lustre/ost/ost_handler.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  *  Copyright (C) 2001-2003 Cluster File Systems, Inc.
   5  *   Author: Peter J. Braam <braam@clusterfs.com>
   6  *   Author: Phil Schwan <phil@clusterfs.com>
   7  *
   8  *   This file is part of Lustre, http://www.lustre.org.
   9  *
  10  *   Lustre is free software; you can redistribute it and/or
  11  *   modify it under the terms of version 2 of the GNU General Public
  12  *   License as published by the Free Software Foundation.
  13  *
  14  *   Lustre is distributed in the hope that it will be useful,
  15  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  *   GNU General Public License for more details.
  18  *
  19  *   You should have received a copy of the GNU General Public License
  20  *   along with Lustre; if not, write to the Free Software
  21  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  22  *
  23  *  Storage Target Handling functions
  24  *  Lustre Object Server Module (OST)
  25  *
  26  *  This server is single threaded at present (but can easily be multi
  27  *  threaded). For testing and management it is treated as an
  28  *  obd_device, although it does not export a full OBD method table
  29  *  (the requests are coming in over the wire, so object target
  30  *  modules do not have a full method table.)
  31  */
  32
  33 #ifndef EXPORT_SYMTAB
  34 # define EXPORT_SYMTAB
  35 #endif
  36 #define DEBUG_SUBSYSTEM S_OST
  37
  38 #include <linux/module.h>
  39 #include <linux/obd_ost.h>
  40 #include <linux/lustre_net.h>
  41 #include <linux/lustre_dlm.h>
  42 #include <linux/lustre_export.h>
  43 #include <linux/init.h>
  44 #include <linux/lprocfs_status.h>
  45 #include <linux/lustre_commit_confd.h>
  46 #include <portals/list.h>
  47
  48 void oti_init(struct obd_trans_info *oti, struct ptlrpc_request *req)
  49 {
  50         if (oti == NULL)
  51                 return;
  52         memset(oti, 0, sizeof *oti);
  53
  54         if (req->rq_repmsg && req->rq_reqmsg != 0)
  55                 oti->oti_transno = req->rq_repmsg->transno;
  56 }
  57
  58 void oti_to_request(struct obd_trans_info *oti, struct ptlrpc_request *req)
  59 {
  60         struct oti_req_ack_lock *ack_lock;
  61         int i;
  62
  63         if (oti == NULL)
  64                 return;
  65
  66         if (req->rq_repmsg)
  67                 req->rq_repmsg->transno = oti->oti_transno;
  68
  69         /* XXX 4 == entries in oti_ack_locks??? */
  70         for (ack_lock = oti->oti_ack_locks, i = 0; i < 4; i++, ack_lock++) {
  71                 if (!ack_lock->mode)
  72                         break;
  73                 ldlm_put_lock_into_req(req, &ack_lock->lock, ack_lock->mode);
  74         }
  75 }
  76
  77 static int ost_destroy(struct obd_export *exp, struct ptlrpc_request *req,
  78                        struct obd_trans_info *oti)
  79 {
  80         struct ost_body *body, *repbody;
  81         int rc, size = sizeof(*body);
  82         ENTRY;
  83
  84         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
  85         if (body == NULL)
  86                 RETURN(-EFAULT);
  87
  88         rc = lustre_pack_reply(req, 1, &size, NULL);
  89         if (rc)
  90                 RETURN(rc);
  91
  92         if (body->oa.o_valid & OBD_MD_FLCOOKIE)
  93                 oti->oti_logcookies = obdo_logcookie(&body->oa);
  94         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
  95         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
  96         req->rq_status = obd_destroy(exp, &body->oa, NULL, oti);
  97         RETURN(0);
  98 }
  99
 100 static int ost_getattr(struct obd_export *exp, struct ptlrpc_request *req)
 101 {
 102         struct ost_body *body, *repbody;
 103         int rc, size = sizeof(*body);
 104         ENTRY;
 105
 106         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 107         if (body == NULL)
 108                 RETURN(-EFAULT);
 109
 110         rc = lustre_pack_reply(req, 1, &size, NULL);
 111         if (rc)
 112                 RETURN(rc);
 113
 114         repbody = lustre_msg_buf (req->rq_repmsg, 0, sizeof(*repbody));
 115         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
 116         req->rq_status = obd_getattr(exp, &repbody->oa, NULL);
 117         RETURN(0);
 118 }
 119
 120 static int ost_statfs(struct ptlrpc_request *req)
 121 {
 122         struct obd_statfs *osfs;
 123         int rc, size = sizeof(*osfs);
 124         ENTRY;
 125
 126         rc = lustre_pack_reply(req, 1, &size, NULL);
 127         if (rc)
 128                 RETURN(rc);
 129
 130         osfs = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*osfs));
 131
 132         req->rq_status = obd_statfs(req->rq_export->exp_obd, osfs, jiffies-HZ);
 133         if (req->rq_status != 0)
 134                 CERROR("ost: statfs failed: rc %d\n", req->rq_status);
 135
 136         RETURN(0);
 137 }
 138
 139 static int ost_create(struct obd_export *exp, struct ptlrpc_request *req,
 140                       struct obd_trans_info *oti)
 141 {
 142         struct ost_body *body, *repbody;
 143         int rc, size = sizeof(*repbody);
 144         ENTRY;
 145
 146         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 147         if (body == NULL)
 148                 RETURN(-EFAULT);
 149
 150         rc = lustre_pack_reply(req, 1, &size, NULL);
 151         if (rc)
 152                 RETURN(rc);
 153
 154         repbody = lustre_msg_buf (req->rq_repmsg, 0, sizeof(*repbody));
 155         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
 156         oti->oti_logcookies = obdo_logcookie(&repbody->oa);
 157         req->rq_status = obd_create(exp, &repbody->oa, NULL, oti);
 158         //obd_log_cancel(conn, NULL, 1, oti->oti_logcookies, 0);
 159         RETURN(0);
 160 }
 161
 162 static int ost_punch(struct obd_export *exp, struct ptlrpc_request *req,
 163                      struct obd_trans_info *oti)
 164 {
 165         struct ost_body *body, *repbody;
 166         int rc, size = sizeof(*repbody);
 167         ENTRY;
 168
 169         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 170         if (body == NULL)
 171                 RETURN(-EFAULT);
 172
 173         if ((body->oa.o_valid & (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS)) !=
 174             (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS))
 175                 RETURN(-EINVAL);
 176
 177         rc = lustre_pack_reply(req, 1, &size, NULL);
 178         if (rc)
 179                 RETURN(rc);
 180
 181         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
 182         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
 183         req->rq_status = obd_punch(exp, &repbody->oa, NULL, repbody->oa.o_size,
 184                                    repbody->oa.o_blocks, oti);
 185         RETURN(0);
 186 }
 187
 188 static int ost_sync(struct obd_export *exp, struct ptlrpc_request *req)
 189 {
 190         struct ost_body *body, *repbody;
 191         int rc, size = sizeof(*repbody);
 192         ENTRY;
 193
 194         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 195         if (body == NULL)
 196                 RETURN(-EFAULT);
 197
 198         rc = lustre_pack_reply(req, 1, &size, NULL);
 199         if (rc)
 200                 RETURN(rc);
 201
 202         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
 203         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
 204         req->rq_status = obd_sync(exp, &repbody->oa, NULL, repbody->oa.o_size,
 205                                   repbody->oa.o_blocks);
 206         RETURN(0);
 207 }
 208
 209 static int ost_setattr(struct obd_export *exp, struct ptlrpc_request *req,
 210                        struct obd_trans_info *oti)
 211 {
 212         struct ost_body *body, *repbody;
 213         int rc, size = sizeof(*repbody);
 214         ENTRY;
 215
 216         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 217         if (body == NULL)
 218                 RETURN(-EFAULT);
 219
 220         rc = lustre_pack_reply(req, 1, &size, NULL);
 221         if (rc)
 222                 RETURN(rc);
 223
 224         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
 225         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
 226
 227         req->rq_status = obd_setattr(exp, &repbody->oa, NULL, oti);
 228         RETURN(0);
 229 }
 230
 231 static int ost_bulk_timeout(void *data)
 232 {
 233         ENTRY;
 234         /* We don't fail the connection here, because having the export
 235          * killed makes the (vital) call to commitrw very sad.
 236          */
 237         RETURN(1);
 238 }
 239
 240 static int get_per_page_niobufs(struct obd_ioobj *ioo, int nioo,
 241                                 struct niobuf_remote *rnb, int nrnb,
 242                                 struct niobuf_remote **pp_rnbp)
 243 {
 244         /* Copy a remote niobuf, splitting it into page-sized chunks
 245          * and setting ioo[i].ioo_bufcnt accordingly */
 246         struct niobuf_remote *pp_rnb;
 247         int   i;
 248         int   j;
 249         int   page;
 250         int   rnbidx = 0;
 251         int   npages = 0;
 252
 253         /* first count and check the number of pages required */
 254         for (i = 0; i < nioo; i++)
 255                 for (j = 0; j < ioo->ioo_bufcnt; j++, rnbidx++) {
 256                         obd_off offset = rnb[rnbidx].offset;
 257                         obd_off p0 = offset >> PAGE_SHIFT;
 258                         obd_off pn = (offset + rnb[rnbidx].len - 1)>>PAGE_SHIFT;
 259
 260                         LASSERT(rnbidx < nrnb);
 261
 262                         npages += (pn + 1 - p0);
 263
 264                         if (rnb[rnbidx].len == 0) {
 265                                 CERROR("zero len BRW: obj %d objid "LPX64
 266                                        " buf %u\n", i, ioo[i].ioo_id, j);
 267                                 return -EINVAL;
 268                         }
 269                         if (j > 0 &&
 270                             rnb[rnbidx].offset <= rnb[rnbidx-1].offset) {
 271                                 CERROR("unordered BRW: obj %d objid "LPX64
 272                                        " buf %u offset "LPX64" <= "LPX64"\n",
 273                                        i, ioo[i].ioo_id, j, rnb[rnbidx].offset,
 274                                        rnb[rnbidx].offset);
 275                                 return -EINVAL;
 276                         }
 277                 }
 278
 279         LASSERT(rnbidx == nrnb);
 280
 281         if (npages == nrnb) {       /* all niobufs are for single pages */
 282                 *pp_rnbp = rnb;
 283                 return npages;
 284         }
 285
 286         OBD_ALLOC(pp_rnb, sizeof(*pp_rnb) * npages);
 287         if (pp_rnb == NULL)
 288                 return -ENOMEM;
 289
 290         /* now do the actual split */
 291         page = rnbidx = 0;
 292         for (i = 0; i < nioo; i++) {
 293                 int  obj_pages = 0;
 294
 295                 for (j = 0; j < ioo[i].ioo_bufcnt; j++, rnbidx++) {
 296                         obd_off off = rnb[rnbidx].offset;
 297                         int     nob = rnb[rnbidx].len;
 298
 299                         LASSERT(rnbidx < nrnb);
 300                         do {
 301                                 obd_off  poff = off & (PAGE_SIZE - 1);
 302                                 int      pnob = (poff + nob > PAGE_SIZE) ?
 303                                                 PAGE_SIZE - poff : nob;
 304
 305                                 LASSERT(page < npages);
 306                                 pp_rnb[page].len = pnob;
 307                                 pp_rnb[page].offset = off;
 308                                 pp_rnb[page].flags = rnb->flags;
 309
 310                                 CDEBUG(D_PAGE, "   obj %d id "LPX64
 311                                        "page %d(%d) "LPX64" for %d\n",
 312                                        i, ioo[i].ioo_id, obj_pages, page,
 313                                        pp_rnb[page].offset, pp_rnb[page].len);
 314                                 page++;
 315                                 obj_pages++;
 316
 317                                 off += pnob;
 318                                 nob -= pnob;
 319                         } while (nob > 0);
 320                         LASSERT(nob == 0);
 321                 }
 322                 ioo[i].ioo_bufcnt = obj_pages;
 323         }
 324         LASSERT(page == npages);
 325
 326         *pp_rnbp = pp_rnb;
 327         return npages;
 328 }
 329
 330 static void free_per_page_niobufs (int npages, struct niobuf_remote *pp_rnb,
 331                                    struct niobuf_remote *rnb)
 332 {
 333         if (pp_rnb == rnb)                      /* didn't allocate above */
 334                 return;
 335
 336         OBD_FREE(pp_rnb, sizeof(*pp_rnb) * npages);
 337 }
 338
 339 #if CHECKSUM_BULK
 340 obd_count ost_checksum_bulk(struct ptlrpc_bulk_desc *desc)
 341 {
 342         obd_count cksum = 0;
 343         struct ptlrpc_bulk_page *bp;
 344
 345         list_for_each_entry(bp, &desc->bd_page_list, bp_link) {
 346                 ost_checksum(&cksum, kmap(bp->bp_page) + bp->bp_pageoffset,
 347                              bp->bp_buflen);
 348                 kunmap(bp->bp_page);
 349         }
 350
 351         return cksum;
 352 }
 353 #endif
 354
 355 static int ost_brw_read(struct ptlrpc_request *req)
 356 {
 357         struct ptlrpc_bulk_desc *desc;
 358         struct niobuf_remote    *remote_nb;
 359         struct niobuf_remote    *pp_rnb;
 360         struct niobuf_local     *local_nb;
 361         struct obd_ioobj        *ioo;
 362         struct ost_body         *body, *repbody;
 363         struct l_wait_info       lwi;
 364         struct obd_trans_info    oti = { 0 };
 365         char                     str[PTL_NALFMT_SIZE];
 366         int                      size[1] = { sizeof(*body) };
 367         int                      comms_error = 0;
 368         int                      niocount;
 369         int                      npages;
 370         int                      nob = 0;
 371         int                      rc;
 372         int                      i;
 373         ENTRY;
 374
 375         if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_READ_BULK))
 376                 GOTO(out, rc = -EIO);
 377
 378         OBD_FAIL_TIMEOUT(OBD_FAIL_OST_BRW_PAUSE_BULK | OBD_FAIL_ONCE,
 379                          (obd_timeout + 1) / 4);
 380
 381         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 382         if (body == NULL) {
 383                 CERROR("Missing/short ost_body\n");
 384                 GOTO(out, rc = -EFAULT);
 385         }
 386
 387         /* BUG 974: when we send back cache grants, don't clear this flag */
 388         body->oa.o_valid &= ~OBD_MD_FLRDEV;
 389
 390         ioo = lustre_swab_reqbuf(req, 1, sizeof(*ioo), lustre_swab_obd_ioobj);
 391         if (ioo == NULL) {
 392                 CERROR("Missing/short ioobj\n");
 393                 GOTO(out, rc = -EFAULT);
 394         }
 395
 396         niocount = ioo->ioo_bufcnt;
 397         remote_nb = lustre_swab_reqbuf(req, 2, niocount * sizeof(*remote_nb),
 398                                        lustre_swab_niobuf_remote);
 399         if (remote_nb == NULL) {
 400                 CERROR("Missing/short niobuf\n");
 401                 GOTO(out, rc = -EFAULT);
 402         }
 403         if (lustre_msg_swabbed(req->rq_reqmsg)) { /* swab remaining niobufs */
 404                 for (i = 1; i < niocount; i++)
 405                         lustre_swab_niobuf_remote (&remote_nb[i]);
 406         }
 407
 408         size[0] = sizeof(*body);
 409         rc = lustre_pack_reply(req, 1, size, NULL);
 410         if (rc)
 411                 GOTO(out, rc);
 412
 413         /* FIXME all niobuf splitting should be done in obdfilter if needed */
 414         /* CAVEAT EMPTOR this sets ioo->ioo_bufcnt to # pages */
 415         npages = get_per_page_niobufs(ioo, 1, remote_nb, niocount, &pp_rnb);
 416         if (npages < 0)
 417                 GOTO(out, rc = npages);
 418
 419         OBD_ALLOC(local_nb, sizeof(*local_nb) * npages);
 420         if (local_nb == NULL)
 421                 GOTO(out_pp_rnb, rc = -ENOMEM);
 422
 423         desc = ptlrpc_prep_bulk_exp(req, BULK_PUT_SOURCE, OST_BULK_PORTAL);
 424         if (desc == NULL)
 425                 GOTO(out_local, rc = -ENOMEM);
 426
 427         rc = obd_preprw(OBD_BRW_READ, req->rq_export, &body->oa, 1,
 428                         ioo, npages, pp_rnb, local_nb, &oti);
 429         if (rc != 0)
 430                 GOTO(out_bulk, rc);
 431
 432         nob = 0;
 433         for (i = 0; i < npages; i++) {
 434                 int page_rc = local_nb[i].rc;
 435
 436                 if (page_rc < 0) {              /* error */
 437                         rc = page_rc;
 438                         break;
 439                 }
 440
 441                 LASSERT(page_rc <= pp_rnb[i].len);
 442                 nob += page_rc;
 443                 if (page_rc != 0) {             /* some data! */
 444                         LASSERT (local_nb[i].page != NULL);
 445                         rc = ptlrpc_prep_bulk_page(desc, local_nb[i].page,
 446                                                    pp_rnb[i].offset& ~PAGE_MASK,
 447                                                    page_rc);
 448                         if (rc != 0)
 449                                 break;
 450                 }
 451
 452                 if (page_rc != pp_rnb[i].len) { /* short read */
 453                         /* All subsequent pages should be 0 */
 454                         while(++i < npages)
 455                                 LASSERT(local_nb[i].rc == 0);
 456                         break;
 457                 }
 458         }
 459
 460         if (rc == 0) {
 461                 rc = ptlrpc_bulk_put(desc);
 462                 if (rc == 0) {
 463                         lwi = LWI_TIMEOUT(obd_timeout * HZ / 4,
 464                                           ost_bulk_timeout, desc);
 465                         rc = l_wait_event(desc->bd_waitq,
 466                                           ptlrpc_bulk_complete(desc), &lwi);
 467                         if (rc) {
 468                                 LASSERT(rc == -ETIMEDOUT);
 469                                 DEBUG_REQ(D_ERROR, req, "timeout on bulk PUT");
 470                                 ptlrpc_abort_bulk(desc);
 471                         }
 472                 } else {
 473                         DEBUG_REQ(D_ERROR, req, "bulk PUT failed: rc %d\n", rc);
 474                 }
 475                 comms_error = rc != 0;
 476         }
 477
 478         /* Must commit after prep above in all cases */
 479         rc = obd_commitrw(OBD_BRW_READ, req->rq_export, &body->oa, 1,
 480                           ioo, npages, local_nb, &oti);
 481
 482         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
 483         memcpy(&repbody->oa, &body->oa, sizeof(repbody->oa));
 484
 485 #if CHECKSUM_BULK
 486         if (rc == 0) {
 487                 repbody->oa.o_nlink = ost_checksum_bulk(desc);
 488                 repbody->oa.o_valid |= OBD_MD_FLCKSUM;
 489         }
 490 #endif
 491
 492  out_bulk:
 493         ptlrpc_free_bulk(desc);
 494  out_local:
 495         OBD_FREE(local_nb, sizeof(*local_nb) * npages);
 496  out_pp_rnb:
 497         free_per_page_niobufs(npages, pp_rnb, remote_nb);
 498  out:
 499         LASSERT(rc <= 0);
 500         if (rc == 0) {
 501                 req->rq_status = nob;
 502                 ptlrpc_reply(req);
 503         } else if (!comms_error) {
 504                 /* only reply if comms OK */
 505                 req->rq_status = rc;
 506                 ptlrpc_error(req);
 507         } else {
 508                 if (req->rq_repmsg != NULL) {
 509                         /* reply out callback would free */
 510                         OBD_FREE(req->rq_repmsg, req->rq_replen);
 511                 }
 512                 if (req->rq_reqmsg->conn_cnt == req->rq_export->exp_conn_cnt) {
 513                         CERROR("bulk IO comms error: "
 514                                "evicting %s@%s nid "LPX64" (%s)\n",
 515                                req->rq_export->exp_client_uuid.uuid,
 516                                req->rq_connection->c_remote_uuid.uuid,
 517                                req->rq_connection->c_peer.peer_nid,
 518                                portals_nid2str(req->rq_connection->c_peer.peer_ni->pni_number,
 519                                                req->rq_connection->c_peer.peer_nid,
 520                                                str));
 521                         ptlrpc_fail_export(req->rq_export);
 522                 } else {
 523                         CERROR("ignoring bulk IO comms error: "
 524                                "client reconnected %s@%s nid "LPX64" (%s)\n",
 525                                req->rq_export->exp_client_uuid.uuid,
 526                                req->rq_connection->c_remote_uuid.uuid,
 527                                req->rq_connection->c_peer.peer_nid,
 528                                portals_nid2str(req->rq_connection->c_peer.peer_ni->pni_number,
 529                                                req->rq_connection->c_peer.peer_nid,
 530                                                str));
 531                 }
 532         }
 533
 534         RETURN(rc);
 535 }
 536
 537 static int ost_brw_write(struct ptlrpc_request *req, struct obd_trans_info *oti)
 538 {
 539         struct ptlrpc_bulk_desc *desc;
 540         struct niobuf_remote    *remote_nb;
 541         struct niobuf_remote    *pp_rnb;
 542         struct niobuf_local     *local_nb;
 543         struct obd_ioobj        *ioo;
 544         struct ost_body         *body, *repbody;
 545         struct l_wait_info       lwi;
 546         __u32                   *rcs;
 547         int                      size[2] = { sizeof(*body) };
 548         int                      objcount, niocount, npages;
 549         int                      comms_error = 0;
 550         int                      rc, rc2, swab, i, j;
 551         char                    str[PTL_NALFMT_SIZE];
 552         ENTRY;
 553
 554         if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_WRITE_BULK))
 555                 GOTO(out, rc = -EIO);
 556
 557         /* pause before transaction has been started */
 558         OBD_FAIL_TIMEOUT(OBD_FAIL_OST_BRW_PAUSE_BULK | OBD_FAIL_ONCE,
 559                          (obd_timeout + 1) / 4);
 560
 561         swab = lustre_msg_swabbed(req->rq_reqmsg);
 562         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 563         if (body == NULL) {
 564                 CERROR("Missing/short ost_body\n");
 565                 GOTO(out, rc = -EFAULT);
 566         }
 567
 568         /* BUG 974: when we send back cache grants, don't clear this flag */
 569         body->oa.o_valid &= ~OBD_MD_FLRDEV;
 570
 571         LASSERT_REQSWAB(req, 1);
 572         objcount = req->rq_reqmsg->buflens[1] / sizeof(*ioo);
 573         if (objcount == 0) {
 574                 CERROR("Missing/short ioobj\n");
 575                 GOTO(out, rc = -EFAULT);
 576         }
 577         ioo = lustre_msg_buf (req->rq_reqmsg, 1, objcount * sizeof(*ioo));
 578         LASSERT (ioo != NULL);
 579         for (niocount = i = 0; i < objcount; i++) {
 580                 if (swab)
 581                         lustre_swab_obd_ioobj (&ioo[i]);
 582                 if (ioo[i].ioo_bufcnt == 0) {
 583                         CERROR("ioo[%d] has zero bufcnt\n", i);
 584                         GOTO(out, rc = -EFAULT);
 585                 }
 586                 niocount += ioo[i].ioo_bufcnt;
 587         }
 588
 589         remote_nb = lustre_swab_reqbuf(req, 2, niocount * sizeof(*remote_nb),
 590                                        lustre_swab_niobuf_remote);
 591         if (remote_nb == NULL) {
 592                 CERROR("Missing/short niobuf\n");
 593                 GOTO(out, rc = -EFAULT);
 594         }
 595         if (swab) {                             /* swab the remaining niobufs */
 596                 for (i = 1; i < niocount; i++)
 597                         lustre_swab_niobuf_remote (&remote_nb[i]);
 598         }
 599
 600         size[1] = niocount * sizeof(*rcs);
 601         rc = lustre_pack_reply(req, 2, size, NULL);
 602         if (rc != 0)
 603                 GOTO(out, rc);
 604         rcs = lustre_msg_buf(req->rq_repmsg, 1, niocount * sizeof(*rcs));
 605
 606         /* FIXME all niobuf splitting should be done in obdfilter if needed */
 607         /* CAVEAT EMPTOR this sets ioo->ioo_bufcnt to # pages */
 608         npages = get_per_page_niobufs(ioo, objcount,remote_nb,niocount,&pp_rnb);
 609         if (npages < 0)
 610                 GOTO(out, rc = npages);
 611
 612         OBD_ALLOC(local_nb, sizeof(*local_nb) * npages);
 613         if (local_nb == NULL)
 614                 GOTO(out_pp_rnb, rc = -ENOMEM);
 615
 616         desc = ptlrpc_prep_bulk_exp(req, BULK_GET_SINK, OST_BULK_PORTAL);
 617         if (desc == NULL)
 618                 GOTO(out_local, rc = -ENOMEM);
 619
 620         rc = obd_preprw(OBD_BRW_WRITE, req->rq_export, &body->oa, objcount,
 621                         ioo, npages, pp_rnb, local_nb, oti);
 622         if (rc != 0)
 623                 GOTO(out_bulk, rc);
 624
 625         /* NB Having prepped, we must commit... */
 626
 627         for (i = 0; i < npages; i++) {
 628                 rc = ptlrpc_prep_bulk_page(desc, local_nb[i].page,
 629                                            pp_rnb[i].offset & (PAGE_SIZE - 1),
 630                                            pp_rnb[i].len);
 631                 if (rc != 0)
 632                         break;
 633         }
 634
 635         if (rc == 0) {
 636                 rc = ptlrpc_bulk_get(desc);
 637                 if (rc == 0) {
 638                         lwi = LWI_TIMEOUT(obd_timeout * HZ / 4,
 639                                           ost_bulk_timeout, desc);
 640                         rc = l_wait_event(desc->bd_waitq,
 641                                           ptlrpc_bulk_complete(desc), &lwi);
 642                         if (rc) {
 643                                 LASSERT(rc == -ETIMEDOUT);
 644                                 DEBUG_REQ(D_ERROR, req, "timeout on bulk GET");
 645                                 ptlrpc_abort_bulk(desc);
 646                         }
 647                 } else {
 648                         DEBUG_REQ(D_ERROR, req, "bulk GET failed: rc %d\n", rc);
 649                 }
 650                 comms_error = rc != 0;
 651         }
 652
 653         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
 654         memcpy(&repbody->oa, &body->oa, sizeof(repbody->oa));
 655
 656 #if CHECKSUM_BULK
 657         if (rc == 0 && (body->oa.o_valid & OBD_MD_FLCKSUM) != 0) {
 658                 static int cksum_counter;
 659                 obd_count client_cksum = body->oa.o_nlink;
 660                 obd_count cksum = ost_checksum_bulk(desc);
 661
 662                 portals_nid2str(req->rq_connection->c_peer.peer_ni->pni_number,
 663                                 req->rq_connection->c_peer.peer_nid, str);
 664                 if (client_cksum != cksum) {
 665                         CERROR("Bad checksum: client %x, server %x, client NID "
 666                                LPX64" (%s)\n", client_cksum, cksum,
 667                                req->rq_connection->c_peer.peer_nid, str);
 668                         cksum_counter = 1;
 669                         repbody->oa.o_nlink = cksum;
 670                 } else {
 671                         cksum_counter++;
 672                         if ((cksum_counter & (-cksum_counter)) == cksum_counter)
 673                                 CWARN("Checksum %u from "LPX64": %x OK\n",
 674                                       cksum_counter,
 675                                       req->rq_connection->c_peer.peer_nid,
 676                                       cksum);
 677                 }
 678         }
 679 #endif
 680         /* Must commit after prep above in all cases */
 681         rc2 = obd_commitrw(OBD_BRW_WRITE, req->rq_export, &repbody->oa,
 682                            objcount, ioo, npages, local_nb, oti);
 683
 684         if (rc == 0) {
 685                 /* set per-requested niobuf return codes */
 686                 for (i = j = 0; i < niocount; i++) {
 687                         int nob = remote_nb[i].len;
 688
 689                         rcs[i] = 0;
 690                         do {
 691                                 LASSERT(j < npages);
 692                                 if (local_nb[j].rc < 0)
 693                                         rcs[i] = local_nb[j].rc;
 694                                 nob -= pp_rnb[j].len;
 695                                 j++;
 696                         } while (nob > 0);
 697                         LASSERT(nob == 0);
 698                 }
 699                 LASSERT(j == npages);
 700         }
 701         if (rc == 0)
 702                 rc = rc2;
 703
 704  out_bulk:
 705         ptlrpc_free_bulk(desc);
 706  out_local:
 707         OBD_FREE(local_nb, sizeof(*local_nb) * npages);
 708  out_pp_rnb:
 709         free_per_page_niobufs(npages, pp_rnb, remote_nb);
 710  out:
 711         if (rc == 0) {
 712                 oti_to_request(oti, req);
 713                 rc = ptlrpc_reply(req);
 714         } else if (!comms_error) {
 715                 /* Only reply if there was no comms problem with bulk */
 716                 req->rq_status = rc;
 717                 ptlrpc_error(req);
 718         } else {
 719                 if (req->rq_repmsg != NULL) {
 720                         /* reply out callback would free */
 721                         OBD_FREE (req->rq_repmsg, req->rq_replen);
 722                 }
 723                 if (req->rq_reqmsg->conn_cnt == req->rq_export->exp_conn_cnt) {
 724                         CERROR("bulk IO comms error: "
 725                                "evicting %s@%s nid "LPX64" (%s)\n",
 726                                req->rq_export->exp_client_uuid.uuid,
 727                                req->rq_connection->c_remote_uuid.uuid,
 728                                req->rq_connection->c_peer.peer_nid,
 729                                portals_nid2str(req->rq_connection->c_peer.peer_ni->pni_number,
 730                                                req->rq_connection->c_peer.peer_nid,
 731                                                str));
 732                         ptlrpc_fail_export(req->rq_export);
 733                 } else {
 734                         CERROR("ignoring bulk IO comms error: "
 735                                "client reconnected %s@%s nid "LPX64" (%s)\n",
 736                                req->rq_export->exp_client_uuid.uuid,
 737                                req->rq_connection->c_remote_uuid.uuid,
 738                                req->rq_connection->c_peer.peer_nid,
 739                                portals_nid2str(req->rq_connection->c_peer.peer_ni->pni_number,
 740                                                req->rq_connection->c_peer.peer_nid,
 741                                                str));
 742                 }
 743         }
 744         RETURN(rc);
 745 }
 746
 747 static int ost_san_brw(struct ptlrpc_request *req, int cmd)
 748 {
 749         struct niobuf_remote *remote_nb, *res_nb, *pp_rnb;
 750         struct obd_ioobj *ioo;
 751         struct ost_body *body, *repbody;
 752         int rc, i, objcount, niocount, size[2] = {sizeof(*body)}, npages;
 753         int swab;
 754         ENTRY;
 755
 756         /* XXX not set to use latest protocol */
 757
 758         swab = lustre_msg_swabbed(req->rq_reqmsg);
 759         body = lustre_swab_reqbuf(req, 0, sizeof(*body), lustre_swab_ost_body);
 760         if (body == NULL) {
 761                 CERROR("Missing/short ost_body\n");
 762                 GOTO(out, rc = -EFAULT);
 763         }
 764
 765         ioo = lustre_swab_reqbuf(req, 1, sizeof(*ioo), lustre_swab_obd_ioobj);
 766         if (ioo == NULL) {
 767                 CERROR("Missing/short ioobj\n");
 768                 GOTO(out, rc = -EFAULT);
 769         }
 770         objcount = req->rq_reqmsg->buflens[1] / sizeof(*ioo);
 771         niocount = ioo[0].ioo_bufcnt;
 772         for (i = 1; i < objcount; i++) {
 773                 if (swab)
 774                         lustre_swab_obd_ioobj (&ioo[i]);
 775                 niocount += ioo[i].ioo_bufcnt;
 776         }
 777
 778         remote_nb = lustre_swab_reqbuf(req, 2, niocount * sizeof(*remote_nb),
 779                                        lustre_swab_niobuf_remote);
 780         if (remote_nb == NULL) {
 781                 CERROR("Missing/short niobuf\n");
 782                 GOTO(out, rc = -EFAULT);
 783         }
 784         if (swab) {                             /* swab the remaining niobufs */
 785                 for (i = 1; i < niocount; i++)
 786                         lustre_swab_niobuf_remote (&remote_nb[i]);
 787         }
 788
 789         /* CAVEAT EMPTOR this sets ioo->ioo_bufcnt to # pages */
 790         npages = get_per_page_niobufs(ioo, objcount,remote_nb,niocount,&pp_rnb);
 791         if (npages < 0)
 792                 GOTO (out, rc = npages);
 793
 794         size[1] = npages * sizeof(*pp_rnb);
 795         rc = lustre_pack_reply(req, 2, size, NULL);
 796         if (rc)
 797                 GOTO(out_pp_rnb, rc);
 798
 799         req->rq_status = obd_san_preprw(cmd, req->rq_export, &body->oa,
 800                                         objcount, ioo, npages, pp_rnb);
 801
 802         if (req->rq_status)
 803                 GOTO(out_pp_rnb, rc = 0);
 804
 805         repbody = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*repbody));
 806         memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
 807
 808         res_nb = lustre_msg_buf(req->rq_repmsg, 1, size[1]);
 809         memcpy(res_nb, remote_nb, size[1]);
 810         rc = 0;
 811 out_pp_rnb:
 812         free_per_page_niobufs(npages, pp_rnb, remote_nb);
 813 out:
 814         if (rc) {
 815                 OBD_FREE(req->rq_repmsg, req->rq_replen);
 816                 req->rq_repmsg = NULL;
 817                 req->rq_status = rc;
 818                 ptlrpc_error(req);
 819         } else
 820                 ptlrpc_reply(req);
 821
 822         return rc;
 823 }
 824
 825
 826 static int ost_set_info(struct obd_export *exp, struct ptlrpc_request *req)
 827 {
 828         char *key;
 829         int keylen, rc = 0;
 830         ENTRY;
 831
 832         key = lustre_msg_buf(req->rq_reqmsg, 0, 1);
 833         if (key == NULL) {
 834                 DEBUG_REQ(D_HA, req, "no set_info key");
 835                 RETURN(-EFAULT);
 836         }
 837         keylen = req->rq_reqmsg->buflens[0];
 838
 839         rc = lustre_pack_reply(req, 0, NULL, NULL);
 840         if (rc)
 841                 RETURN(rc);
 842
 843         rc = obd_set_info(exp, keylen, key, 0, NULL);
 844         req->rq_repmsg->status = 0;
 845         RETURN(rc);
 846 }
 847
 848 static int ost_get_info(struct obd_export *exp, struct ptlrpc_request *req)
 849 {
 850         char *key;
 851         int keylen, rc = 0, size = sizeof(obd_id);
 852         obd_id *reply;
 853         ENTRY;
 854
 855         key = lustre_msg_buf(req->rq_reqmsg, 0, 1);
 856         if (key == NULL) {
 857                 DEBUG_REQ(D_HA, req, "no get_info key");
 858                 RETURN(-EFAULT);
 859         }
 860         keylen = req->rq_reqmsg->buflens[0];
 861
 862         if (keylen < strlen("last_id") || memcmp(key, "last_id", 7) != 0)
 863                 RETURN(-EPROTO);
 864
 865         rc = lustre_pack_reply(req, 1, &size, NULL);
 866         if (rc)
 867                 RETURN(rc);
 868
 869         reply = lustre_msg_buf(req->rq_repmsg, 0, sizeof(*reply));
 870         rc = obd_get_info(exp, keylen, key, &size, reply);
 871         req->rq_repmsg->status = 0;
 872         RETURN(rc);
 873 }
 874
 875 static int ost_filter_recovery_request(struct ptlrpc_request *req,
 876                                        struct obd_device *obd, int *process)
 877 {
 878         switch (req->rq_reqmsg->opc) {
 879         case OST_CONNECT: /* This will never get here, but for completeness. */
 880         case OST_DISCONNECT:
 881                *process = 1;
 882                RETURN(0);
 883
 884         case OBD_PING:
 885         case OST_CREATE:
 886         case OST_DESTROY:
 887         case OST_PUNCH:
 888         case OST_SETATTR:
 889         case OST_SYNC:
 890         case OST_WRITE:
 891         case OBD_LOG_CANCEL:
 892         case LDLM_ENQUEUE:
 893                 *process = target_queue_recovery_request(req, obd);
 894                 RETURN(0);
 895
 896         default:
 897                 DEBUG_REQ(D_ERROR, req, "not permitted during recovery");
 898                 *process = 0;
 899                 /* XXX what should we set rq_status to here? */
 900                 req->rq_status = -EAGAIN;
 901                 RETURN(ptlrpc_error(req));
 902         }
 903 }
 904
 905
 906
 907 static int ost_handle(struct ptlrpc_request *req)
 908 {
 909         struct obd_trans_info trans_info = { 0, };
 910         struct obd_trans_info *oti = &trans_info;
 911         int should_process, fail = OBD_FAIL_OST_ALL_REPLY_NET, rc = 0;
 912         struct obd_export *exp = NULL;
 913         ENTRY;
 914
 915         LASSERT(current->journal_info == NULL);
 916         /* XXX identical to MDS */
 917         if (req->rq_reqmsg->opc != OST_CONNECT) {
 918                 struct obd_device *obd;
 919                 int abort_recovery, recovering;
 920
 921                 exp = req->rq_export;
 922
 923                 if (exp == NULL) {
 924                         CDEBUG(D_HA, "operation %d on unconnected OST\n",
 925                                req->rq_reqmsg->opc);
 926                         req->rq_status = -ENOTCONN;
 927                         GOTO(out, rc = -ENOTCONN);
 928                 }
 929
 930                 obd = exp->exp_obd;
 931
 932                 /* Check for aborted recovery. */
 933                 spin_lock_bh(&obd->obd_processing_task_lock);
 934                 abort_recovery = obd->obd_abort_recovery;
 935                 recovering = obd->obd_recovering;
 936                 spin_unlock_bh(&obd->obd_processing_task_lock);
 937                 if (abort_recovery) {
 938                         target_abort_recovery(obd);
 939                 } else if (recovering) {
 940                         rc = ost_filter_recovery_request(req, obd,
 941                                                          &should_process);
 942                         if (rc || !should_process)
 943                                 RETURN(rc);
 944                 }
 945         }
 946
 947         if (strcmp(req->rq_obd->obd_type->typ_name, "ost") != 0)
 948                 GOTO(out, rc = -EINVAL);
 949
 950         oti_init(oti, req);
 951
 952         switch (req->rq_reqmsg->opc) {
 953         case OST_CONNECT:
 954                 CDEBUG(D_INODE, "connect\n");
 955                 OBD_FAIL_RETURN(OBD_FAIL_OST_CONNECT_NET, 0);
 956                 rc = target_handle_connect(req, ost_handle);
 957                 break;
 958         case OST_DISCONNECT:
 959                 CDEBUG(D_INODE, "disconnect\n");
 960                 OBD_FAIL_RETURN(OBD_FAIL_OST_DISCONNECT_NET, 0);
 961                 rc = target_handle_disconnect(req);
 962                 break;
 963         case OST_CREATE:
 964                 CDEBUG(D_INODE, "create\n");
 965                 OBD_FAIL_RETURN(OBD_FAIL_OST_CREATE_NET, 0);
 966                 rc = ost_create(exp, req, oti);
 967                 break;
 968         case OST_DESTROY:
 969                 CDEBUG(D_INODE, "destroy\n");
 970                 OBD_FAIL_RETURN(OBD_FAIL_OST_DESTROY_NET, 0);
 971                 rc = ost_destroy(exp, req, oti);
 972                 break;
 973         case OST_GETATTR:
 974                 CDEBUG(D_INODE, "getattr\n");
 975                 OBD_FAIL_RETURN(OBD_FAIL_OST_GETATTR_NET, 0);
 976                 rc = ost_getattr(exp, req);
 977                 break;
 978         case OST_SETATTR:
 979                 CDEBUG(D_INODE, "setattr\n");
 980                 OBD_FAIL_RETURN(OBD_FAIL_OST_SETATTR_NET, 0);
 981                 rc = ost_setattr(exp, req, oti);
 982                 break;
 983         case OST_WRITE:
 984                 CDEBUG(D_INODE, "write\n");
 985                 OBD_FAIL_RETURN(OBD_FAIL_OST_BRW_NET, 0);
 986                 rc = ost_brw_write(req, oti);
 987                 LASSERT(current->journal_info == NULL);
 988                 /* ost_brw sends its own replies */
 989                 RETURN(rc);
 990         case OST_READ:
 991                 CDEBUG(D_INODE, "read\n");
 992                 OBD_FAIL_RETURN(OBD_FAIL_OST_BRW_NET, 0);
 993                 rc = ost_brw_read(req);
 994                 LASSERT(current->journal_info == NULL);
 995                 /* ost_brw sends its own replies */
 996                 RETURN(rc);
 997         case OST_SAN_READ:
 998                 CDEBUG(D_INODE, "san read\n");
 999                 OBD_FAIL_RETURN(OBD_FAIL_OST_BRW_NET, 0);
1000                 rc = ost_san_brw(req, OBD_BRW_READ);
1001                 /* ost_san_brw sends its own replies */
1002                 RETURN(rc);
1003         case OST_SAN_WRITE:
1004                 CDEBUG(D_INODE, "san write\n");
1005                 OBD_FAIL_RETURN(OBD_FAIL_OST_BRW_NET, 0);
1006                 rc = ost_san_brw(req, OBD_BRW_WRITE);
1007                 /* ost_san_brw sends its own replies */
1008                 RETURN(rc);
1009         case OST_PUNCH:
1010                 CDEBUG(D_INODE, "punch\n");
1011                 OBD_FAIL_RETURN(OBD_FAIL_OST_PUNCH_NET, 0);
1012                 rc = ost_punch(exp, req, oti);
1013                 break;
1014         case OST_STATFS:
1015                 CDEBUG(D_INODE, "statfs\n");
1016                 OBD_FAIL_RETURN(OBD_FAIL_OST_STATFS_NET, 0);
1017                 rc = ost_statfs(req);
1018                 break;
1019         case OST_SYNC:
1020                 CDEBUG(D_INODE, "sync\n");
1021                 OBD_FAIL_RETURN(OBD_FAIL_OST_SYNC_NET, 0);
1022                 rc = ost_sync(exp, req);
1023                 break;
1024         case OST_SET_INFO:
1025                 DEBUG_REQ(D_INODE, req, "set_info");
1026                 rc = ost_set_info(exp, req);
1027                 break;
1028         case OST_GET_INFO:
1029                 DEBUG_REQ(D_INODE, req, "get_info");
1030                 rc = ost_get_info(exp, req);
1031                 break;
1032         case OBD_PING:
1033                 DEBUG_REQ(D_INODE, req, "ping");
1034                 rc = target_handle_ping(req);
1035                 break;
1036 #ifdef ENABLE_ORPHANS
1037         /* FIXME - just reply status */
1038         case LLOG_ORIGIN_CONNECT:
1039                 DEBUG_REQ(D_INODE, req, "log connect\n");
1040                 rc = llog_handle_connect(req);
1041                 req->rq_status = rc;
1042                 rc = lustre_pack_reply(req, 0, NULL, NULL);
1043                 if (rc)
1044                         RETURN(rc);
1045                 RETURN(ptlrpc_reply(req));
1046                 //break;
1047         case OBD_LOG_CANCEL:
1048                 CDEBUG(D_INODE, "log cancel\n");
1049                 OBD_FAIL_RETURN(OBD_FAIL_OBD_LOG_CANCEL_NET, 0);
1050                 rc = llog_origin_handle_cancel(req);
1051                 req->rq_status = rc;
1052                 rc = lustre_pack_reply(req, 0, NULL, NULL);
1053                 if (rc)
1054                         RETURN(rc);
1055                 RETURN(ptlrpc_reply(req));
1056                 //break;
1057 #endif
1058         case LDLM_ENQUEUE:
1059                 CDEBUG(D_INODE, "enqueue\n");
1060                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_ENQUEUE, 0);
1061                 rc = ldlm_handle_enqueue(req, ldlm_server_completion_ast,
1062                                          ldlm_server_blocking_ast);
1063                 fail = OBD_FAIL_OST_LDLM_REPLY_NET;
1064                 break;
1065         case LDLM_CONVERT:
1066                 CDEBUG(D_INODE, "convert\n");
1067                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CONVERT, 0);
1068                 rc = ldlm_handle_convert(req);
1069                 break;
1070         case LDLM_CANCEL:
1071                 CDEBUG(D_INODE, "cancel\n");
1072                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CANCEL, 0);
1073                 rc = ldlm_handle_cancel(req);
1074                 break;
1075         case LDLM_BL_CALLBACK:
1076         case LDLM_CP_CALLBACK:
1077                 CDEBUG(D_INODE, "callback\n");
1078                 CERROR("callbacks should not happen on OST\n");
1079                 /* fall through */
1080         default:
1081                 CERROR("Unexpected opcode %d\n", req->rq_reqmsg->opc);
1082                 req->rq_status = -ENOTSUPP;
1083                 rc = ptlrpc_error(req);
1084                 RETURN(rc);
1085         }
1086
1087         LASSERT(current->journal_info == NULL);
1088
1089         EXIT;
1090         /* If we're DISCONNECTing, the export_data is already freed */
1091         if (!rc && req->rq_reqmsg->opc != OST_DISCONNECT) {
1092                 struct obd_device *obd  = req->rq_export->exp_obd;
1093                 if (!obd->obd_no_transno) {
1094                         req->rq_repmsg->last_committed =
1095                                 obd->obd_last_committed;
1096                 } else {
1097                         DEBUG_REQ(D_IOCTL, req,
1098                                   "not sending last_committed update");
1099                 }
1100                 CDEBUG(D_INFO, "last_committed "LPU64", xid "LPX64"\n",
1101                        obd->obd_last_committed, req->rq_xid);
1102         }
1103
1104 out:
1105         if (lustre_msg_get_flags(req->rq_reqmsg) & MSG_LAST_REPLAY) {
1106                 struct obd_device *obd = req->rq_export->exp_obd;
1107
1108                 if (obd && obd->obd_recovering) {
1109                         DEBUG_REQ(D_HA, req, "LAST_REPLAY, queuing reply");
1110                         return target_queue_final_reply(req, rc);
1111                 }
1112                 /* Lost a race with recovery; let the error path DTRT. */
1113                 rc = req->rq_status = -ENOTCONN;
1114         }
1115
1116         if (!rc)
1117                 oti_to_request(oti, req);
1118
1119         target_send_reply(req, rc, fail);
1120         return 0;
1121 }
1122
1123 static int ost_setup(struct obd_device *obddev, obd_count len, void *buf)
1124 {
1125         struct ost_obd *ost = &obddev->u.ost;
1126         int rc;
1127         ENTRY;
1128
1129         /* Get rid of unneeded supplementary groups */
1130         current->ngroups = 0;
1131         memset(current->groups, 0, sizeof(current->groups));
1132
1133         rc = llog_start_commit_thread();
1134         if (rc < 0)
1135                 RETURN(rc);
1136
1137         ost->ost_service = ptlrpc_init_svc(OST_NEVENTS, OST_NBUFS,
1138                                            OST_BUFSIZE, OST_MAXREQSIZE,
1139                                            OST_REQUEST_PORTAL, OSC_REPLY_PORTAL,
1140                                            ost_handle, "ost",
1141                                            obddev->obd_proc_entry);
1142         if (ost->ost_service == NULL) {
1143                 CERROR("failed to start service\n");
1144                 RETURN(-ENOMEM);
1145         }
1146
1147         rc = ptlrpc_start_n_threads(obddev, ost->ost_service, OST_NUM_THREADS,
1148                                  "ll_ost");
1149         if (rc)
1150                 GOTO(out, rc = -EINVAL);
1151
1152         ost->ost_create_service =
1153                 ptlrpc_init_svc(OST_NEVENTS, OST_NBUFS, OST_BUFSIZE,
1154                                 OST_MAXREQSIZE, OST_CREATE_PORTAL,
1155                                 OSC_REPLY_PORTAL, ost_handle, "ost_create",
1156                                 obddev->obd_proc_entry);
1157         if (ost->ost_create_service == NULL) {
1158                 CERROR("failed to start OST create service\n");
1159                 GOTO(out, rc = -ENOMEM);
1160         }
1161
1162         rc = ptlrpc_start_n_threads(obddev, ost->ost_create_service, 1,
1163                                     "ll_ost_create");
1164         if (rc)
1165                 GOTO(out_create, rc = -EINVAL);
1166
1167         RETURN(0);
1168
1169 out_create:
1170         ptlrpc_unregister_service(ost->ost_create_service);
1171 out:
1172         ptlrpc_unregister_service(ost->ost_service);
1173         RETURN(rc);
1174 }
1175
1176 static int ost_cleanup(struct obd_device *obddev, int flags)
1177 {
1178         struct ost_obd *ost = &obddev->u.ost;
1179         int err = 0;
1180         ENTRY;
1181
1182         if (obddev->obd_recovering)
1183                 target_cancel_recovery_timer(obddev);
1184
1185         ptlrpc_stop_all_threads(ost->ost_service);
1186         ptlrpc_unregister_service(ost->ost_service);
1187
1188         ptlrpc_stop_all_threads(ost->ost_create_service);
1189         ptlrpc_unregister_service(ost->ost_create_service);
1190
1191         RETURN(err);
1192 }
1193
1194 int ost_attach(struct obd_device *dev, obd_count len, void *data)
1195 {
1196         struct lprocfs_static_vars lvars;
1197
1198         lprocfs_init_vars(ost,&lvars);
1199         return lprocfs_obd_attach(dev, lvars.obd_vars);
1200 }
1201
1202 int ost_detach(struct obd_device *dev)
1203 {
1204         return lprocfs_obd_detach(dev);
1205 }
1206
1207 /* use obd ops to offer management infrastructure */
1208 static struct obd_ops ost_obd_ops = {
1209         o_owner:        THIS_MODULE,
1210         o_attach:       ost_attach,
1211         o_detach:       ost_detach,
1212         o_setup:        ost_setup,
1213         o_cleanup:      ost_cleanup,
1214 };
1215
1216 static int __init ost_init(void)
1217 {
1218         struct lprocfs_static_vars lvars;
1219         ENTRY;
1220
1221         lprocfs_init_vars(ost,&lvars);
1222         RETURN(class_register_type(&ost_obd_ops, lvars.module_vars,
1223                                    LUSTRE_OST_NAME));
1224 }
1225
1226 static void /*__exit*/ ost_exit(void)
1227 {
1228         class_unregister_type(LUSTRE_OST_NAME);
1229 }
1230
1231 MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
1232 MODULE_DESCRIPTION("Lustre Object Storage Target (OST) v0.01");
1233 MODULE_LICENSE("GPL");
1234
1235 module_init(ost_init);
1236 module_exit(ost_exit);