lustre/obdecho/echo.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.gnu.org/licenses/gpl-2.0.html
  19  *
  20  * GPL HEADER END
  21  */
  22 /*
  23  * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Use is subject to license terms.
  25  *
  26  * Copyright (c) 2010, 2017, Intel Corporation.
  27  */
  28 /*
  29  * This file is part of Lustre, http://www.lustre.org/
  30  * Lustre is a trademark of Sun Microsystems, Inc.
  31  *
  32  * lustre/obdecho/echo.c
  33  *
  34  * Author: Peter Braam <braam@clusterfs.com>
  35  * Author: Andreas Dilger <adilger@clusterfs.com>
  36  */
  37
  38 #define DEBUG_SUBSYSTEM S_ECHO
  39
  40 #include <obd_support.h>
  41 #include <obd_class.h>
  42 #include <lustre_debug.h>
  43 #include <lustre_dlm.h>
  44 #include <lprocfs_status.h>
  45
  46 #include "echo_internal.h"
  47
  48 /* The echo objid needs to be below 2^32, because regular FID numbers are
  49  * limited to 2^32 objects in f_oid for the FID_SEQ_ECHO range. b=23335 */
  50 #define ECHO_INIT_OID        0x10000000ULL
  51 #define ECHO_HANDLE_MAGIC    0xabcd0123fedc9876ULL
  52
  53 #define ECHO_PERSISTENT_PAGES (ECHO_PERSISTENT_SIZE >> PAGE_SHIFT)
  54 static struct page *echo_persistent_pages[ECHO_PERSISTENT_PAGES];
  55
  56 enum {
  57         LPROC_ECHO_READ_BYTES = 1,
  58         LPROC_ECHO_WRITE_BYTES = 2,
  59         LPROC_ECHO_LAST = LPROC_ECHO_WRITE_BYTES +1
  60 };
  61
  62 struct echo_srv_device {
  63         struct lu_device esd_dev;
  64         struct lu_target esd_lut;
  65 };
  66
  67 static inline struct echo_srv_device *echo_srv_dev(struct lu_device *d)
  68 {
  69         return container_of0(d, struct echo_srv_device, esd_dev);
  70 }
  71
  72 static inline struct obd_device *echo_srv_obd(struct echo_srv_device *esd)
  73 {
  74         return esd->esd_dev.ld_obd;
  75 }
  76
  77 static int echo_connect(const struct lu_env *env,
  78                         struct obd_export **exp, struct obd_device *obd,
  79                         struct obd_uuid *cluuid, struct obd_connect_data *data,
  80                         void *localdata)
  81 {
  82         struct lustre_handle conn = { 0 };
  83         int rc;
  84
  85         data->ocd_connect_flags &= ECHO_CONNECT_SUPPORTED;
  86
  87         if (data->ocd_connect_flags & OBD_CONNECT_FLAGS2)
  88                 data->ocd_connect_flags2 &= ECHO_CONNECT_SUPPORTED2;
  89
  90         rc = class_connect(&conn, obd, cluuid);
  91         if (rc) {
  92                 CERROR("can't connect %d\n", rc);
  93                 return rc;
  94         }
  95         *exp = class_conn2export(&conn);
  96
  97         return 0;
  98 }
  99
 100 static int echo_disconnect(struct obd_export *exp)
 101 {
 102         LASSERT (exp != NULL);
 103
 104         return server_disconnect_export(exp);
 105 }
 106
 107 static int echo_init_export(struct obd_export *exp)
 108 {
 109         return ldlm_init_export(exp);
 110 }
 111
 112 static int echo_destroy_export(struct obd_export *exp)
 113 {
 114         ENTRY;
 115
 116         target_destroy_export(exp);
 117         ldlm_destroy_export(exp);
 118
 119         RETURN(0);
 120 }
 121
 122 static u64 echo_next_id(struct obd_device *obddev)
 123 {
 124         u64 id;
 125
 126         spin_lock(&obddev->u.echo.eo_lock);
 127         id = ++obddev->u.echo.eo_lastino;
 128         spin_unlock(&obddev->u.echo.eo_lock);
 129
 130         return id;
 131 }
 132
 133 static void
 134 echo_page_debug_setup(struct page *page, int rw, u64 id,
 135                       __u64 offset, int len)
 136 {
 137         int   page_offset = offset & ~PAGE_MASK;
 138         char *addr        = ((char *)kmap(page)) + page_offset;
 139
 140         if (len % OBD_ECHO_BLOCK_SIZE != 0)
 141                 CERROR("Unexpected block size %d\n", len);
 142
 143         while (len > 0) {
 144                 if (rw & OBD_BRW_READ)
 145                         block_debug_setup(addr, OBD_ECHO_BLOCK_SIZE,
 146                                           offset, id);
 147                 else
 148                         block_debug_setup(addr, OBD_ECHO_BLOCK_SIZE,
 149                                           0xecc0ecc0ecc0ecc0ULL,
 150                                           0xecc0ecc0ecc0ecc0ULL);
 151
 152                 addr   += OBD_ECHO_BLOCK_SIZE;
 153                 offset += OBD_ECHO_BLOCK_SIZE;
 154                 len    -= OBD_ECHO_BLOCK_SIZE;
 155         }
 156
 157         kunmap(page);
 158 }
 159
 160 static int
 161 echo_page_debug_check(struct page *page, u64 id,
 162                       __u64 offset, int len)
 163 {
 164         int   page_offset = offset & ~PAGE_MASK;
 165         char *addr        = ((char *)kmap(page)) + page_offset;
 166         int   rc          = 0;
 167         int   rc2;
 168
 169         if (len % OBD_ECHO_BLOCK_SIZE != 0)
 170                 CERROR("Unexpected block size %d\n", len);
 171
 172         while (len > 0) {
 173                 rc2 = block_debug_check("echo", addr, OBD_ECHO_BLOCK_SIZE,
 174                                         offset, id);
 175
 176                 if (rc2 != 0 && rc == 0)
 177                         rc = rc2;
 178
 179                 addr   += OBD_ECHO_BLOCK_SIZE;
 180                 offset += OBD_ECHO_BLOCK_SIZE;
 181                 len    -= OBD_ECHO_BLOCK_SIZE;
 182         }
 183
 184         kunmap(page);
 185
 186         return rc;
 187 }
 188
 189 static int echo_map_nb_to_lb(struct obdo *oa, struct obd_ioobj *obj,
 190                              struct niobuf_remote *nb, int *pages,
 191                              struct niobuf_local *lb, int cmd, int *left)
 192 {
 193         gfp_t gfp_mask = (ostid_id(&obj->ioo_oid) & 1) ?
 194                         GFP_HIGHUSER : GFP_KERNEL;
 195         int ispersistent = ostid_id(&obj->ioo_oid) == ECHO_PERSISTENT_OBJID;
 196         int debug_setup = (!ispersistent &&
 197                            (oa->o_valid & OBD_MD_FLFLAGS) != 0 &&
 198                            (oa->o_flags & OBD_FL_DEBUG_CHECK) != 0);
 199         struct niobuf_local *res = lb;
 200         u64 offset = nb->rnb_offset;
 201         int len = nb->rnb_len;
 202
 203         while (len > 0) {
 204                 int plen = PAGE_SIZE - (offset & (PAGE_SIZE-1));
 205                 if (len < plen)
 206                         plen = len;
 207
 208                 /* check for local buf overflow */
 209                 if (*left == 0)
 210                         return -EINVAL;
 211
 212                 res->lnb_file_offset = offset;
 213                 res->lnb_len = plen;
 214                 LASSERT((res->lnb_file_offset & ~PAGE_MASK) +
 215                         res->lnb_len <= PAGE_SIZE);
 216
 217                 if (ispersistent &&
 218                     ((res->lnb_file_offset >> PAGE_SHIFT) <
 219                       ECHO_PERSISTENT_PAGES)) {
 220                         res->lnb_page =
 221                                 echo_persistent_pages[res->lnb_file_offset >>
 222                                                       PAGE_SHIFT];
 223                         /* Take extra ref so __free_pages() can be called OK */
 224                         get_page(res->lnb_page);
 225                 } else {
 226                         res->lnb_page = alloc_page(gfp_mask);
 227                         if (res->lnb_page == NULL) {
 228                                 CERROR("can't get page for id " DOSTID"\n",
 229                                        POSTID(&obj->ioo_oid));
 230                                 return -ENOMEM;
 231                         }
 232                 }
 233
 234                 CDEBUG(D_PAGE, "$$$$ get page %p @ %llu for %d\n",
 235                        res->lnb_page, res->lnb_file_offset, res->lnb_len);
 236
 237                 if (cmd & OBD_BRW_READ)
 238                         res->lnb_rc = res->lnb_len;
 239
 240                 if (debug_setup)
 241                         echo_page_debug_setup(res->lnb_page, cmd,
 242                                               ostid_id(&obj->ioo_oid),
 243                                               res->lnb_file_offset,
 244                                               res->lnb_len);
 245
 246                 offset += plen;
 247                 len -= plen;
 248                 res++;
 249
 250                 (*left)--;
 251                 (*pages)++;
 252         }
 253
 254         return 0;
 255 }
 256
 257 static int echo_finalize_lb(struct obdo *oa, struct obd_ioobj *obj,
 258                             struct niobuf_remote *rb, int *pgs,
 259                             struct niobuf_local *lb, int verify)
 260 {
 261         struct niobuf_local *res = lb;
 262         u64 start = rb->rnb_offset >> PAGE_SHIFT;
 263         u64 end   = (rb->rnb_offset + rb->rnb_len + PAGE_SIZE - 1) >>
 264                     PAGE_SHIFT;
 265         int     count  = (int)(end - start);
 266         int     rc     = 0;
 267         int     i;
 268
 269         for (i = 0; i < count; i++, (*pgs) ++, res++) {
 270                 struct page *page = res->lnb_page;
 271                 void       *addr;
 272
 273                 if (page == NULL) {
 274                         CERROR("null page objid %llu:%p, buf %d/%d\n",
 275                                ostid_id(&obj->ioo_oid), page, i,
 276                                obj->ioo_bufcnt);
 277                         return -EFAULT;
 278                 }
 279
 280                 addr = kmap(page);
 281
 282                 CDEBUG(D_PAGE, "$$$$ use page %p, addr %p@%llu\n",
 283                        res->lnb_page, addr, res->lnb_file_offset);
 284
 285                 if (verify) {
 286                         int vrc = echo_page_debug_check(page,
 287                                                         ostid_id(&obj->ioo_oid),
 288                                                         res->lnb_file_offset,
 289                                                         res->lnb_len);
 290                         /* check all the pages always */
 291                         if (vrc != 0 && rc == 0)
 292                                 rc = vrc;
 293                 }
 294
 295                 kunmap(page);
 296                 /* NB see comment above regarding persistent pages */
 297                 __free_page(page);
 298         }
 299
 300         return rc;
 301 }
 302
 303 static int echo_preprw(const struct lu_env *env, int cmd,
 304                        struct obd_export *export, struct obdo *oa,
 305                        int objcount, struct obd_ioobj *obj,
 306                        struct niobuf_remote *nb, int *pages,
 307                        struct niobuf_local *res)
 308 {
 309         struct obd_device *obd;
 310         int tot_bytes = 0;
 311         int rc = 0;
 312         int i, left;
 313         ENTRY;
 314
 315         obd = export->exp_obd;
 316         if (obd == NULL)
 317                 RETURN(-EINVAL);
 318
 319         /* Temp fix to stop falling foul of osc_announce_cached() */
 320         oa->o_valid &= ~(OBD_MD_FLBLOCKS | OBD_MD_FLGRANT);
 321
 322         memset(res, 0, sizeof(*res) * *pages);
 323
 324         CDEBUG(D_PAGE, "%s %d obdos with %d IOs\n",
 325                cmd == OBD_BRW_READ ? "reading" : "writing", objcount, *pages);
 326
 327         left = *pages;
 328         *pages = 0;
 329
 330         for (i = 0; i < objcount; i++, obj++) {
 331                 int j;
 332
 333                 for (j = 0 ; j < obj->ioo_bufcnt ; j++, nb++) {
 334
 335                         rc = echo_map_nb_to_lb(oa, obj, nb, pages,
 336                                                res + *pages, cmd, &left);
 337                         if (rc)
 338                                 GOTO(preprw_cleanup, rc);
 339
 340                         tot_bytes += nb->rnb_len;
 341                 }
 342         }
 343
 344         atomic_add(*pages, &obd->u.echo.eo_prep);
 345
 346         if (cmd & OBD_BRW_READ)
 347                 lprocfs_counter_add(obd->obd_stats, LPROC_ECHO_READ_BYTES,
 348                                     tot_bytes);
 349         else
 350                 lprocfs_counter_add(obd->obd_stats, LPROC_ECHO_WRITE_BYTES,
 351                                     tot_bytes);
 352
 353         CDEBUG(D_PAGE, "%d pages allocated after prep\n",
 354                atomic_read(&obd->u.echo.eo_prep));
 355
 356         RETURN(0);
 357
 358 preprw_cleanup:
 359         /* It is possible that we would rather handle errors by  allow
 360          * any already-set-up pages to complete, rather than tearing them
 361          * all down again.  I believe that this is what the in-kernel
 362          * prep/commit operations do.
 363          */
 364         CERROR("cleaning up %u pages (%d obdos)\n", *pages, objcount);
 365         for (i = 0; i < *pages; i++) {
 366                 kunmap(res[i].lnb_page);
 367                 /* NB if this is a persistent page, __free_page() will just
 368                  * lose the extra ref gained above */
 369                 __free_page(res[i].lnb_page);
 370                 res[i].lnb_page = NULL;
 371                 atomic_dec(&obd->u.echo.eo_prep);
 372         }
 373
 374         return rc;
 375 }
 376
 377 static int echo_commitrw(const struct lu_env *env, int cmd,
 378                          struct obd_export *export, struct obdo *oa,
 379                          int objcount, struct obd_ioobj *obj,
 380                          struct niobuf_remote *rb, int niocount,
 381                          struct niobuf_local *res, int rc)
 382 {
 383         struct obd_device *obd;
 384         int pgs = 0;
 385         int i;
 386         ENTRY;
 387
 388         obd = export->exp_obd;
 389         if (obd == NULL)
 390                 RETURN(-EINVAL);
 391
 392         if (rc)
 393                 GOTO(commitrw_cleanup, rc);
 394
 395         if ((cmd & OBD_BRW_RWMASK) == OBD_BRW_READ) {
 396                 CDEBUG(D_PAGE, "reading %d obdos with %d IOs\n",
 397                        objcount, niocount);
 398         } else {
 399                 CDEBUG(D_PAGE, "writing %d obdos with %d IOs\n",
 400                        objcount, niocount);
 401         }
 402
 403         if (niocount && res == NULL) {
 404                 CERROR("NULL res niobuf with niocount %d\n", niocount);
 405                 RETURN(-EINVAL);
 406         }
 407
 408         for (i = 0; i < objcount; i++, obj++) {
 409                 int verify = (rc == 0 &&
 410                              ostid_id(&obj->ioo_oid) != ECHO_PERSISTENT_OBJID &&
 411                               (oa->o_valid & OBD_MD_FLFLAGS) != 0 &&
 412                               (oa->o_flags & OBD_FL_DEBUG_CHECK) != 0);
 413                 int j;
 414
 415                 for (j = 0 ; j < obj->ioo_bufcnt ; j++, rb++) {
 416                         int vrc = echo_finalize_lb(oa, obj, rb, &pgs, &res[pgs],
 417                                                    verify);
 418                         if (vrc == 0)
 419                                 continue;
 420
 421                         if (vrc == -EFAULT)
 422                                 GOTO(commitrw_cleanup, rc = vrc);
 423
 424                         if (rc == 0)
 425                                 rc = vrc;
 426                 }
 427
 428         }
 429
 430         atomic_sub(pgs, &obd->u.echo.eo_prep);
 431
 432         CDEBUG(D_PAGE, "%d pages remain after commit\n",
 433                atomic_read(&obd->u.echo.eo_prep));
 434         RETURN(rc);
 435
 436 commitrw_cleanup:
 437         atomic_sub(pgs, &obd->u.echo.eo_prep);
 438
 439         CERROR("cleaning up %d pages (%d obdos)\n",
 440                niocount - pgs - 1, objcount);
 441
 442         while (pgs < niocount) {
 443                 struct page *page = res[pgs++].lnb_page;
 444
 445                 if (page == NULL)
 446                         continue;
 447
 448                 /* NB see comment above regarding persistent pages */
 449                 __free_page(page);
 450                 atomic_dec(&obd->u.echo.eo_prep);
 451         }
 452         return rc;
 453 }
 454
 455 LPROC_SEQ_FOPS_RO_TYPE(echo, uuid);
 456 static struct lprocfs_vars lprocfs_echo_obd_vars[] = {
 457         { .name =       "uuid",
 458           .fops =       &echo_uuid_fops         },
 459         { NULL }
 460 };
 461
 462 struct obd_ops echo_obd_ops = {
 463         .o_owner           = THIS_MODULE,
 464         .o_connect         = echo_connect,
 465         .o_disconnect      = echo_disconnect,
 466         .o_init_export     = echo_init_export,
 467         .o_destroy_export  = echo_destroy_export,
 468         .o_preprw          = echo_preprw,
 469         .o_commitrw        = echo_commitrw,
 470 };
 471
 472 /**
 473  * Echo Server request handler for OST_CREATE RPC.
 474  *
 475  * This is part of request processing. Its simulates the object
 476  * creation on OST.
 477  *
 478  * \param[in] tsi       target session environment for this request
 479  *
 480  * \retval              0 if successful
 481  * \retval              negative value on error
 482  */
 483 static int esd_create_hdl(struct tgt_session_info *tsi)
 484 {
 485         const struct obdo *oa = &tsi->tsi_ost_body->oa;
 486         struct obd_device *obd = tsi->tsi_exp->exp_obd;
 487         struct ost_body *repbody;
 488         struct obdo *rep_oa;
 489
 490         ENTRY;
 491
 492         repbody = req_capsule_server_get(tsi->tsi_pill, &RMF_OST_BODY);
 493         if (repbody == NULL)
 494                 RETURN(-ENOMEM);
 495
 496         if (!(oa->o_mode & S_IFMT)) {
 497                 CERROR("%s: no type is set in obdo!\n",
 498                        tsi->tsi_exp->exp_obd->obd_name);
 499                 RETURN(-ENOENT);
 500         }
 501
 502         if (!(oa->o_valid & OBD_MD_FLTYPE)) {
 503                 CERROR("%s: invalid o_valid in obdo: %#llx\n",
 504                        tsi->tsi_exp->exp_obd->obd_name, oa->o_valid);
 505                 RETURN(-EINVAL);
 506         }
 507
 508         rep_oa = &repbody->oa;
 509
 510         if (!fid_seq_is_echo(ostid_seq(&oa->o_oi))) {
 511                 CERROR("%s: invalid seq %#llx\n",
 512                        tsi->tsi_exp->exp_obd->obd_name, ostid_seq(&oa->o_oi));
 513                 return -EINVAL;
 514         }
 515
 516         ostid_set_seq_echo(&rep_oa->o_oi);
 517         ostid_set_id(&rep_oa->o_oi, echo_next_id(obd));
 518
 519         CDEBUG(D_INFO, "%s: Create object "DOSTID"\n",
 520                tsi->tsi_exp->exp_obd->obd_name, POSTID(&rep_oa->o_oi));
 521
 522         rep_oa->o_valid |= OBD_MD_FLID | OBD_MD_FLGROUP;
 523
 524         RETURN(0);
 525 }
 526
 527 /**
 528  * Echo Server request handler for OST_DESTROY RPC.
 529  *
 530  * This is Echo Server part of request handling. It simulates the objects
 531  * destroy on OST.
 532  *
 533  * \param[in] tsi       target session environment for this request
 534  *
 535  * \retval              0 if successful
 536  * \retval              negative value on error
 537  */
 538 static int esd_destroy_hdl(struct tgt_session_info *tsi)
 539 {
 540         const struct obdo *oa = &tsi->tsi_ost_body->oa;
 541         struct obd_device *obd = tsi->tsi_exp->exp_obd;
 542         struct ost_body *repbody;
 543         u64 oid;
 544
 545         ENTRY;
 546
 547         oid = ostid_id(&oa->o_oi);
 548         LASSERT(oid != 0);
 549
 550         if (!(oa->o_valid & OBD_MD_FLID)) {
 551                 CERROR("%s: obdo missing FLID valid flag: %#llx\n",
 552                        tsi->tsi_exp->exp_obd->obd_name, oa->o_valid);
 553                 RETURN(-EINVAL);
 554         }
 555
 556         repbody = req_capsule_server_get(tsi->tsi_pill, &RMF_OST_BODY);
 557
 558         if (ostid_id(&oa->o_oi) > obd->u.echo.eo_lastino ||
 559             ostid_id(&oa->o_oi) < ECHO_INIT_OID) {
 560                 CERROR("%s: bad objid to destroy: "DOSTID"\n",
 561                        tsi->tsi_exp->exp_obd->obd_name, POSTID(&oa->o_oi));
 562                 RETURN(-EINVAL);
 563         }
 564
 565         CDEBUG(D_INFO, "%s: Destroy object "DOSTID"\n",
 566                tsi->tsi_exp->exp_obd->obd_name, POSTID(&oa->o_oi));
 567
 568         repbody->oa.o_oi = oa->o_oi;
 569         RETURN(0);
 570 }
 571
 572 /**
 573  * Echo Server request handler for OST_GETATTR RPC.
 574  *
 575  * This is Echo Server part of request handling. It returns an object
 576  * attributes to the client. All objects have the same attributes in
 577  * Echo Server.
 578  *
 579  * \param[in] tsi       target session environment for this request
 580  *
 581  * \retval              0 if successful
 582  * \retval              negative value on error
 583  */
 584 static int esd_getattr_hdl(struct tgt_session_info *tsi)
 585 {
 586         const struct obdo *oa = &tsi->tsi_ost_body->oa;
 587         struct obd_device *obd = tsi->tsi_exp->exp_obd;
 588         struct ost_body *repbody;
 589
 590         ENTRY;
 591
 592         if (!(oa->o_valid & OBD_MD_FLID)) {
 593                 CERROR("%s: obdo missing FLID valid flag: %#llx\n",
 594                        tsi->tsi_exp->exp_obd->obd_name, oa->o_valid);
 595                 RETURN(-EINVAL);
 596         }
 597
 598         repbody = req_capsule_server_get(tsi->tsi_pill, &RMF_OST_BODY);
 599         if (repbody == NULL)
 600                 RETURN(-ENOMEM);
 601
 602         repbody->oa.o_oi = oa->o_oi;
 603         repbody->oa.o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
 604
 605         obdo_cpy_md(&repbody->oa, &obd->u.echo.eo_oa, oa->o_valid);
 606
 607         repbody->oa.o_valid |= OBD_MD_FLFLAGS;
 608         repbody->oa.o_flags = OBD_FL_FLUSH;
 609
 610         RETURN(0);
 611 }
 612
 613 /**
 614  * Echo Server request handler for OST_SETATTR RPC.
 615  *
 616  * This is Echo Server part of request handling. It sets common
 617  * attributes from request to the Echo Server objects.
 618  *
 619  * \param[in] tsi       target session environment for this request
 620  *
 621  * \retval              0 if successful
 622  * \retval              negative value on error
 623  */
 624 static int esd_setattr_hdl(struct tgt_session_info *tsi)
 625 {
 626         struct ost_body *body = tsi->tsi_ost_body;
 627         struct obd_device *obd = tsi->tsi_exp->exp_obd;
 628         struct ost_body *repbody;
 629
 630         ENTRY;
 631
 632         if (!(body->oa.o_valid & OBD_MD_FLID)) {
 633                 CERROR("%s: obdo missing FLID valid flag: %#llx\n",
 634                        tsi->tsi_exp->exp_obd->obd_name,
 635                        body->oa.o_valid);
 636                 RETURN(-EINVAL);
 637         }
 638
 639         repbody = req_capsule_server_get(tsi->tsi_pill, &RMF_OST_BODY);
 640         if (repbody == NULL)
 641                 RETURN(-ENOMEM);
 642
 643         repbody->oa.o_oi = body->oa.o_oi;
 644         repbody->oa.o_valid = OBD_MD_FLID | OBD_MD_FLGROUP;
 645
 646         obd->u.echo.eo_oa = body->oa;
 647
 648         RETURN(0);
 649 }
 650
 651 #define OBD_FAIL_OST_READ_NET   OBD_FAIL_OST_BRW_NET
 652 #define OBD_FAIL_OST_WRITE_NET  OBD_FAIL_OST_BRW_NET
 653 #define OST_BRW_READ    OST_READ
 654 #define OST_BRW_WRITE   OST_WRITE
 655
 656 /**
 657  * Table of Echo Server specific request handlers
 658  *
 659  * This table contains all opcodes accepted by Echo Server and
 660  * specifies handlers for them. The tgt_request_handler()
 661  * uses such table from each target to process incoming
 662  * requests.
 663  */
 664 static struct tgt_handler esd_tgt_handlers[] = {
 665 TGT_RPC_HANDLER(OST_FIRST_OPC, 0, OST_CONNECT, tgt_connect,
 666                 &RQF_CONNECT, LUSTRE_OBD_VERSION),
 667 TGT_RPC_HANDLER(OST_FIRST_OPC, 0, OST_DISCONNECT, tgt_disconnect,
 668                 &RQF_OST_DISCONNECT, LUSTRE_OBD_VERSION),
 669 TGT_OST_HDL(HABEO_CORPUS | HABEO_REFERO, OST_GETATTR, esd_getattr_hdl),
 670 TGT_OST_HDL(HABEO_CORPUS | HABEO_REFERO | MUTABOR, OST_SETATTR,
 671             esd_setattr_hdl),
 672 TGT_OST_HDL(HABEO_REFERO | MUTABOR, OST_CREATE, esd_create_hdl),
 673 TGT_OST_HDL(HABEO_REFERO | MUTABOR, OST_DESTROY, esd_destroy_hdl),
 674 TGT_OST_HDL(HABEO_CORPUS | HABEO_REFERO, OST_BRW_READ, tgt_brw_read),
 675 TGT_OST_HDL(HABEO_CORPUS | MUTABOR, OST_BRW_WRITE, tgt_brw_write),
 676 };
 677
 678 static struct tgt_opc_slice esd_common_slice[] = {
 679         {
 680                 .tos_opc_start  = OST_FIRST_OPC,
 681                 .tos_opc_end    = OST_LAST_OPC,
 682                 .tos_hs         = esd_tgt_handlers
 683         },
 684         {
 685                 .tos_opc_start  = OBD_FIRST_OPC,
 686                 .tos_opc_end    = OBD_LAST_OPC,
 687                 .tos_hs         = tgt_obd_handlers
 688         },
 689         {
 690                 .tos_opc_start  = LDLM_FIRST_OPC,
 691                 .tos_opc_end    = LDLM_LAST_OPC,
 692                 .tos_hs         = tgt_dlm_handlers
 693         },
 694         {
 695                 .tos_opc_start  = SEC_FIRST_OPC,
 696                 .tos_opc_end    = SEC_LAST_OPC,
 697                 .tos_hs         = tgt_sec_ctx_handlers
 698         },
 699         {
 700                 .tos_hs         = NULL
 701         }
 702 };
 703
 704 /**
 705  * lu_device_operations matrix for ECHO SRV device is NULL,
 706  * this device is just serving incoming requests immediately
 707  * without building a stack of lu_devices.
 708  */
 709 static struct lu_device_operations echo_srv_lu_ops = { 0 };
 710
 711 /**
 712  * Initialize Echo Server device with parameters in the config log \a cfg.
 713  *
 714  * This is the main starting point of Echo Server initialization. It fills all
 715  * parameters with their initial values and starts Echo Server.
 716  *
 717  * \param[in] env       execution environment
 718  * \param[in] m         Echo Server device
 719  * \param[in] ldt       LU device type of Echo Server
 720  * \param[in] cfg       configuration log
 721  *
 722  * \retval              0 if successful
 723  * \retval              negative value on error
 724  */
 725 static int echo_srv_init0(const struct lu_env *env,
 726                           struct echo_srv_device *esd,
 727                           struct lu_device_type *ldt, struct lustre_cfg *cfg)
 728 {
 729         const char *dev = lustre_cfg_string(cfg, 0);
 730         struct obd_device *obd;
 731         char ns_name[48];
 732         int rc;
 733
 734         ENTRY;
 735
 736         obd = class_name2obd(dev);
 737         if (obd == NULL) {
 738                 CERROR("Cannot find obd with name %s\n", dev);
 739                 RETURN(-ENODEV);
 740         }
 741
 742         spin_lock_init(&obd->u.echo.eo_lock);
 743         obd->u.echo.eo_lastino = ECHO_INIT_OID;
 744
 745         esd->esd_dev.ld_ops = &echo_srv_lu_ops;
 746         esd->esd_dev.ld_obd = obd;
 747         /* set this lu_device to obd, because error handling need it */
 748         obd->obd_lu_dev = &esd->esd_dev;
 749
 750         /* No connection accepted until configurations will finish */
 751         spin_lock(&obd->obd_dev_lock);
 752         obd->obd_no_conn = 1;
 753         spin_unlock(&obd->obd_dev_lock);
 754
 755         /* non-replayable target */
 756         obd->obd_replayable = 0;
 757
 758         snprintf(ns_name, sizeof(ns_name), "echotgt-%s", obd->obd_uuid.uuid);
 759         obd->obd_namespace = ldlm_namespace_new(obd, ns_name,
 760                                                 LDLM_NAMESPACE_SERVER,
 761                                                 LDLM_NAMESPACE_MODEST,
 762                                                 LDLM_NS_TYPE_OST);
 763         if (obd->obd_namespace == NULL)
 764                 RETURN(-ENOMEM);
 765
 766         obd->obd_vars = lprocfs_echo_obd_vars;
 767         if (!lprocfs_obd_setup(obd, true) &&
 768             lprocfs_alloc_obd_stats(obd, LPROC_ECHO_LAST) == 0) {
 769                 lprocfs_counter_init(obd->obd_stats, LPROC_ECHO_READ_BYTES,
 770                                      LPROCFS_CNTR_AVGMINMAX,
 771                                      "read_bytes", "bytes");
 772                 lprocfs_counter_init(obd->obd_stats, LPROC_ECHO_WRITE_BYTES,
 773                                      LPROCFS_CNTR_AVGMINMAX,
 774                                      "write_bytes", "bytes");
 775         }
 776
 777         ptlrpc_init_client(LDLM_CB_REQUEST_PORTAL, LDLM_CB_REPLY_PORTAL,
 778                            "echo_ldlm_cb_client", &obd->obd_ldlm_client);
 779
 780         rc = tgt_init(env, &esd->esd_lut, obd, NULL, esd_common_slice,
 781                       OBD_FAIL_OST_ALL_REQUEST_NET,
 782                       OBD_FAIL_OST_ALL_REPLY_NET);
 783         if (rc)
 784                 GOTO(err_out, rc);
 785
 786         spin_lock(&obd->obd_dev_lock);
 787         obd->obd_no_conn = 0;
 788         spin_unlock(&obd->obd_dev_lock);
 789
 790         RETURN(0);
 791
 792 err_out:
 793         ldlm_namespace_free(obd->obd_namespace, NULL, obd->obd_force);
 794         obd->obd_namespace = NULL;
 795
 796         lprocfs_obd_cleanup(obd);
 797         lprocfs_free_obd_stats(obd);
 798         RETURN(rc);
 799 }
 800
 801 /**
 802  * Stop the Echo Server device.
 803  *
 804  * This function stops the Echo Server device and all its subsystems.
 805  * This is the end of Echo Server lifecycle.
 806  *
 807  * \param[in] env       execution environment
 808  * \param[in] esd               ESD device
 809  */
 810 static void echo_srv_fini(const struct lu_env *env,
 811                           struct echo_srv_device *esd)
 812 {
 813         struct obd_device *obd = echo_srv_obd(esd);
 814         struct lu_device *d = &esd->esd_dev;
 815         int leaked;
 816
 817         ENTRY;
 818
 819         class_disconnect_exports(obd);
 820         if (obd->obd_namespace != NULL)
 821                 ldlm_namespace_free_prior(obd->obd_namespace, NULL,
 822                                           obd->obd_force);
 823
 824         obd_exports_barrier(obd);
 825         obd_zombie_barrier();
 826
 827         tgt_fini(env, &esd->esd_lut);
 828
 829         if (obd->obd_namespace != NULL) {
 830                 ldlm_namespace_free_post(obd->obd_namespace);
 831                 obd->obd_namespace = NULL;
 832         }
 833
 834         lprocfs_obd_cleanup(obd);
 835         lprocfs_free_obd_stats(obd);
 836
 837         leaked = atomic_read(&obd->u.echo.eo_prep);
 838         if (leaked != 0)
 839                 CERROR("%d prep/commitrw pages leaked\n", leaked);
 840
 841         LASSERT(atomic_read(&d->ld_ref) == 0);
 842         EXIT;
 843 }
 844
 845 /**
 846  * Implementation of lu_device_type_operations::ldto_device_fini.
 847  *
 848  * Finalize device. Dual to echo_srv_device_init(). It is called from
 849  * obd_precleanup() and stops the current device.
 850  *
 851  * \param[in] env       execution environment
 852  * \param[in] d         LU device of ESD
 853  *
 854  * \retval              NULL
 855  */
 856 static struct lu_device *echo_srv_device_fini(const struct lu_env *env,
 857                                               struct lu_device *d)
 858 {
 859         ENTRY;
 860         echo_srv_fini(env, echo_srv_dev(d));
 861         RETURN(NULL);
 862 }
 863
 864 /**
 865  * Implementation of lu_device_type_operations::ldto_device_free.
 866  *
 867  * Free Echo Server device. Dual to echo_srv_device_alloc().
 868  *
 869  * \param[in] env       execution environment
 870  * \param[in] d         LU device of ESD
 871  *
 872  * \retval              NULL
 873  */
 874 static struct lu_device *echo_srv_device_free(const struct lu_env *env,
 875                                               struct lu_device *d)
 876 {
 877         struct echo_srv_device *esd = echo_srv_dev(d);
 878
 879         lu_device_fini(&esd->esd_dev);
 880         OBD_FREE_PTR(esd);
 881         RETURN(NULL);
 882 }
 883
 884 /**
 885  * Implementation of lu_device_type_operations::ldto_device_alloc.
 886  *
 887  * This function allocates the new Echo Server device. It is called from
 888  * obd_setup() if OBD device had lu_device_type defined.
 889  *
 890  * \param[in] env       execution environment
 891  * \param[in] t         lu_device_type of ESD device
 892  * \param[in] cfg       configuration log
 893  *
 894  * \retval              pointer to the lu_device of just allocated OFD
 895  * \retval              ERR_PTR of return value on error
 896  */
 897 static struct lu_device *echo_srv_device_alloc(const struct lu_env *env,
 898                                                struct lu_device_type *t,
 899                                                struct lustre_cfg *cfg)
 900 {
 901         struct echo_srv_device *esd;
 902         struct lu_device *l;
 903         int rc;
 904
 905         OBD_ALLOC_PTR(esd);
 906         if (esd == NULL)
 907                 return ERR_PTR(-ENOMEM);
 908
 909         l = &esd->esd_dev;
 910         lu_device_init(l, t);
 911         rc = echo_srv_init0(env, esd, t, cfg);
 912         if (rc != 0) {
 913                 echo_srv_device_free(env, l);
 914                 l = ERR_PTR(rc);
 915         }
 916
 917         return l;
 918 }
 919
 920 static const struct lu_device_type_operations echo_srv_type_ops = {
 921         .ldto_device_alloc = echo_srv_device_alloc,
 922         .ldto_device_free = echo_srv_device_free,
 923         .ldto_device_fini = echo_srv_device_fini
 924 };
 925
 926 struct lu_device_type echo_srv_type = {
 927         .ldt_tags = LU_DEVICE_DT,
 928         .ldt_name = LUSTRE_ECHO_NAME,
 929         .ldt_ops = &echo_srv_type_ops,
 930         .ldt_ctx_tags = LCT_DT_THREAD,
 931 };
 932
 933 void echo_persistent_pages_fini(void)
 934 {
 935         int i;
 936
 937         for (i = 0; i < ECHO_PERSISTENT_PAGES; i++)
 938                 if (echo_persistent_pages[i] != NULL) {
 939                         __free_page(echo_persistent_pages[i]);
 940                         echo_persistent_pages[i] = NULL;
 941                 }
 942 }
 943
 944 int echo_persistent_pages_init(void)
 945 {
 946         struct page *pg;
 947         int          i;
 948
 949         for (i = 0; i < ECHO_PERSISTENT_PAGES; i++) {
 950                 gfp_t gfp_mask = (i < ECHO_PERSISTENT_PAGES/2) ?
 951                         GFP_KERNEL : GFP_HIGHUSER;
 952
 953                 pg = alloc_page(gfp_mask);
 954                 if (pg == NULL) {
 955                         echo_persistent_pages_fini();
 956                         return -ENOMEM;
 957                 }
 958
 959                 memset(kmap(pg), 0, PAGE_SIZE);
 960                 kunmap(pg);
 961
 962                 echo_persistent_pages[i] = pg;
 963         }
 964
 965         return 0;
 966 }