Whamcloud - gitweb
land v0.9.1 on HEAD, in preparation for a 1.0.x branch
[fs/lustre-release.git] / lustre / llite / namei.c
index 17afcef..e768933 100644 (file)
@@ -51,13 +51,29 @@ static int ll_test_inode(struct inode *inode, unsigned long ino, void *opaque)
 static int ll_test_inode(struct inode *inode, void *opaque)
 #endif
 {
+        static int last_ino, last_gen, last_count;
         struct lustre_md *md = opaque;
 
-        if (!(md->body->valid & (OBD_MD_FLGENER | OBD_MD_FLID)))
-                CERROR("invalid generation\n");
-        CDEBUG(D_VFSTRACE, "comparing inode %p ino %lu/%u to body %u/%u\n",
-               inode, inode->i_ino, inode->i_generation, 
-               md->body->ino, md->body->generation);
+        if (!(md->body->valid & (OBD_MD_FLGENER | OBD_MD_FLID))) {
+                CERROR("MDS body missing inum or generation\n");
+                return 0;
+        }
+
+        if (last_ino == md->body->ino && last_gen == md->body->generation &&
+            last_count < 500) {
+                last_count++;
+        } else {
+                if (last_count > 1)
+                        CDEBUG(D_VFSTRACE, "compared %u/%u %u times\n",
+                               last_ino, last_gen, last_count);
+                last_count = 0;
+                last_ino = md->body->ino;
+                last_gen = md->body->generation;
+                CDEBUG(D_VFSTRACE,
+                       "comparing inode %p ino %lu/%u to body %u/%u\n",
+                       inode, inode->i_ino, inode->i_generation,
+                       md->body->ino, md->body->generation);
+        }
 
         if (inode->i_generation != md->body->generation)
                 return 0;
@@ -95,13 +111,13 @@ struct inode *ll_iget(struct super_block *sb, ino_t hash,
         LASSERT(hash != 0);
         inode = iget5_locked(sb, hash, ll_test_inode, ll_set_inode, md);
 
-        if (!inode)
-                return (NULL);              /* removed ERR_PTR(-ENOMEM) -eeb */
-
-        if (inode->i_state & I_NEW)
-                unlock_new_inode(inode);
+        if (inode) {
+                if (inode->i_state & I_NEW)
+                        unlock_new_inode(inode);
+                CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino,
+                       inode->i_generation, inode);
+        }
 
-        // XXX Coda always fills inodes, should Lustre?
         return inode;
 }
 #else
@@ -111,64 +127,18 @@ struct inode *ll_iget(struct super_block *sb, ino_t hash,
         struct inode *inode;
         LASSERT(hash != 0);
         inode = iget4(sb, hash, ll_test_inode, md);
-        CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino,
-               inode->i_generation, inode);
+        if (inode)
+                CDEBUG(D_VFSTRACE, "inode: %lu/%u(%p)\n", inode->i_ino,
+                       inode->i_generation, inode);
         return inode;
 }
 #endif
 
-static int ll_intent_to_lock_mode(struct lookup_intent *it)
-{
-        /* CREAT needs to be tested before open (both could be set) */
-        if (it->it_op & IT_CREAT)
-                return LCK_PW;
-        else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
-                return LCK_PR;
-
-        LBUG();
-        RETURN(-EINVAL);
-}
-
-int ll_it_open_error(int phase, struct lookup_intent *it)
-{
-        if (it_disposition(it, DISP_OPEN_OPEN)) {
-                if (phase == DISP_OPEN_OPEN)
-                        return it->it_status;
-                else
-                        return 0;
-        }
-
-        if (it_disposition(it, DISP_OPEN_CREATE)) {
-                if (phase == DISP_OPEN_CREATE)
-                        return it->it_status;
-                else
-                        return 0;
-        }
-
-        if (it_disposition(it, DISP_LOOKUP_EXECD)) {
-                if (phase == DISP_LOOKUP_EXECD)
-                        return it->it_status;
-                else
-                        return 0;
-        }
-
-        if (it_disposition(it, DISP_IT_EXECD)) {
-                if (phase == DISP_IT_EXECD)
-                        return it->it_status;
-                else
-                        return 0;
-        }
-        CERROR("it disp: %X, status: %d\n", it->it_disposition, it->it_status);
-        LBUG();
-        return 0;
-}
-
 int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
                         void *data, int flag)
 {
         int rc;
         struct lustre_handle lockh;
-        struct inode *inode = lock->l_data;
         ENTRY;
 
         switch (flag) {
@@ -181,9 +151,15 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
                 }
                 break;
         case LDLM_CB_CANCELING: {
+                struct inode *inode = ll_inode_from_lock(lock);
+
                 /* Invalidate all dentries associated with this inode */
                 if (inode == NULL)
                         break;
+
+                clear_bit(LLI_F_HAVE_MDS_SIZE_LOCK,
+                          &(ll_i2info(inode)->lli_flags));
+
                 if (lock->l_resource->lr_name.name[0] != inode->i_ino ||
                     lock->l_resource->lr_name.name[1] != inode->i_generation) {
                         LDLM_ERROR(lock, "data mismatch with ino %lu/%u",
@@ -196,10 +172,10 @@ int ll_mdc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
                         ll_invalidate_inode_pages(inode);
                 }
 
-#warning FIXME: we should probably free this inode if there are no aliases
                 if (inode->i_sb->s_root &&
                     inode != inode->i_sb->s_root->d_inode)
                         ll_unhash_aliases(inode);
+                iput(inode);
                 break;
         }
         default:
@@ -220,228 +196,23 @@ int ll_mdc_cancel_unused(struct lustre_handle *conn, struct inode *inode,
                                       opaque));
 }
 
-void ll_prepare_mdc_op_data(struct mdc_op_data *data,
-                            struct inode *i1,
-                            struct inode *i2,
-                            const char *name,
-                            int namelen,
+
+void ll_prepare_mdc_op_data(struct mdc_op_data *data, struct inode *i1,
+                            struct inode *i2, const char *name, int namelen,
                             int mode)
 {
         LASSERT(i1);
 
-        data->ino1 = i1->i_ino;
-        data->gen1 = i1->i_generation;
-        data->typ1 = i1->i_mode & S_IFMT;
-        data->gid1 = i1->i_gid;
+        ll_i2uctxt(&data->ctxt, i1, i2);
+        ll_inode2fid(&data->fid1, i1);
 
-        if (i2) {
-                data->ino2 = i2->i_ino;
-                data->gen2 = i2->i_generation;
-                data->typ2 = i2->i_mode & S_IFMT;
-                data->gid2 = i2->i_gid;
-        } else {
-                data->ino2 = 0;
-        }
+        if (i2)
+                ll_inode2fid(&data->fid2, i2);
 
         data->name = name;
         data->namelen = namelen;
-        data->mode = mode;
-}
-
-/* 
- *This long block is all about fixing up the local state so that it is
- *correct as of the moment _before_ the operation was applied; that
- *way, the VFS will think that everything is normal and call Lustre's
- *regular VFS methods.
- *
- * If we're performing a creation, that means that unless the creation
- * failed with EEXIST, we should fake up a negative dentry.
- *
- * For everything else, we want to lookup to succeed.
- *
- * One additional note: if CREATE or OPEN succeeded, we add an extra
- * reference to the request because we need to keep it around until
- * ll_create/ll_open gets called.
- *
- * The server will return to us, in it_disposition, an indication of
- * exactly what it_status refers to.
- *
- * If DISP_OPEN_OPEN is set, then it_status refers to the open() call,
- * otherwise if DISP_OPEN_CREATE is set, then it status is the
- * creation failure mode.  In either case, one of DISP_LOOKUP_NEG or
- * DISP_LOOKUP_POS will be set, indicating whether the child lookup
- * was successful.
- *
- * Else, if DISP_LOOKUP_EXECD then it_status is the rc of the child
- * lookup.
- */
-int ll_intent_lock(struct inode *parent, struct dentry **de,
-                   struct lookup_intent *it, int flags, intent_finish_cb intent_finish)
-{
-        struct dentry *dentry = *de;
-        struct inode *inode = dentry->d_inode;
-        struct ll_sb_info *sbi = ll_i2sbi(parent);
-        struct lustre_handle lockh;
-        struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
-        struct ptlrpc_request *request;
-        int rc = 0;
-        struct mds_body *mds_body;
-        int mode;
-        obd_id ino = 0;
-        ENTRY;
-
-#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
-        if (it && it->it_magic != INTENT_MAGIC) { 
-                CERROR("WARNING: uninitialized intent\n");
-                LBUG();
-                intent_init(it, IT_LOOKUP, 0);
-        }
-        if (it->it_op == IT_GETATTR || 
-            it->it_op == 0)
-                it->it_op = IT_LOOKUP;
-        
-#endif
-        if (!it ||it->it_op == IT_GETXATTR)
-                it = &lookup_it;
-
-        it->it_op_release = ll_intent_release;
-
-        CDEBUG(D_DLMTRACE, "name: %*s, intent: %s\n", dentry->d_name.len,
-               dentry->d_name.name, ldlm_it2str(it->it_op));
-        
-        if (dentry->d_name.len > EXT2_NAME_LEN)
-                RETURN(-ENAMETOOLONG);
-
-        /* This function may be called twice, we only once want to
-           execute the request associated with the intent. If it was
-           done already, we skip past this and use the results. */ 
-        if (!it_disposition(it, DISP_ENQ_COMPLETE)) {
-                struct mdc_op_data op_data;
-
-                ll_prepare_mdc_op_data(&op_data, parent, dentry->d_inode,
-                                       dentry->d_name.name, dentry->d_name.len,
-                                       0);
-
-                rc = mdc_enqueue(&sbi->ll_mdc_conn, LDLM_PLAIN, it,
-                                 ll_intent_to_lock_mode(it), &op_data,
-                                 &lockh, NULL, 0, ldlm_completion_ast,
-                                 ll_mdc_blocking_ast, NULL);
-                if (rc < 0)
-                        RETURN(rc);
-                memcpy(it->it_lock_handle, &lockh, sizeof(lockh));
-        }
-        request = it->it_data;
-        LASSERT(request != NULL);
-
-        if (!it_disposition(it, DISP_IT_EXECD)) {
-                /* The server failed before it even started executing the
-                 * intent, i.e. because it couldn't unpack the request. */
-                LASSERT(it->it_status != 0);
-                GOTO(drop_req, rc = it->it_status);
-        }
-        rc = ll_it_open_error(DISP_IT_EXECD, it);
-        if (rc)
-                GOTO(drop_req, rc);
-
-        mds_body = lustre_msg_buf(request->rq_repmsg, 1, sizeof(*mds_body));
-        LASSERT(mds_body != NULL);           /* mdc_enqueue checked */
-        LASSERT_REPSWABBED(request, 1); /* mdc_enqueue swabbed */
-
-        /* XXX everything with fids please, no ino's inode's etc */
-        ino = mds_body->fid1.id;
-        mode = mds_body->mode;
-
-        /*We were called from revalidate2: did we find the same inode?*/
-        if (inode && 
-            (ino != inode->i_ino ||
-             mds_body->fid1.generation != inode->i_generation)) {
-                it_set_disposition(it, DISP_ENQ_COMPLETE);
-                RETURN(-ESTALE);
-        }
-
-        /* If we're doing an IT_OPEN which did not result in an actual
-         * successful open, then we need to remove the bit which saves
-         * this request for unconditional replay. */
-        if (it->it_op & IT_OPEN) {
-                if (!it_disposition(it, DISP_OPEN_OPEN) ||
-                    it->it_status != 0) {
-                        unsigned long flags;
-                
-                        spin_lock_irqsave (&request->rq_lock, flags);
-                        request->rq_replay = 0;
-                        spin_unlock_irqrestore (&request->rq_lock, flags);
-                }
-        }
-
-        rc = ll_it_open_error(DISP_LOOKUP_EXECD, it);
-        if (rc)
-                GOTO(drop_req, rc);
-        
-        /* keep requests around for the multiple phases of the call
-         * this shows the DISP_XX must guarantee we make it into the call 
-         */ 
-        if (it_disposition(it, DISP_OPEN_CREATE))
-                ptlrpc_request_addref(request);
-        if (it_disposition(it, DISP_OPEN_OPEN))
-                ptlrpc_request_addref(request);
-        
-        if (it->it_op & IT_CREAT) {
-                /* XXX this belongs in ll_create_iit */
-        } else if (it->it_op == IT_OPEN) {
-                LASSERT(!it_disposition(it, DISP_OPEN_CREATE));
-        } else 
-                LASSERT(it->it_op & (IT_GETATTR | IT_LOOKUP));
-
-        if (intent_finish != NULL) {
-                struct lustre_handle old_lock;
-                struct ldlm_lock *lock;
-
-                rc = intent_finish(request, parent, de, it, 1, ino);
-                dentry = *de; /* intent_finish may change *de */
-                inode = dentry->d_inode;
-                if (rc != 0)
-                        GOTO(drop_lock, rc);
-
-                /* The intent processing may well have given us a lock different
-                 * from the one we requested.  If we already have a matching
-                 * lock, then cancel the new one.  (We have to do this here,
-                 * instead of in mdc_enqueue, because we need to use the child's
-                 * inode as the l_data to match, and that's not available until
-                 * intent_finish has performed the iget().) */
-                lock = ldlm_handle2lock(&lockh);
-                if (lock) {
-                        LDLM_DEBUG(lock, "matching against this");
-                        LDLM_LOCK_PUT(lock);
-                        memcpy(&old_lock, &lockh, sizeof(lockh));
-                        if (ldlm_lock_match(NULL,
-                                            LDLM_FL_BLOCK_GRANTED |
-                                            LDLM_FL_MATCH_DATA,
-                                            NULL, LDLM_PLAIN, NULL, 0, LCK_NL,
-                                            inode, &old_lock)) {
-                                ldlm_lock_decref_and_cancel(&lockh,
-                                                            it->it_lock_mode);
-                                memcpy(&lockh, &old_lock, sizeof(old_lock));
-                                memcpy(it->it_lock_handle, &lockh,
-                                       sizeof(lockh));
-                        }
-                }
-
-        }
-        ptlrpc_req_finished(request);
-
-        CDEBUG(D_DENTRY, "D_IT dentry %p intent: %s status %d disp %x\n",
-               dentry, ldlm_it2str(it->it_op), it->it_status, it->it_disposition);
-        
-        /* drop IT_LOOKUP locks */
-        if (it->it_op == IT_LOOKUP)
-                ll_intent_release(it);
-        RETURN(rc);
-
- drop_lock:
-        ll_intent_release(it);
- drop_req:
-        ptlrpc_req_finished(request);
-        RETURN(rc);
+        data->create_mode = mode;
+        data->mod_time = LTIME_S(CURRENT_TIME);
 }
 
 /* Search "inode"'s alias list for a dentry that has the same name and parent as
@@ -487,11 +258,12 @@ struct dentry *ll_find_alias(struct inode *inode, struct dentry *de)
         return de;
 }
 
-static int
-lookup2_finish(struct ptlrpc_request *request,
-               struct inode *parent, struct dentry **de,
-               struct lookup_intent *it, int offset, obd_id ino)
+static int lookup_it_finish(struct ptlrpc_request *request, int offset,
+                            struct lookup_intent *it, void *data)
 {
+        struct it_cb_data *icbd = data;
+        struct dentry **de = icbd->icbd_childp;
+        struct inode *parent = icbd->icbd_parent;
         struct ll_sb_info *sbi = ll_i2sbi(parent);
         struct dentry *dentry = *de, *saved = *de;
         struct inode *inode = NULL;
@@ -500,24 +272,13 @@ lookup2_finish(struct ptlrpc_request *request,
         /* NB 1 request reference will be taken away by ll_intent_lock()
          * when I return */
         if (!it_disposition(it, DISP_LOOKUP_NEG)) {
-                struct lustre_md md;
                 ENTRY;
 
-                rc =mdc_req2lustre_md(request, offset, &sbi->ll_osc_conn, &md);
-                if (rc) 
+                rc = ll_prep_inode(sbi->ll_osc_exp, &inode, request, offset,
+                                   dentry->d_sb);
+                if (rc)
                         RETURN(rc);
 
-                inode = ll_iget(dentry->d_sb, ino, &md);
-                if (!inode) {
-                        /* free the lsm if we allocated one above */
-                        if (md.lsm != NULL)
-                                obd_free_memmd(&sbi->ll_osc_conn, &md.lsm);
-                        RETURN(-ENOMEM);
-                } else if (md.lsm != NULL &&
-                           ll_i2info(inode)->lli_smd != md.lsm) {
-                        obd_free_memmd(&sbi->ll_osc_conn, &md.lsm);
-                }
-
                 /* If this is a stat, get the authoritative file size */
                 if (it->it_op == IT_GETATTR && S_ISREG(inode->i_mode) &&
                     ll_i2info(inode)->lli_smd != NULL) {
@@ -538,14 +299,6 @@ lookup2_finish(struct ptlrpc_request *request,
                 }
 
                 dentry = *de = ll_find_alias(inode, dentry);
-
-                /* We asked for a lock on the directory, and may have been
-                 * granted a lock on the inode.  Just in case, fixup the data
-                 * pointer. */
-                CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
-                       inode, inode->i_ino, inode->i_generation);
-                ldlm_lock_set_data((struct lustre_handle*)it->it_lock_handle,
-                                   inode);
         } else {
                 ENTRY;
         }
@@ -559,161 +312,103 @@ lookup2_finish(struct ptlrpc_request *request,
         RETURN(0);
 }
 
+
 static struct dentry *ll_lookup_it(struct inode *parent, struct dentry *dentry,
                                    struct lookup_intent *it, int flags)
 {
         struct dentry *save = dentry, *retval;
+        struct ll_fid pfid;
+        struct ll_uctxt ctxt;
+        struct it_cb_data icbd;
+        struct ptlrpc_request *req = NULL;
+        struct lookup_intent lookup_it = { .it_op = IT_LOOKUP };
         int rc;
         ENTRY;
 
+        if (dentry->d_name.len > EXT3_NAME_LEN)
+                RETURN(ERR_PTR(-ENAMETOOLONG));
+
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p),intent=%s\n",
                dentry->d_name.name, parent->i_ino, parent->i_generation,
                parent, LL_IT2STR(it));
 
-        if (d_mountpoint(dentry)) { 
+        if (d_mountpoint(dentry))
                 CERROR("Tell Peter, lookup on mtpt, it %s\n", LL_IT2STR(it));
-        }
 
-        rc = ll_intent_lock(parent, &dentry, it, flags, lookup2_finish);
-        if (rc < 0) {
-                CDEBUG(D_INFO, "ll_intent_lock: %d\n", rc);
+        ll_frob_intent(&it, &lookup_it);
+
+        icbd.icbd_childp = &dentry;
+        icbd.icbd_parent = parent;
+        ll_inode2fid(&pfid, parent);
+        ll_i2uctxt(&ctxt, parent, NULL);
+
+        rc = mdc_intent_lock(ll_i2mdcexp(parent), &ctxt, &pfid,
+                             dentry->d_name.name, dentry->d_name.len, NULL, 0,
+                             NULL, it, flags, &req, ll_mdc_blocking_ast);
+        if (rc < 0)
+                GOTO(out, retval = ERR_PTR(rc));
+
+        rc = lookup_it_finish(req, 1, it, &icbd);
+        if (rc != 0) {
+                ll_intent_release(it);
                 GOTO(out, retval = ERR_PTR(rc));
         }
 
+        ll_lookup_finish_locks(it, dentry);
+
         if (dentry == save)
                 GOTO(out, retval = NULL);
         else
                 GOTO(out, retval = dentry);
  out:
+        if (req)
+                ptlrpc_req_finished(req);
         return retval;
 }
 
 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
-static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry, 
+static struct dentry *ll_lookup_nd(struct inode *parent, struct dentry *dentry,
                                    struct nameidata *nd)
 {
         struct dentry *de;
         ENTRY;
 
         if (nd->flags & LOOKUP_LAST && !(nd->flags & LOOKUP_LINK_NOTLAST))
-                de = ll_lookup_it(parent, dentry, &nd->it, nd->flags);
-        else 
+                de = ll_lookup_it(parent, dentry, &nd->intent, nd->flags);
+        else
                 de = ll_lookup_it(parent, dentry, NULL, 0);
 
         RETURN(de);
 }
 #endif
 
-static int ll_mdc_unlink(struct inode *dir, struct inode *child, __u32 mode,
-                         const char *name, int len)
-{
-        struct ptlrpc_request *request = NULL;
-        struct mds_body *body;
-        struct lov_mds_md *eadata;
-        struct lov_stripe_md *lsm = NULL;
-        struct obd_trans_info oti = { 0 };
-        struct mdc_op_data op_data;
-        struct obdo *oa;
-        int rc;
-        ENTRY;
-
-        ll_prepare_mdc_op_data(&op_data, dir, child, name, len, mode);
-        rc = mdc_unlink(&ll_i2sbi(dir)->ll_mdc_conn, &op_data, &request);
-        if (rc)
-                GOTO(out, rc);
-        /* req is swabbed so this is safe */
-        body = lustre_msg_buf(request->rq_repmsg, 0, sizeof(*body));
-
-        if (!(body->valid & OBD_MD_FLEASIZE))
-                GOTO(out, rc = 0);
-
-        if (body->eadatasize == 0) {
-                CERROR("OBD_MD_FLEASIZE set but eadatasize zero\n");
-                GOTO(out, rc = -EPROTO);
-        }
-
-        /* The MDS sent back the EA because we unlinked the last reference
-         * to this file. Use this EA to unlink the objects on the OST.
-         * It's opaque so we don't swab here; we leave it to obd_unpackmd() to
-         * check it is complete and sensible. */
-        eadata = lustre_swab_repbuf(request, 1, body->eadatasize, NULL);
-        LASSERT(eadata != NULL);
-        if (eadata == NULL) {
-                CERROR("Can't unpack MDS EA data\n");
-                GOTO(out, rc = -EPROTO);
-        }
-
-        rc = obd_unpackmd(ll_i2obdconn(dir), &lsm, eadata, body->eadatasize);
-        if (rc < 0) {
-                CERROR("obd_unpackmd: %d\n", rc);
-                GOTO(out, rc);
-        }
-        LASSERT(rc >= sizeof(*lsm));
-
-        oa = obdo_alloc();
-        if (oa == NULL)
-                GOTO(out_free_memmd, rc = -ENOMEM);
-
-        oa->o_id = lsm->lsm_object_id;
-        oa->o_mode = body->mode & S_IFMT;
-        oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE;
-
-        if (body->valid & OBD_MD_FLCOOKIE) {
-                oa->o_valid |= OBD_MD_FLCOOKIE;
-                oti.oti_logcookies = lustre_msg_buf(request->rq_repmsg, 3,
-                                                    body->eadatasize);
-        }
-
-        rc = obd_destroy(ll_i2obdconn(dir), oa, lsm, &oti);
-        obdo_free(oa);
-        if (rc)
-                CERROR("obd destroy objid "LPX64": rc %d\n",
-                       lsm->lsm_object_id, rc);
- out_free_memmd:
-        obd_free_memmd(ll_i2obdconn(dir), &lsm);
- out:
-        ptlrpc_req_finished(request);
-        return rc;
-}
-
 /* We depend on "mode" being set with the proper file type/umask by now */
 static struct inode *ll_create_node(struct inode *dir, const char *name,
                                     int namelen, const void *data, int datalen,
                                     int mode, __u64 extra,
                                     struct lookup_intent *it)
 {
-        struct inode *inode;
+        struct inode *inode = NULL;
         struct ptlrpc_request *request = NULL;
         struct ll_sb_info *sbi = ll_i2sbi(dir);
-        struct lustre_md md;
         int rc;
         ENTRY;
 
-        LASSERT(it && it->it_disposition);
-
-        ll_invalidate_inode_pages(dir);
+        LASSERT(it && it->d.lustre.it_disposition);
 
-        request = it->it_data;
-        rc = mdc_req2lustre_md(request, 1, &sbi->ll_osc_conn, &md);
-        if (rc) { 
+        request = it->d.lustre.it_data;
+        rc = ll_prep_inode(sbi->ll_osc_exp, &inode, request, 1, dir->i_sb);
+        if (rc)
                 GOTO(out, inode = ERR_PTR(rc));
-        }
 
-        inode = ll_iget(dir->i_sb, md.body->ino, &md);
-        if (!inode || is_bad_inode(inode)) {
-                /* XXX might need iput() for bad inode */
-                int rc = -EIO;
-                CERROR("new_inode -fatal: rc %d\n", rc);
-                LBUG();
-                GOTO(out, rc);
-        }
         LASSERT(list_empty(&inode->i_dentry));
 
-        CDEBUG(D_DLMTRACE, "setting l_data to inode %p (%lu/%u)\n",
+        /* We asked for a lock on the directory, but were granted a
+         * lock on the inode.  Since we finally have an inode pointer,
+         * stuff it in the lock. */
+        CDEBUG(D_DLMTRACE, "setting l_ast_data to inode %p (%lu/%u)\n",
                inode, inode->i_ino, inode->i_generation);
-        ldlm_lock_set_data((struct lustre_handle*)it->it_lock_handle,
-                           inode);
-
+        mdc_set_lock_data(&it->d.lustre.it_lock_handle, inode);
         EXIT;
  out:
         ptlrpc_req_finished(request);
@@ -734,10 +429,11 @@ static struct inode *ll_create_node(struct inode *dir, const char *name,
  * If the create succeeds, we fill in the inode information
  * with d_instantiate().
  */
-static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, struct lookup_intent *it)
+static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode,
+                        struct lookup_intent *it)
 {
         struct inode *inode;
-        struct ptlrpc_request *request = it->it_data;
+        struct ptlrpc_request *request = it->d.lustre.it_data;
         int rc = 0;
         ENTRY;
 
@@ -745,11 +441,9 @@ static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, stru
                dentry->d_name.name, dir->i_ino, dir->i_generation, dir,
                LL_IT2STR(it));
 
-        rc = ll_it_open_error(DISP_OPEN_CREATE, it);
-        if (rc) {
-                ptlrpc_req_finished(request);
+        rc = it_open_error(DISP_OPEN_CREATE, it);
+        if (rc)
                 RETURN(rc);
-        }
 
         mdc_store_inode_generation(request, 2, 1);
         inode = ll_create_node(dir, dentry->d_name.name, dentry->d_name.len,
@@ -765,17 +459,16 @@ static int ll_create_it(struct inode *dir, struct dentry *dentry, int mode, stru
 #if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
 static int ll_create_nd(struct inode *dir, struct dentry *dentry, int mode, struct nameidata *nd)
 {
-        return ll_create_it(dir, dentry, mode, &nd->it);
+        return ll_create_it(dir, dentry, mode, &nd->intent);
 }
 #endif
 
 static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev)
 {
+        struct ptlrpc_request *request = NULL;
         struct inode *dir = nd->dentry->d_inode;
         const char *name = nd->last.name;
         int len = nd->last.len;
-        struct ptlrpc_request *request = NULL;
-        time_t time = LTIME_S(CURRENT_TIME);
         struct ll_sb_info *sbi = ll_i2sbi(dir);
         struct mdc_op_data op_data;
         int err = -EMLINK;
@@ -784,22 +477,22 @@ static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev)
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
                name, dir->i_ino, dir->i_generation, dir);
 
-        if (dir->i_nlink >= EXT2_LINK_MAX)
+        if (dir->i_nlink >= EXT3_LINK_MAX)
                 RETURN(err);
 
         mode &= ~current->fs->umask;
 
         switch (mode & S_IFMT) {
-        case 0: 
+        case 0:
         case S_IFREG:
                 mode |= S_IFREG; /* for mode = 0 case, fallthrough */
-        case S_IFCHR: 
+        case S_IFCHR:
         case S_IFBLK:
-        case S_IFIFO: 
+        case S_IFIFO:
         case S_IFSOCK:
                 ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
-                err = mdc_create(&sbi->ll_mdc_conn, &op_data, NULL, 0, mode,
-                                 current->fsuid, current->fsgid, time,
+                err = mdc_create(sbi->ll_mdc_exp, &op_data, NULL, 0, mode,
+                                 current->fsuid, current->fsgid,
                                  rdev, &request);
                 ptlrpc_req_finished(request);
                 break;
@@ -812,13 +505,62 @@ static int ll_mknod_raw(struct nameidata *nd, int mode, dev_t rdev)
         RETURN(err);
 }
 
+static int ll_mknod(struct inode *dir, struct dentry *child, int mode,
+                    ll_dev_t rdev)
+{
+        struct ptlrpc_request *request = NULL;
+        struct inode *inode = NULL;
+        const char *name = child->d_name.name;
+        int len = child->d_name.len;
+        struct ll_sb_info *sbi = ll_i2sbi(dir);
+        struct mdc_op_data op_data;
+        int err = -EMLINK;
+        ENTRY;
+
+        CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
+               name, dir->i_ino, dir->i_generation, dir);
+
+        if (dir->i_nlink >= EXT3_LINK_MAX)
+                RETURN(err);
+
+        mode &= ~current->fs->umask;
+
+        switch (mode & S_IFMT) {
+        case 0:
+        case S_IFREG:
+                mode |= S_IFREG; /* for mode = 0 case, fallthrough */
+        case S_IFCHR:
+        case S_IFBLK:
+        case S_IFIFO:
+        case S_IFSOCK:
+                ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+                err = mdc_create(sbi->ll_mdc_exp, &op_data, NULL, 0, mode,
+                                 current->fsuid, current->fsgid,
+                                 rdev, &request);
+                err = ll_prep_inode(sbi->ll_osc_exp, &inode, request, 0,
+                                    child->d_sb);
+                if (err)
+                        GOTO(out_err, err);
+                break;
+        case S_IFDIR:
+                RETURN(-EPERM);
+                break;
+        default:
+                RETURN(-EINVAL);
+        }
+
+        d_instantiate(child, inode);
+ out_err:
+        ptlrpc_req_finished(request);
+        RETURN(err);
+}
+
 static int ll_symlink_raw(struct nameidata *nd, const char *tgt)
 {
         struct inode *dir = nd->dentry->d_inode;
         const char *name = nd->last.name;
         int len = nd->last.len;
         struct ptlrpc_request *request = NULL;
-        time_t time = LTIME_S(CURRENT_TIME);
         struct ll_sb_info *sbi = ll_i2sbi(dir);
         struct mdc_op_data op_data;
         int err = -EMLINK;
@@ -827,13 +569,13 @@ static int ll_symlink_raw(struct nameidata *nd, const char *tgt)
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p),target=%s\n",
                name, dir->i_ino, dir->i_generation, dir, tgt);
 
-        if (dir->i_nlink >= EXT2_LINK_MAX)
+        if (dir->i_nlink >= EXT3_LINK_MAX)
                 RETURN(err);
 
         ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
-        err = mdc_create(&sbi->ll_mdc_conn, &op_data,
+        err = mdc_create(sbi->ll_mdc_exp, &op_data,
                          tgt, strlen(tgt) + 1, S_IFLNK | S_IRWXUGO,
-                         current->fsuid, current->fsgid, time, 0, &request);
+                         current->fsuid, current->fsgid, 0, &request);
         ptlrpc_req_finished(request);
         RETURN(err);
 }
@@ -855,7 +597,7 @@ static int ll_link_raw(struct nameidata *srcnd, struct nameidata *tgtnd)
                dir->i_ino, dir->i_generation, dir, name);
 
         ll_prepare_mdc_op_data(&op_data, src, dir, name, len, 0);
-        err = mdc_link(&sbi->ll_mdc_conn, &op_data, &request);
+        err = mdc_link(sbi->ll_mdc_exp, &op_data, &request);
         ptlrpc_req_finished(request);
 
         RETURN(err);
@@ -868,7 +610,6 @@ static int ll_mkdir_raw(struct nameidata *nd, int mode)
         const char *name = nd->last.name;
         int len = nd->last.len;
         struct ptlrpc_request *request = NULL;
-        time_t time = LTIME_S(CURRENT_TIME);
         struct ll_sb_info *sbi = ll_i2sbi(dir);
         struct mdc_op_data op_data;
         int err = -EMLINK;
@@ -876,13 +617,13 @@ static int ll_mkdir_raw(struct nameidata *nd, int mode)
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
                name, dir->i_ino, dir->i_generation, dir);
 
-        if (dir->i_nlink >= EXT2_LINK_MAX)
+        if (dir->i_nlink >= EXT3_LINK_MAX)
                 RETURN(err);
 
         mode = (mode & (S_IRWXUGO|S_ISVTX) & ~current->fs->umask) | S_IFDIR;
         ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
-        err = mdc_create(&sbi->ll_mdc_conn, &op_data, NULL, 0, mode,
-                         current->fsuid, current->fsgid, time, 0, &request);
+        err = mdc_create(sbi->ll_mdc_exp, &op_data, NULL, 0, mode,
+                         current->fsuid, current->fsgid, 0, &request);
         ptlrpc_req_finished(request);
         RETURN(err);
 }
@@ -892,26 +633,109 @@ static int ll_rmdir_raw(struct nameidata *nd)
         struct inode *dir = nd->dentry->d_inode;
         const char *name = nd->last.name;
         int len = nd->last.len;
+        struct ptlrpc_request *request = NULL;
+        struct mdc_op_data op_data;
         int rc;
         ENTRY;
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
                name, dir->i_ino, dir->i_generation, dir);
 
-        rc = ll_mdc_unlink(dir, NULL, S_IFDIR, name, len);
+        ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, S_IFDIR);
+        rc = mdc_unlink(ll_i2sbi(dir)->ll_mdc_exp, &op_data, &request);
+        ptlrpc_req_finished(request);
         RETURN(rc);
 }
 
+int ll_objects_destroy(struct ptlrpc_request *request, struct inode *dir)
+{
+        struct mds_body *body;
+        struct lov_mds_md *eadata;
+        struct lov_stripe_md *lsm = NULL;
+        struct obd_trans_info oti = { 0 };
+        struct obdo *oa;
+        int rc;
+        ENTRY;
+
+        /* req is swabbed so this is safe */
+        body = lustre_msg_buf(request->rq_repmsg, 0, sizeof(*body));
+
+        if (!(body->valid & OBD_MD_FLEASIZE))
+                RETURN(0);
+
+        if (body->eadatasize == 0) {
+                CERROR("OBD_MD_FLEASIZE set but eadatasize zero\n");
+                GOTO(out, rc = -EPROTO);
+        }
+
+        /* The MDS sent back the EA because we unlinked the last reference
+         * to this file. Use this EA to unlink the objects on the OST.
+         * It's opaque so we don't swab here; we leave it to obd_unpackmd() to
+         * check it is complete and sensible. */
+        eadata = lustre_swab_repbuf(request, 1, body->eadatasize, NULL);
+        LASSERT(eadata != NULL);
+        if (eadata == NULL) {
+                CERROR("Can't unpack MDS EA data\n");
+                GOTO(out, rc = -EPROTO);
+        }
+
+        rc = obd_unpackmd(ll_i2obdexp(dir), &lsm, eadata, body->eadatasize);
+        if (rc < 0) {
+                CERROR("obd_unpackmd: %d\n", rc);
+                GOTO(out, rc);
+        }
+        LASSERT(rc >= sizeof(*lsm));
+
+        oa = obdo_alloc();
+        if (oa == NULL)
+                GOTO(out_free_memmd, rc = -ENOMEM);
+
+        oa->o_id = lsm->lsm_object_id;
+        oa->o_mode = body->mode & S_IFMT;
+        oa->o_valid = OBD_MD_FLID | OBD_MD_FLTYPE;
+
+        if (body->valid & OBD_MD_FLCOOKIE) {
+                oa->o_valid |= OBD_MD_FLCOOKIE;
+                oti.oti_logcookies =
+                        lustre_msg_buf(request->rq_repmsg, 2,
+                                       sizeof(struct llog_cookie) *
+                                       lsm->lsm_stripe_count);
+                if (oti.oti_logcookies == NULL) {
+                        oa->o_valid &= ~OBD_MD_FLCOOKIE;
+                        body->valid &= ~OBD_MD_FLCOOKIE;
+                }
+        }
+
+        rc = obd_destroy(ll_i2obdexp(dir), oa, lsm, &oti);
+        obdo_free(oa);
+        if (rc)
+                CERROR("obd destroy objid "LPX64" error %d\n",
+                       lsm->lsm_object_id, rc);
+ out_free_memmd:
+        obd_free_memmd(ll_i2obdexp(dir), &lsm);
+ out:
+        return rc;
+}
+
 static int ll_unlink_raw(struct nameidata *nd)
 {
         struct inode *dir = nd->dentry->d_inode;
         const char *name = nd->last.name;
         int len = nd->last.len;
+        struct ptlrpc_request *request = NULL;
+        struct mdc_op_data op_data;
         int rc;
         ENTRY;
         CDEBUG(D_VFSTRACE, "VFS Op:name=%s,dir=%lu/%u(%p)\n",
                name, dir->i_ino, dir->i_generation, dir);
 
-        rc = ll_mdc_unlink(dir, NULL, S_IFREG, name, len);
+        ll_prepare_mdc_op_data(&op_data, dir, NULL, name, len, 0);
+        rc = mdc_unlink(ll_i2sbi(dir)->ll_mdc_exp, &op_data, &request);
+        if (rc)
+                GOTO(out, rc);
+
+        rc = ll_objects_destroy(request, dir);
+ out:
+        ptlrpc_req_finished(request);
         RETURN(rc);
 }
 
@@ -933,8 +757,12 @@ static int ll_rename_raw(struct nameidata *oldnd, struct nameidata *newnd)
                src, newname, tgt->i_ino, tgt->i_generation, tgt);
 
         ll_prepare_mdc_op_data(&op_data, src, tgt, NULL, 0, 0);
-        err = mdc_rename(&sbi->ll_mdc_conn, &op_data,
+        err = mdc_rename(sbi->ll_mdc_exp, &op_data,
                          oldname, oldlen, newname, newlen, &request);
+        if (!err) {
+                err = ll_objects_destroy(request, src);
+        }
+
         ptlrpc_req_finished(request);
 
         RETURN(err);
@@ -947,6 +775,7 @@ struct inode_operations ll_dir_inode_operations = {
         mkdir_raw:          ll_mkdir_raw,
         rmdir_raw:          ll_rmdir_raw,
         mknod_raw:          ll_mknod_raw,
+        mknod:              ll_mknod,
         rename_raw:         ll_rename_raw,
         setattr:         ll_setattr,
         setattr_raw:     ll_setattr_raw,
@@ -955,8 +784,8 @@ struct inode_operations ll_dir_inode_operations = {
         lookup_it:            ll_lookup_it,
         revalidate_it:      ll_inode_revalidate_it,
 #else
-        lookup_it:          ll_lookup_nd,
-        create_nd:          ll_create_nd,
+        lookup:          ll_lookup_nd,
+        create:          ll_create_nd,
         getattr_it:         ll_getattr,
 #endif
 };