lustre/osc/osc_request.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  * Copyright (C) 2001, 2002 Cluster File Systems, Inc.
   5  *
   6  *  This code is issued under the GNU General Public License.
   7  *  See the file COPYING in this distribution
   8  *
   9  *  Author Peter Braam <braam@clusterfs.com>
  10  *
  11  *  This server is single threaded at present (but can easily be multi
  12  *  threaded). For testing and management it is treated as an
  13  *  obd_device, although it does not export a full OBD method table
  14  *  (the requests are coming in over the wire, so object target
  15  *  modules do not have a full method table.)
  16  *
  17  */
  18
  19 #define EXPORT_SYMTAB
  20 #define DEBUG_SUBSYSTEM S_OSC
  21
  22 #include <linux/module.h>
  23 #include <linux/lustre_dlm.h>
  24 #include <linux/lustre_mds.h> /* for mds_objid */
  25 #include <linux/obd_ost.h>
  26 #include <linux/obd_lov.h>
  27
  28 static void osc_con2cl(struct lustre_handle *conn, struct ptlrpc_client **cl,
  29                        struct ptlrpc_connection **connection)
  30 {
  31         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
  32         *cl = osc->osc_client;
  33         *connection = osc->osc_conn;
  34 }
  35
  36 static void osc_con2dlmcl(struct lustre_handle *conn, struct ptlrpc_client **cl,
  37                           struct ptlrpc_connection **connection)
  38 {
  39         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
  40         *cl = osc->osc_ldlm_client;
  41         *connection = osc->osc_conn;
  42 }
  43
  44 static int osc_connect(struct lustre_handle *conn, struct obd_device *obd)
  45 {
  46         struct osc_obd *osc = &obd->u.osc;
  47         struct obd_import *import;
  48         struct ptlrpc_request *request;
  49         char *tmp = osc->osc_target_uuid;
  50         int rc, size = sizeof(osc->osc_target_uuid);
  51         ENTRY;
  52
  53         OBD_ALLOC(import, sizeof(*import));
  54         if (!import)
  55                 RETURN(-ENOMEM);
  56
  57         MOD_INC_USE_COUNT;
  58         rc = class_connect(conn, obd);
  59         if (rc)
  60                 RETURN(rc);
  61
  62         request = ptlrpc_prep_req(osc->osc_client, osc->osc_conn,
  63                                   OST_CONNECT, 1, &size, &tmp);
  64         if (!request)
  65                 GOTO(out_disco, rc = -ENOMEM);
  66
  67         request->rq_level = LUSTRE_CONN_NEW;
  68         request->rq_replen = lustre_msg_size(0, NULL);
  69
  70         rc = ptlrpc_queue_wait(request);
  71         rc = ptlrpc_check_status(request, rc);
  72         if (rc) {
  73                 CERROR("%s failed: rc = %d\n", __FUNCTION__, rc);
  74                 GOTO(out, rc);
  75         }
  76
  77         /* XXX: Make this a handle */
  78         osc->osc_connh.addr = request->rq_repmsg->addr;
  79         osc->osc_connh.cookie = request->rq_repmsg->cookie;
  80
  81         EXIT;
  82  out:
  83         ptlrpc_free_req(request);
  84  out_disco:
  85         if (rc) {
  86                 class_disconnect(conn);
  87                 MOD_DEC_USE_COUNT;
  88         }
  89         return rc;
  90 }
  91
  92 static int osc_disconnect(struct lustre_handle *conn)
  93 {
  94         struct ptlrpc_request *request;
  95         struct ptlrpc_client *cl;
  96         struct ptlrpc_connection *connection;
  97         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
  98         int rc;
  99         ENTRY;
 100
 101         osc_con2cl(conn, &cl, &connection);
 102         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 103                                   OST_DISCONNECT, 0, NULL, NULL);
 104         if (!request)
 105                 RETURN(-ENOMEM);
 106         request->rq_replen = lustre_msg_size(0, NULL);
 107
 108         rc = ptlrpc_queue_wait(request);
 109         if (rc)
 110                 GOTO(out, rc);
 111         rc = class_disconnect(conn);
 112         if (!rc)
 113                 MOD_DEC_USE_COUNT;
 114
 115  out:
 116         ptlrpc_free_req(request);
 117         return rc;
 118 }
 119
 120 static int osc_getattr(struct lustre_handle *conn, struct obdo *oa)
 121 {
 122         struct ptlrpc_request *request;
 123         struct ptlrpc_client *cl;
 124         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 125         struct ptlrpc_connection *connection;
 126         struct ost_body *body;
 127         int rc, size = sizeof(*body);
 128         ENTRY;
 129
 130         osc_con2cl(conn, &cl, &connection);
 131         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 132                                    OST_GETATTR, 1, &size, NULL);
 133         if (!request)
 134                 RETURN(-ENOMEM);
 135
 136         body = lustre_msg_buf(request->rq_reqmsg, 0);
 137         memcpy(&body->oa, oa, sizeof(*oa));
 138         body->oa.o_valid = ~0;
 139
 140         request->rq_replen = lustre_msg_size(1, &size);
 141
 142         rc = ptlrpc_queue_wait(request);
 143         rc = ptlrpc_check_status(request, rc);
 144         if (rc) {
 145                 CERROR("%s failed: rc = %d\n", __FUNCTION__, rc);
 146                 GOTO(out, rc);
 147         }
 148
 149         body = lustre_msg_buf(request->rq_repmsg, 0);
 150         CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
 151         if (oa)
 152                 memcpy(oa, &body->oa, sizeof(*oa));
 153
 154         EXIT;
 155  out:
 156         ptlrpc_free_req(request);
 157         return rc;
 158 }
 159
 160 static int osc_open(struct lustre_handle *conn, struct obdo *oa,
 161                     struct lov_stripe_md *md)
 162 {
 163         struct ptlrpc_request *request;
 164         struct ptlrpc_client *cl;
 165         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 166         struct ptlrpc_connection *connection;
 167         struct ost_body *body;
 168         int rc, size = sizeof(*body);
 169         ENTRY;
 170
 171         osc_con2cl(conn, &cl, &connection);
 172         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 173                                    OST_OPEN, 1, &size, NULL);
 174         if (!request)
 175                 RETURN(-ENOMEM);
 176
 177         body = lustre_msg_buf(request->rq_reqmsg, 0);
 178         memcpy(&body->oa, oa, sizeof(*oa));
 179         body->oa.o_valid = (OBD_MD_FLMODE | OBD_MD_FLID);
 180
 181         request->rq_replen = lustre_msg_size(1, &size);
 182
 183         rc = ptlrpc_queue_wait(request);
 184         rc = ptlrpc_check_status(request, rc);
 185         if (rc)
 186                 GOTO(out, rc);
 187
 188         body = lustre_msg_buf(request->rq_repmsg, 0);
 189         CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
 190         if (oa)
 191                 memcpy(oa, &body->oa, sizeof(*oa));
 192
 193         EXIT;
 194  out:
 195         ptlrpc_free_req(request);
 196         return rc;
 197 }
 198
 199 static int osc_close(struct lustre_handle *conn, struct obdo *oa,
 200                      struct lov_stripe_md *md)
 201 {
 202         struct ptlrpc_request *request;
 203         struct ptlrpc_client *cl;
 204         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 205         struct ptlrpc_connection *connection;
 206         struct ost_body *body;
 207         int rc, size = sizeof(*body);
 208         ENTRY;
 209
 210         osc_con2cl(conn, &cl, &connection);
 211         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 212                                    OST_CLOSE, 1, &size, NULL);
 213         if (!request)
 214                 RETURN(-ENOMEM);
 215
 216         oa->o_id = md->lmd_object_id;
 217         oa->o_mode = S_IFREG;
 218         oa->o_valid = (OBD_MD_FLMODE | OBD_MD_FLID);
 219         body = lustre_msg_buf(request->rq_reqmsg, 0);
 220         memcpy(&body->oa, oa, sizeof(*oa));
 221
 222         request->rq_replen = lustre_msg_size(1, &size);
 223
 224         rc = ptlrpc_queue_wait(request);
 225         rc = ptlrpc_check_status(request, rc);
 226         if (rc)
 227                 GOTO(out, rc);
 228
 229         body = lustre_msg_buf(request->rq_repmsg, 0);
 230         CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
 231         if (oa)
 232                 memcpy(oa, &body->oa, sizeof(*oa));
 233
 234         EXIT;
 235  out:
 236         ptlrpc_free_req(request);
 237         return rc;
 238 }
 239
 240 static int osc_setattr(struct lustre_handle *conn, struct obdo *oa)
 241 {
 242         struct ptlrpc_request *request;
 243         struct ptlrpc_client *cl;
 244         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 245         struct ptlrpc_connection *connection;
 246         struct ost_body *body;
 247         int rc, size = sizeof(*body);
 248         ENTRY;
 249
 250         osc_con2cl(conn, &cl, &connection);
 251         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 252                                   OST_SETATTR, 1, &size, NULL);
 253         if (!request)
 254                 RETURN(-ENOMEM);
 255
 256         body = lustre_msg_buf(request->rq_reqmsg, 0);
 257         memcpy(&body->oa, oa, sizeof(*oa));
 258
 259         request->rq_replen = lustre_msg_size(1, &size);
 260
 261         rc = ptlrpc_queue_wait(request);
 262         rc = ptlrpc_check_status(request, rc);
 263         GOTO(out, rc);
 264
 265  out:
 266         ptlrpc_free_req(request);
 267         return rc;
 268 }
 269
 270 static int osc_create(struct lustre_handle *conn, struct obdo *oa,
 271                       struct lov_stripe_md **ea)
 272 {
 273         struct ptlrpc_request *request;
 274         struct ptlrpc_client *cl;
 275         struct ptlrpc_connection *connection;
 276         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 277         struct ost_body *body;
 278         int rc, size = sizeof(*body);
 279         ENTRY;
 280
 281         if (!oa) {
 282                 CERROR("oa NULL\n");
 283                 RETURN(-EINVAL);
 284         }
 285
 286         if (!ea) {
 287                 LBUG();
 288         }
 289
 290         if (!*ea) {
 291                 OBD_ALLOC(*ea, oa->o_easize);
 292                 if (!*ea)
 293                         RETURN(-ENOMEM);
 294                 (*ea)->lmd_size = oa->o_easize;
 295         }
 296
 297         osc_con2cl(conn, &cl, &connection);
 298         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 299                                   OST_CREATE, 1, &size, NULL);
 300         if (!request)
 301                 RETURN(-ENOMEM);
 302
 303         body = lustre_msg_buf(request->rq_reqmsg, 0);
 304         memcpy(&body->oa, oa, sizeof(*oa));
 305
 306         request->rq_replen = lustre_msg_size(1, &size);
 307
 308         rc = ptlrpc_queue_wait(request);
 309         rc = ptlrpc_check_status(request, rc);
 310         if (rc)
 311                 GOTO(out, rc);
 312
 313         body = lustre_msg_buf(request->rq_repmsg, 0);
 314         memcpy(oa, &body->oa, sizeof(*oa));
 315
 316         (*ea)->lmd_object_id = oa->o_id;
 317         (*ea)->lmd_stripe_count = 1;
 318         EXIT;
 319  out:
 320         ptlrpc_free_req(request);
 321         return rc;
 322 }
 323
 324 static int osc_punch(struct lustre_handle *conn, struct obdo *oa,
 325                      struct lov_stripe_md *md, obd_size count,
 326                      obd_off offset)
 327 {
 328         struct ptlrpc_request *request;
 329         struct ptlrpc_client *cl;
 330         struct ptlrpc_connection *connection;
 331         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 332         struct ost_body *body;
 333         int rc, size = sizeof(*body);
 334         ENTRY;
 335
 336         if (!oa) {
 337                 CERROR("oa NULL\n");
 338                 RETURN(-EINVAL);
 339         }
 340         osc_con2cl(conn, &cl, &connection);
 341         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 342                                    OST_PUNCH, 1, &size, NULL);
 343         if (!request)
 344                 RETURN(-ENOMEM);
 345
 346         body = lustre_msg_buf(request->rq_reqmsg, 0);
 347         memcpy(&body->oa, oa, sizeof(*oa));
 348         body->oa.o_blocks = count;
 349         body->oa.o_valid |= OBD_MD_FLBLOCKS;
 350
 351         request->rq_replen = lustre_msg_size(1, &size);
 352
 353         rc = ptlrpc_queue_wait(request);
 354         rc = ptlrpc_check_status(request, rc);
 355         if (rc)
 356                 GOTO(out, rc);
 357
 358         body = lustre_msg_buf(request->rq_repmsg, 0);
 359         memcpy(oa, &body->oa, sizeof(*oa));
 360
 361         EXIT;
 362  out:
 363         ptlrpc_free_req(request);
 364         return rc;
 365 }
 366
 367 static int osc_destroy(struct lustre_handle *conn, struct obdo *oa,
 368                        struct lov_stripe_md *ea)
 369 {
 370         struct ptlrpc_request *request;
 371         struct ptlrpc_client *cl;
 372         struct ptlrpc_connection *connection;
 373         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 374         struct ost_body *body;
 375         int rc, size = sizeof(*body);
 376         ENTRY;
 377
 378         if (!oa) {
 379                 CERROR("oa NULL\n");
 380                 RETURN(-EINVAL);
 381         }
 382         osc_con2cl(conn, &cl, &connection);
 383         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 384                                    OST_DESTROY, 1, &size, NULL);
 385         if (!request)
 386                 RETURN(-ENOMEM);
 387
 388         body = lustre_msg_buf(request->rq_reqmsg, 0);
 389         memcpy(&body->oa, oa, sizeof(*oa));
 390         body->oa.o_valid = ~0;
 391
 392         request->rq_replen = lustre_msg_size(1, &size);
 393
 394         rc = ptlrpc_queue_wait(request);
 395         rc = ptlrpc_check_status(request, rc);
 396         if (rc)
 397                 GOTO(out, rc);
 398
 399         body = lustre_msg_buf(request->rq_repmsg, 0);
 400         memcpy(oa, &body->oa, sizeof(*oa));
 401
 402         EXIT;
 403  out:
 404         ptlrpc_free_req(request);
 405         return rc;
 406 }
 407
 408 struct osc_brw_cb_data {
 409         struct page **buf;
 410         bulk_callback_t callback;
 411         void *cb_data;
 412         void *obd_data;
 413         size_t obd_size;
 414 };
 415
 416 static void brw_finish(struct ptlrpc_bulk_desc *desc, void *data)
 417 {
 418         struct osc_brw_cb_data *cb_data = data;
 419         int i;
 420         ENTRY;
 421
 422         if (desc->b_flags & PTL_RPC_FL_INTR)
 423                 CERROR("got signal\n");
 424
 425         for (i = 0; i < desc->b_page_count; i++)
 426                 kunmap(cb_data->buf[i]);
 427
 428         if (cb_data->callback)
 429                 (cb_data->callback)(desc, cb_data->cb_data);
 430
 431         ptlrpc_bulk_decref(desc);
 432         if (cb_data->obd_data)
 433                 OBD_FREE(cb_data->obd_data, cb_data->obd_size);
 434         OBD_FREE(cb_data, sizeof(*cb_data));
 435         EXIT;
 436 }
 437
 438 static int osc_brw_read(struct lustre_handle *conn,
 439                         struct lov_stripe_md *md, obd_count page_count,
 440                         struct page **page_array,
 441                         obd_size *count, obd_off *offset, obd_flag *flags,
 442                         bulk_callback_t callback)
 443 {
 444         struct ptlrpc_client *cl;
 445         struct ptlrpc_connection *connection;
 446         struct ptlrpc_request *request = NULL;
 447         struct ptlrpc_bulk_desc *desc = NULL;
 448         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 449         struct ost_body *body;
 450         int rc, j, size[3] = {sizeof(*body)};
 451         void *iooptr, *nioptr;
 452         struct osc_brw_cb_data *cb_data = NULL;
 453
 454         ENTRY;
 455
 456         size[1] = sizeof(struct obd_ioobj);
 457         size[2] = page_count * sizeof(struct niobuf_remote);
 458
 459         osc_con2cl(conn, &cl, &connection);
 460         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 461                                    OST_BRW, 3, size, NULL);
 462         if (!request)
 463                 RETURN(-ENOMEM);
 464
 465         body = lustre_msg_buf(request->rq_reqmsg, 0);
 466         body->data = OBD_BRW_READ;
 467
 468         desc = ptlrpc_prep_bulk(connection);
 469         if (!desc)
 470                 GOTO(out_free, rc = -ENOMEM);
 471         desc->b_portal = OST_BULK_PORTAL;
 472         desc->b_cb = brw_finish;
 473         OBD_ALLOC(cb_data, sizeof(*cb_data));
 474         if (!cb_data)
 475                 GOTO(out_free, rc = -ENOMEM);
 476         cb_data->buf = NULL;
 477         cb_data->callback = callback;
 478         desc->b_cb_data = cb_data;
 479         /* XXX end almost identical to brw_write case */
 480
 481         iooptr = lustre_msg_buf(request->rq_reqmsg, 1);
 482         nioptr = lustre_msg_buf(request->rq_reqmsg, 2);
 483         ost_pack_ioo(&iooptr, md, page_count);
 484         for (j = 0; j < page_count; j++) {
 485                 struct ptlrpc_bulk_page *bulk;
 486                 bulk = ptlrpc_prep_bulk_page(desc);
 487                 if (bulk == NULL)
 488                         GOTO(out_unmap, rc = -ENOMEM);
 489
 490                 spin_lock(&connection->c_lock);
 491                 bulk->b_xid = ++connection->c_xid_out;
 492                 spin_unlock(&connection->c_lock);
 493
 494                 bulk->b_buf = kmap(page_array[j]);
 495                 bulk->b_page = page_array[j];
 496                 bulk->b_buflen = PAGE_SIZE;
 497                 ost_pack_niobuf(&nioptr, offset[j], count[j],
 498                                 flags[j], bulk->b_xid);
 499         }
 500
 501         /*
 502          * Register the bulk first, because the reply could arrive out of order,
 503          * and we want to be ready for the bulk data.
 504          *
 505          * One reference is released by the bulk callback, the other when
 506          * we finish sleeping on it (if we don't have a callback).
 507          */
 508         atomic_set(&desc->b_refcount, callback ? 1 : 2);
 509         rc = ptlrpc_register_bulk(desc);
 510         if (rc)
 511                 GOTO(out_unmap, rc);
 512
 513         request->rq_replen = lustre_msg_size(1, size);
 514         rc = ptlrpc_queue_wait(request);
 515         rc = ptlrpc_check_status(request, rc);
 516         if (rc) {
 517                 ptlrpc_bulk_decref(desc);
 518                 GOTO(out_unmap, rc);
 519         }
 520
 521         /* Callbacks cause asynchronous handling. */
 522         if (callback)
 523                 RETURN(0);
 524
 525         l_wait_event_killable(desc->b_waitq, ptlrpc_check_bulk_received(desc));
 526         ptlrpc_bulk_decref(desc);
 527         if (desc->b_flags & PTL_RPC_FL_INTR)
 528                 RETURN(-EINTR);
 529
 530         RETURN(0);
 531
 532         /* Clean up on error. */
 533  out_unmap:
 534         for (j = 0; j < desc->b_page_count; j++)
 535                         kunmap(pagearray[j]);
 536  out_free:
 537         if (cb_data)
 538                 OBD_FREE(cb_data, sizeof(*cb_data));
 539         ptlrpc_free_bulk(desc);
 540         ptlrpc_free_req(request);
 541         return rc;
 542 }
 543
 544 static int osc_brw_write(struct lustre_handle *conn,
 545                          struct lov_stripe_md *md, obd_count page_count,
 546                          struct page **pagearray, obd_size *count,
 547                          obd_off *offset, obd_flag *flags,
 548                          bulk_callback_t callback)
 549 {
 550         struct ptlrpc_client *cl;
 551         struct ptlrpc_connection *connection;
 552         struct ptlrpc_request *request = NULL;
 553         struct ptlrpc_bulk_desc *desc = NULL;
 554         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 555         struct ost_body *body;
 556         struct niobuf_local *local = NULL;
 557         struct niobuf_remote *remote;
 558         struct osc_brw_cb_data *cb_data = NULL;
 559         int rc, j, size[3] = {sizeof(*body)};
 560         void *iooptr, *nioptr;
 561         ENTRY;
 562
 563         size[1] = sizeof(struct obd_ioobj);
 564         size[2] = page_count * sizeof(*remote);
 565
 566         osc_con2cl(conn, &cl, &connection);
 567         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 568                                    OST_BRW, 3, size, NULL);
 569         if (!request)
 570                 RETURN(-ENOMEM);
 571
 572         body = lustre_msg_buf(request->rq_reqmsg, 0);
 573         body->data = OBD_BRW_WRITE;
 574
 575         OBD_ALLOC(local, page_count * sizeof(*local));
 576         if (!local)
 577                 GOTO(out_free, rc = -ENOMEM);
 578
 579         desc = ptlrpc_prep_bulk(connection);
 580         if (!desc)
 581                 GOTO(out_free, rc = -ENOMEM);
 582         desc->b_portal = OSC_BULK_PORTAL;
 583         desc->b_cb = brw_finish;
 584         OBD_ALLOC(cb_data, sizeof(*cb_data));
 585         if (!cb_data)
 586                 GOTO(out_free, rc = -ENOMEM);
 587         cb_data->buf = pagearray;
 588         cb_data->callback = callback;
 589         desc->b_cb_data = cb_data;
 590         /* XXX end almost identical to brw_read case */
 591         cb_data->obd_data = local;
 592         cb_data->obd_size = page_count * sizeof(*local);
 593
 594         iooptr = lustre_msg_buf(request->rq_reqmsg, 1);
 595         nioptr = lustre_msg_buf(request->rq_reqmsg, 2);
 596         ost_pack_ioo(&iooptr, md, page_count);
 597         for (j = 0; j < page_count; j++) {
 598                 local[j].addr = kmap(pagearray[j]);
 599                 local[j].offset = offset[j];
 600                 local[j].len = count[j];
 601                 ost_pack_niobuf(&nioptr, offset[j], count[j],
 602                                 flags[j], 0);
 603         }
 604
 605         size[1] = page_count * sizeof(struct niobuf_remote);
 606         request->rq_replen = lustre_msg_size(2, size);
 607         rc = ptlrpc_queue_wait(request);
 608         rc = ptlrpc_check_status(request, rc);
 609         if (rc)
 610                 GOTO(out_unmap, rc);
 611
 612         nioptr = lustre_msg_buf(request->rq_repmsg, 1);
 613         if (!nioptr)
 614                 GOTO(out_unmap, rc = -EINVAL);
 615
 616         if (request->rq_repmsg->buflens[1] !=
 617             page_count * sizeof(struct niobuf_remote)) {
 618                 CERROR("buffer length wrong (%d vs. %d)\n",
 619                        request->rq_repmsg->buflens[1],
 620                        page_count * sizeof(struct niobuf_remote));
 621                 GOTO(out_unmap, rc = -EINVAL);
 622         }
 623
 624
 625         for (j = 0; j < page_count; j++) {
 626                 struct ptlrpc_bulk_page *page;
 627
 628                 ost_unpack_niobuf(&nioptr, &remote);
 629
 630                 page = ptlrpc_prep_bulk_page(desc);
 631                 if (!page)
 632                         GOTO(out_unmap, rc = -ENOMEM);
 633
 634                 page->b_buf = (void *)(unsigned long)local[j].addr;
 635                 page->b_buflen = local[j].len;
 636                 page->b_xid = remote->xid;
 637         }
 638
 639         if (desc->b_page_count != page_count)
 640                 LBUG();
 641
 642         /*
 643          * One is released when the bulk is complete, the other when we finish
 644          * waiting on it.  (Callback cases don't sleep, so only one ref for
 645          * them.)
 646          */
 647         atomic_set(&desc->b_refcount, callback ? 1 : 2);
 648         CDEBUG(D_PAGE, "Set refcount of %p to %d\n", desc,
 649                atomic_read(&desc->b_refcount));
 650         rc = ptlrpc_send_bulk(desc);
 651         if (rc)
 652                 GOTO(out_unmap, rc);
 653
 654         /* Callbacks cause asynchronous handling. */
 655         if (callback)
 656                 RETURN(0);
 657
 658         /* If there's no callback function, sleep here until complete. */
 659         l_wait_event_killable(desc->b_waitq, ptlrpc_check_bulk_sent(desc));
 660         ptlrpc_bulk_decref(desc);
 661         if (desc->b_flags & PTL_RPC_FL_INTR)
 662                 RETURN(-EINTR);
 663         RETURN(0);
 664
 665         /* Clean up on error. */
 666  out_unmap:
 667         for (j = 0; j < page_count; j++)
 668                 kunmap(pagearray[j]);
 669
 670  out_free:
 671         if (cb_data)
 672                 OBD_FREE(cb_data, sizeof(*cb_data));
 673         if (local)
 674                 OBD_FREE(local, page_count * sizeof(*local));
 675         ptlrpc_free_bulk(desc);
 676         ptlrpc_req_finished(request);
 677         return rc;
 678 }
 679
 680 static int osc_brw(int cmd, struct lustre_handle *conn,
 681                    struct lov_stripe_md *md, obd_count page_count,
 682                    struct page **page_array,
 683                    obd_size *count,
 684                    obd_off *offset,
 685                    obd_flag *flags,
 686                    void *callback)
 687 {
 688         if (cmd & OBD_BRW_WRITE)
 689                 return osc_brw_write(conn, md, page_count, page_array, count,
 690                                      offset, flags, (bulk_callback_t)callback);
 691         else
 692                 return osc_brw_read(conn, md, page_count, page_array, count,
 693                                     offset, flags, (bulk_callback_t)callback);
 694 }
 695
 696 static int osc_enqueue(struct lustre_handle *oconn,
 697                        struct lustre_handle *parent_lock, __u64 *res_id,
 698                        __u32 type, void *extentp, int extent_len, __u32 mode,
 699                        int *flags, void *callback, void *data, int datalen,
 700                        struct lustre_handle *lockh)
 701 {
 702         struct obd_device *obddev = class_conn2obd(oconn);
 703         struct osc_obd *osc = &obddev->u.osc;
 704         struct ptlrpc_connection *conn;
 705         struct ptlrpc_client *cl;
 706         struct ldlm_extent *extent = extentp;
 707         int rc;
 708         __u32 mode2;
 709
 710         /* Filesystem locks are given a bit of special treatment: first we
 711          * fixup the lock to start and end on page boundaries. */
 712         extent->start &= PAGE_MASK;
 713         extent->end = (extent->end + PAGE_SIZE - 1) & PAGE_MASK;
 714
 715         /* Next, search for already existing extent locks that will cover us */
 716         osc_con2dlmcl(oconn, &cl, &conn);
 717         rc = ldlm_lock_match(obddev->obd_namespace, res_id, type, extent,
 718                              sizeof(extent), mode, lockh);
 719         if (rc == 1) {
 720                 /* We already have a lock, and it's referenced */
 721                 return 0;
 722         }
 723
 724         /* Next, search for locks that we can upgrade (if we're trying to write)
 725          * or are more than we need (if we're trying to read).  Because the VFS
 726          * and page cache already protect us locally, lots of readers/writers
 727          * can share a single PW lock. */
 728         if (mode == LCK_PW)
 729                 mode2 = LCK_PR;
 730         else
 731                 mode2 = LCK_PW;
 732
 733         rc = ldlm_lock_match(obddev->obd_namespace, res_id, type, extent,
 734                              sizeof(extent), mode2, lockh);
 735         if (rc == 1) {
 736                 int flags;
 737                 /* FIXME: This is not incredibly elegant, but it might
 738                  * be more elegant than adding another parameter to
 739                  * lock_match.  I want a second opinion. */
 740                 ldlm_lock_addref(lockh, mode);
 741                 ldlm_lock_decref(lockh, mode2);
 742
 743                 if (mode == LCK_PR)
 744                         return 0;
 745
 746                 rc = ldlm_cli_convert(cl, lockh, &osc->osc_connh,
 747                                       mode, &flags);
 748                 if (rc)
 749                         LBUG();
 750
 751                 return rc;
 752         }
 753
 754         rc = ldlm_cli_enqueue(cl, conn, &osc->osc_connh,
 755                               NULL, obddev->obd_namespace,
 756                               parent_lock, res_id, type, extent, sizeof(extent),
 757                               mode, flags, callback, data, datalen, lockh);
 758         return rc;
 759 }
 760
 761 static int osc_cancel(struct lustre_handle *oconn, __u32 mode,
 762                       struct lustre_handle *lockh)
 763 {
 764         ENTRY;
 765
 766         ldlm_lock_decref(lockh, mode);
 767
 768         RETURN(0);
 769 }
 770
 771 static int osc_setup(struct obd_device *obddev, obd_count len, void *buf)
 772 {
 773         struct obd_ioctl_data* data = buf;
 774         struct osc_obd *osc = &obddev->u.osc;
 775         char server_uuid[37];
 776         int rc;
 777         ENTRY;
 778
 779         if (data->ioc_inllen1 < 1) {
 780                 CERROR("osc setup requires a TARGET UUID\n");
 781                 RETURN(-EINVAL);
 782         }
 783
 784         if (data->ioc_inllen1 > 37) {
 785                 CERROR("osc TARGET UUID must be less than 38 characters\n");
 786                 RETURN(-EINVAL);
 787         }
 788
 789         if (data->ioc_inllen2 < 1) {
 790                 CERROR("osc setup requires a SERVER UUID\n");
 791                 RETURN(-EINVAL);
 792         }
 793
 794         if (data->ioc_inllen2 > 37) {
 795                 CERROR("osc SERVER UUID must be less than 38 characters\n");
 796                 RETURN(-EINVAL);
 797         }
 798
 799         memcpy(osc->osc_target_uuid, data->ioc_inlbuf1, data->ioc_inllen1);
 800         memcpy(server_uuid, data->ioc_inlbuf2, MIN(data->ioc_inllen2,
 801                                                    sizeof(server_uuid)));
 802
 803         osc->osc_conn = ptlrpc_uuid_to_connection(server_uuid);
 804         if (!osc->osc_conn)
 805                 RETURN(-ENOENT);
 806
 807         obddev->obd_namespace =
 808                 ldlm_namespace_new("osc", LDLM_NAMESPACE_CLIENT);
 809         if (obddev->obd_namespace == NULL)
 810                 GOTO(out_conn, rc = -ENOMEM);
 811
 812         OBD_ALLOC(osc->osc_client, sizeof(*osc->osc_client));
 813         if (osc->osc_client == NULL)
 814                 GOTO(out_ns, rc = -ENOMEM);
 815
 816         OBD_ALLOC(osc->osc_ldlm_client, sizeof(*osc->osc_ldlm_client));
 817         if (osc->osc_ldlm_client == NULL)
 818                 GOTO(out_client, rc = -ENOMEM);
 819
 820         ptlrpc_init_client(NULL, NULL, OST_REQUEST_PORTAL, OSC_REPLY_PORTAL,
 821                            osc->osc_client);
 822         ptlrpc_init_client(NULL, NULL, LDLM_REQUEST_PORTAL, LDLM_REPLY_PORTAL,
 823                            osc->osc_ldlm_client);
 824         osc->osc_client->cli_name = "osc";
 825         osc->osc_ldlm_client->cli_name = "ldlm";
 826
 827         MOD_INC_USE_COUNT;
 828         RETURN(0);
 829
 830  out_client:
 831         OBD_FREE(osc->osc_client, sizeof(*osc->osc_client));
 832  out_ns:
 833         ldlm_namespace_free(obddev->obd_namespace);
 834  out_conn:
 835         ptlrpc_put_connection(osc->osc_conn);
 836         return rc;
 837 }
 838
 839 static int osc_cleanup(struct obd_device * obddev)
 840 {
 841         struct osc_obd *osc = &obddev->u.osc;
 842
 843         ldlm_namespace_free(obddev->obd_namespace);
 844
 845         ptlrpc_cleanup_client(osc->osc_client);
 846         OBD_FREE(osc->osc_client, sizeof(*osc->osc_client));
 847         ptlrpc_cleanup_client(osc->osc_ldlm_client);
 848         OBD_FREE(osc->osc_ldlm_client, sizeof(*osc->osc_ldlm_client));
 849         ptlrpc_put_connection(osc->osc_conn);
 850
 851         MOD_DEC_USE_COUNT;
 852         return 0;
 853 }
 854
 855 static int osc_statfs(struct lustre_handle *conn, struct statfs *sfs)
 856 {
 857         struct ptlrpc_request *request;
 858         struct ptlrpc_client *cl;
 859         struct osc_obd *osc = &class_conn2obd(conn)->u.osc;
 860         struct ptlrpc_connection *connection;
 861         struct obd_statfs *osfs;
 862         int rc, size = sizeof(*osfs);
 863         ENTRY;
 864
 865         osc_con2cl(conn, &cl, &connection);
 866         request = ptlrpc_prep_req2(cl, connection, &osc->osc_connh,
 867                                    OST_STATFS, 0, NULL, NULL);
 868         if (!request)
 869                 RETURN(-ENOMEM);
 870
 871         request->rq_replen = lustre_msg_size(1, &size);
 872
 873         rc = ptlrpc_queue_wait(request);
 874         rc = ptlrpc_check_status(request, rc);
 875         if (rc) {
 876                 CERROR("%s failed: rc = %d\n", __FUNCTION__, rc);
 877                 GOTO(out, rc);
 878         }
 879
 880         osfs = lustre_msg_buf(request->rq_repmsg, 0);
 881         obd_statfs_unpack(osfs, sfs);
 882
 883         EXIT;
 884  out:
 885         ptlrpc_free_req(request);
 886         return rc;
 887 }
 888
 889 struct obd_ops osc_obd_ops = {
 890         o_setup:        osc_setup,
 891         o_cleanup:      osc_cleanup,
 892         o_statfs:       osc_statfs,
 893         o_create:       osc_create,
 894         o_destroy:      osc_destroy,
 895         o_getattr:      osc_getattr,
 896         o_setattr:      osc_setattr,
 897         o_open:         osc_open,
 898         o_close:        osc_close,
 899         o_connect:      osc_connect,
 900         o_disconnect:   osc_disconnect,
 901         o_brw:          osc_brw,
 902         o_punch:        osc_punch,
 903         o_enqueue:      osc_enqueue,
 904         o_cancel:       osc_cancel
 905 };
 906
 907 static int __init osc_init(void)
 908 {
 909         return class_register_type(&osc_obd_ops, LUSTRE_OSC_NAME);
 910 }
 911
 912 static void __exit osc_exit(void)
 913 {
 914         class_unregister_type(LUSTRE_OSC_NAME);
 915 }
 916
 917 MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
 918 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC) v1.0");
 919 MODULE_LICENSE("GPL");
 920
 921 module_init(osc_init);
 922 module_exit(osc_exit);