_blocking_callback cb_blocking)
 {
         struct obd_device *obd = exp->exp_obd;
         struct lmv_obd *lmv = &obd->u.lmv;
         struct mds_body *body = NULL;
-        struct ll_fid rpfid = *pfid;
+        struct lustre_id rpid = *pid;
         struct lmv_obj *obj;
         struct mea *mea;
-        int rc, mds;
+        int rc, mds, loop = 0;
         ENTRY;
 
-        /* IT_LOOKUP is intended to produce name -> fid resolving
-         * (let's call this lookup below) or to confirm requested
-         * resolving is still valid (let's call this revalidation)
-         * cfid != NULL specifies revalidation */
-
-        if (cfid) {
-                /* this is revalidation: we have to check is LOOKUP
-                 * lock still valid for given fid. very important
-                 * part is that we have to choose right mds because
-                 * namespace is per mds */
-                rpfid = *pfid;
-                obj = lmv_grab_obj(obd, pfid, 0);
+        /*
+         * IT_LOOKUP is intended to produce name -> id resolving (let's call
+         * this lookup below) or to confirm requested resolving is still valid
+         * (let's call this revalidation) cid != NULL specifies revalidation.
+         */
+        if (cid) {
+                /*
+                 * this is revalidation: we have to check is LOOKUP lock still
+                 * valid for given id. Very important part is that we have to
+                 * choose right mds because namespace is per mds.
+                 */
+                rpid = *pid;
+                obj = lmv_grab_obj(obd, pid);
                 if (obj) {
-                        mds = raw_name2idx(obj->objcount, (char *) name, len);
-                        rpfid = obj->objs[mds].fid;
+                        mds = raw_name2idx(obj->hashtype, obj->objcount,
+                                           (char *)name, len);
+                        rpid = obj->objs[mds].id;
                         lmv_put_obj(obj);
                 }
-                mds = rpfid.mds;
-                CDEBUG(D_OTHER, "revalidate lookup for %lu/%lu/%lu to %d MDS\n",
-                       (unsigned long) cfid->mds,
-                       (unsigned long) cfid->id,
-                       (unsigned long) cfid->generation, mds);
-                rc = md_intent_lock(lmv->tgts[mds].exp, uctxt, pfid, name,
-                                    len, lmm, lmmsize, cfid, it, flags,
-                                    reqp, cb_blocking);
-                RETURN(rc);
-        }
+                mds = id_group(&rpid);
 
-        mds = pfid->mds;
+                CDEBUG(D_OTHER, "revalidate lookup for "DLID4" to %d MDS\n",
+                       OLID4(cid), mds);
+
+        } else {
+                mds = id_group(pid);
 repeat:
-        /* this is lookup. during lookup we have to update all the
-         * attributes, because returned values will be put in struct
-         * inode */
-
-        obj = lmv_grab_obj(obd, pfid, 0);
-        if (obj && len) {
-                /* directory is already splitted. calculate mds */
-                mds = raw_name2idx(obj->objcount, (char *) name, len);
-                rpfid = obj->objs[mds].fid;
-                lmv_put_obj(obj);
-        }
+                LASSERT(++loop <= 2);
+                
+                /* this is lookup. during lookup we have to update all the
+                 * attributes, because returned values will be put in struct
+                 * inode */
 
-        rc = md_intent_lock(lmv->tgts[mds].exp, uctxt, &rpfid, name,
-                            len, lmm, lmmsize, NULL, it, flags, reqp,
-                            cb_blocking);
+                obj = lmv_grab_obj(obd, pid);
+                if (obj) {
+                        if (len) {
+                                /* directory is already splitted. calculate mds */
+                                mds = raw_name2idx(obj->hashtype, obj->objcount, 
+                                                   (char *)name, len);
+                                rpid = obj->objs[mds].id;
+                                mds = id_group(&rpid);
+                        }
+                        lmv_put_obj(obj);
+                }
+        }
+        rc = md_intent_lock(lmv->tgts[mds].ltd_exp, &rpid, name,
+                            len, lmm, lmmsize, cid, it, flags,
+                            reqp, cb_blocking);
+        if (rc > 0) {
+                LASSERT(cid != 0);
+                RETURN(rc);
+        }
         if (rc > 0) {
-                /* very interesting. it seems object is still valid
-                 * but for some reason llite calls lookup, not revalidate */
-                CWARN("lookup for %lu/%lu/%lu and data should be uptodate\n",
-                       (unsigned long) rpfid.mds,
-                       (unsigned long) rpfid.id,
-                       (unsigned long) rpfid.generation);
+                /* very interesting. it seems object is still valid but for some
+                 * reason llite calls lookup, not revalidate */
+                CDEBUG(D_OTHER, "lookup for "DLID4" and data should be uptodate\n",
+                      OLID4(&rpid));
                 LASSERT(*reqp == NULL);
                 RETURN(rc);
         }
 
         if (rc == 0 && *reqp == NULL) {
                 /* once again, we're asked for lookup, not revalidate */
-                CWARN("lookup for %lu/%lu/%lu and data should be uptodate\n",
-                       (unsigned long) rpfid.mds,
-                       (unsigned long) rpfid.id,
-                       (unsigned long) rpfid.generation);
+                CDEBUG(D_OTHER, "lookup for "DLID4" and data should be uptodate\n",
+                      OLID4(&rpid));
                 RETURN(rc);
         }
        
         if (rc == -ERESTART) {
-                /* directory got splitted since last update. this shouldn't
-                 * be becasue splitting causes lock revocation, so revalidate
-                 * had to fail and lookup on dir had to return mea */
+                /* directory got splitted since last update. this shouldn't be
+                 * becasue splitting causes lock revocation, so revalidate had
+                 * to fail and lookup on dir had to return mea */
                 CWARN("we haven't knew about directory splitting!\n");
                 LASSERT(obj == NULL);
-                rc = lmv_create_obj_from_attrs(exp, &rpfid, NULL);
-                if (rc)
-                        RETURN(rc);
+
+                obj = lmv_create_obj(exp, &rpid, NULL);
+                if (IS_ERR(obj))
+                        RETURN(PTR_ERR(obj));
+                lmv_put_obj(obj);
                 goto repeat;
         }
 
         if (rc < 0)
                 RETURN(rc);
 
-        /* okay, MDS has returned success. probably name has been
-         * resolved in remote inode */
-        rc = lmv_handle_remote_inode(exp, uctxt, lmm, lmmsize, it, flags,
-                                     reqp, cb_blocking);
+        /* okay, MDS has returned success. Probably name has been resolved in
+         * remote inode. */
+        rc = lmv_handle_remote_inode(exp, lmm, lmmsize, it,
+                                     flags, reqp, cb_blocking);
 
-        if (rc == 0 && (mea = is_body_of_splitted_dir(*reqp, 1))) {
+        if (rc == 0 && (mea = lmv_splitted_dir_body(*reqp, 1))) {
                 /* wow! this is splitted dir, we'd like to handle it */
                 body = lustre_msg_buf((*reqp)->rq_repmsg, 1, sizeof(*body));
                 LASSERT(body != NULL);
-                obj = lmv_grab_obj(obd, &body->fid1, 0);
-                if (!obj)
-                        rc = lmv_create_obj_from_attrs(exp, &body->fid1, mea);
+                
+                obj = lmv_grab_obj(obd, &body->id1);
+                if (!obj) {
+                        obj = lmv_create_obj(exp, &body->id1, mea);
+                        if (IS_ERR(obj))
+                                RETURN(PTR_ERR(obj));
+                }
                 lmv_put_obj(obj);
         }
 
         RETURN(rc);
 }
 
-int lmv_intent_lock(struct obd_export *exp, struct ll_uctxt *uctxt,
-                struct ll_fid *pfid, const char *name, int len,
-                void *lmm, int lmmsize, struct ll_fid *cfid,
-                struct lookup_intent *it, int flags,
-                struct ptlrpc_request **reqp,
-                ldlm_blocking_callback cb_blocking)
+int lmv_intent_lock(struct obd_export *exp, struct lustre_id *pid,
+                    const char *name, int len, void *lmm, int lmmsize,
+                    struct lustre_id *cid, struct lookup_intent *it,
+                    int flags, struct ptlrpc_request **reqp,
+                    ldlm_blocking_callback cb_blocking)
 {
         struct obd_device *obd = exp->exp_obd;
         int rc = 0;
         ENTRY;
 
         LASSERT(it);
-        LASSERT(pfid);
+        LASSERT(pid);
+
+        CDEBUG(D_OTHER, "INTENT LOCK '%s' for '%*s' on %lu/%lu -> %lu\n",
+               LL_IT2STR(it), len, name, (unsigned long)id_ino(pid),
+               (unsigned long)id_gen(pid), (unsigned long)id_group(pid));
 
-        CDEBUG(D_OTHER, "INTENT LOCK '%s' for '%*s' on %lu/%lu -> %u\n",
-                        LL_IT2STR(it), len, name, (unsigned long) pfid->id,
-                        (unsigned long) pfid->generation, pfid->mds);
+        rc = lmv_check_connect(obd);
+        if (rc)
+                RETURN(rc);
 
-        lmv_connect(obd);
         if (it->it_op == IT_LOOKUP)
-                rc = lmv_intent_lookup(exp, uctxt, pfid, name, len, lmm,
-                                       lmmsize, cfid, it, flags, reqp,
+                rc = lmv_intent_lookup(exp, pid, name, len, lmm,
+                                       lmmsize, cid, it, flags, reqp,
                                        cb_blocking);
         else if (it->it_op & IT_OPEN)
-                rc = lmv_intent_open(exp, uctxt, pfid, name, len, lmm,
-                                     lmmsize, cfid, it, flags, reqp,
+                rc = lmv_intent_open(exp, pid, name, len, lmm,
+                                     lmmsize, cid, it, flags, reqp,
                                      cb_blocking);
         else if (it->it_op == IT_GETATTR || it->it_op == IT_CHDIR)
-                rc = lmv_intent_getattr(exp, uctxt, pfid, name, len, lmm,
-                                        lmmsize, cfid, it, flags, reqp,
+                rc = lmv_intent_getattr(exp, pid, name, len, lmm,
+                                        lmmsize, cid, it, flags, reqp,
                                         cb_blocking);
         else
                 LBUG();
@@ -579,59 +598,57 @@ int lmv_intent_lock(struct obd_export *exp, struct ll_uctxt *uctxt,
 }
 
 int lmv_revalidate_slaves(struct obd_export *exp, struct ptlrpc_request **reqp,
-                          struct ll_fid *mfid, struct lookup_intent *oit,
+                          struct lustre_id *mid, struct lookup_intent *oit,
                           int master_valid, ldlm_blocking_callback cb_blocking)
 {
         struct obd_device *obd = exp->exp_obd;
         struct ptlrpc_request *mreq = *reqp;
         struct lmv_obd *lmv = &obd->u.lmv;
         struct lustre_handle master_lockh;
-        unsigned long size = 0;
         struct ldlm_lock *lock;
+        unsigned long size = 0;
         struct mds_body *body;
-        struct ll_uctxt uctxt;
         struct lmv_obj *obj;
         int master_lock_mode;
         int i, rc = 0;
         ENTRY;
 
-        /* we have to loop over the subobjects, check validity and update
-         * them from MDSs if needed. it's very useful that we need not to
-         * update all the fields. say, common fields (that are equal on 
-         * all the subojects need not to be update, another fields (i_size,
-         * for example) are cached all the time */
-        obj = lmv_grab_obj(obd, mfid, 0);
-        LASSERT(obj);
+        /* we have to loop over the subobjects, check validity and update them
+         * from MDSs if needed. it's very useful that we need not to update all
+         * the fields. say, common fields (that are equal on all the subojects
+         * need not to be update, another fields (i_size, for example) are
+         * cached all the time */
+        obj = lmv_grab_obj(obd, mid);
+        LASSERT(obj != NULL);
 
         master_lock_mode = 0;
-        uctxt.gid1 = 0;
-        uctxt.gid2 = 0;
+
+        lmv_lock_obj(obj);
+        
         for (i = 0; i < obj->objcount; i++) {
-                struct ll_fid fid = obj->objs[i].fid;
+                struct lustre_id id = obj->objs[i].id;
                 struct lustre_handle *lockh = NULL;
                 struct ptlrpc_request *req = NULL;
                 ldlm_blocking_callback cb;
                 struct lookup_intent it;
                 int master = 0;
 
-                CDEBUG(D_OTHER, "revalidate subobj %lu/%lu/%lu\n",
-                       (unsigned long) fid.mds,
-                       (unsigned long) fid.id,
-                       (unsigned long) fid.generation);
+                CDEBUG(D_OTHER, "revalidate subobj "DLID4"\n",
+                       OLID4(&id));
 
                 memset(&it, 0, sizeof(it));
                 it.it_op = IT_GETATTR;
                 cb = lmv_dirobj_blocking_ast;
 
-                if (fid_equal(&fid, &obj->fid)) {
+                if (id_equal_fid(&id, &obj->id)) {
                         if (master_valid) {
                                 /* lmv_intent_getattr() already checked
                                  * validness and took the lock */
                                 if (mreq) {
-                                        /* it even got the reply
-                                         * refresh attrs from that reply */
+                                        /* it even got the reply refresh attrs
+                                         * from that reply */
                                         body = lustre_msg_buf(mreq->rq_repmsg,
-                                                              1,sizeof(*body));
+                                                              1, sizeof(*body));
                                         LASSERT(body != NULL);
                                         goto update; 
                                 }
@@ -645,36 +662,21 @@ int lmv_revalidate_slaves(struct obd_export *exp, struct ptlrpc_request **reqp,
                 }
 
                 /* is obj valid? */
-                rc = md_intent_lock(lmv->tgts[fid.mds].exp, &uctxt, &fid,
-                                    NULL, 0, NULL, 0, &fid, &it, 0, &req, cb);
+                rc = md_intent_lock(lmv->tgts[id_group(&id)].ltd_exp,
+                                    &id, NULL, 0, NULL, 0, &id, &it, 0, 
+                                    &req, cb);
                 lockh = (struct lustre_handle *) &it.d.lustre.it_lock_handle;
-                if (rc > 0) {
+                if (rc > 0 && req == NULL) {
                         /* nice, this slave is valid */
                         LASSERT(req == NULL);
                         CDEBUG(D_OTHER, "cached\n");
                         goto release_lock;
                 }
 
-                if (rc < 0) {
+                if (rc < 0)
                         /* error during revalidation */
                         GOTO(cleanup, rc);
-                }
 
-                /* rc == 0, this means we have no such a lock and can't
-                 * think obj is still valid. lookup it again */
-                LASSERT(req == NULL);
-                req = NULL;
-                memset(&it, 0, sizeof(it));
-                it.it_op = IT_GETATTR;
-                rc = md_intent_lock(lmv->tgts[fid.mds].exp, &uctxt, &fid,
-                                    NULL, 0, NULL, 0, NULL, &it, 0, &req, cb);
-                lockh = (struct lustre_handle *) &it.d.lustre.it_lock_handle;
-                LASSERT(rc <= 0);
-                if (rc < 0) {
-                        /* error during lookup */
-                        GOTO(cleanup, rc);
-                }
-              
                 if (master) {
                         LASSERT(master_valid == 0);
                         /* save lock on master to be returned to the caller */
@@ -686,14 +688,13 @@ int lmv_revalidate_slaves(struct obd_export *exp, struct ptlrpc_request **reqp,
                         /* this is slave. we want to control it */
                         lock = ldlm_handle2lock(lockh);
                         LASSERT(lock);
-                        lock->l_ast_data = obj;
-                        atomic_inc(&obj->count);
+                        lock->l_ast_data = lmv_get_obj(obj);
                         LDLM_LOCK_PUT(lock);
                 }
 
                 if (*reqp == NULL) {
-                        /* this is first reply, we'll use it to return
-                         * updated data back to the caller */
+                        /* this is first reply, we'll use it to return updated
+                         * data back to the caller */
                         LASSERT(req);
                         ptlrpc_request_addref(req);
                         *reqp = req;
@@ -705,33 +706,38 @@ int lmv_revalidate_slaves(struct obd_export *exp, struct ptlrpc_request **reqp,
                 
 update:
                 obj->objs[i].size = body->size;
+                
                 CDEBUG(D_OTHER, "fresh: %lu\n",
-                       (unsigned long) obj->objs[i].size);
+                       (unsigned long)obj->objs[i].size);
 
                 if (req)
                         ptlrpc_req_finished(req);
 release_lock:
                 size += obj->objs[i].size;
+
                 if (it.d.lustre.it_lock_mode)
                         ldlm_lock_decref(lockh, it.d.lustre.it_lock_mode);
         }
 
         if (*reqp) {
-                /* some attrs got refreshed, we have reply and it's time
-                 * to put fresh attrs to it */
+                /* some attrs got refreshed, we have reply and it's time to put
+                 * fresh attrs to it */
                 CDEBUG(D_OTHER, "return refreshed attrs: size = %lu\n",
-                       (unsigned long) size);
+                       (unsigned long)size);
+                
                 body = lustre_msg_buf((*reqp)->rq_repmsg, 1, sizeof(*body));
                 LASSERT(body);
-                /* FIXME: what about another attributes? */
+
+                /* FIXME: what about other attributes? */
                 body->size = size;
+                
                 if (mreq == NULL) {
-                        /* very important to maintain lli->mds the same
-                         * because of revalidation. mreq == NULL means
-                         * that caller has no reply and the only attr
-                         * we can return is size */
+                        /* very important to maintain id_group(lli->lli_id) the
+                         * same because of revalidation. mreq == NULL means that
+                         * caller has no reply and the only attr we can return
+                         * is size */
                         body->valid = OBD_MD_FLSIZE;
-                        body->mds = obj->fid.mds;
+//                        body->mds = id_group(&obj->id);
                 }
                 if (master_valid == 0) {
                         memcpy(&oit->d.lustre.it_lock_handle,
@@ -746,7 +752,10 @@ release_lock:
                         oit->d.lustre.it_lock_mode = master_lock_mode;
                 rc = 1;
         }
+
+        EXIT;
 cleanup:
-        RETURN(rc);
+        lmv_unlock_obj(obj);
+        lmv_put_obj(obj);
+        return rc;
 }
-