kernel/drivers/staging/lustre/lustre/obdecho/echo_client.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.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2011, 2012, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  */
  36
  37 #define DEBUG_SUBSYSTEM S_ECHO
  38 #include "../../include/linux/libcfs/libcfs.h"
  39
  40 #include "../include/obd.h"
  41 #include "../include/obd_support.h"
  42 #include "../include/obd_class.h"
  43 #include "../include/lustre_debug.h"
  44 #include "../include/lprocfs_status.h"
  45 #include "../include/cl_object.h"
  46 #include "../include/lustre_fid.h"
  47 #include "../include/lustre_acl.h"
  48 #include "../include/lustre_net.h"
  49
  50 #include "echo_internal.h"
  51
  52 /** \defgroup echo_client Echo Client
  53  * @{
  54  */
  55
  56 struct echo_device {
  57         struct cl_device        ed_cl;
  58         struct echo_client_obd *ed_ec;
  59
  60         struct cl_site    ed_site_myself;
  61         struct cl_site   *ed_site;
  62         struct lu_device       *ed_next;
  63         int                  ed_next_islov;
  64 };
  65
  66 struct echo_object {
  67         struct cl_object        eo_cl;
  68         struct cl_object_header eo_hdr;
  69
  70         struct echo_device     *eo_dev;
  71         struct list_head              eo_obj_chain;
  72         struct lov_stripe_md   *eo_lsm;
  73         atomic_t            eo_npages;
  74         int                  eo_deleted;
  75 };
  76
  77 struct echo_object_conf {
  78         struct cl_object_conf  eoc_cl;
  79         struct lov_stripe_md **eoc_md;
  80 };
  81
  82 struct echo_page {
  83         struct cl_page_slice   ep_cl;
  84         struct mutex            ep_lock;
  85         struct page         *ep_vmpage;
  86 };
  87
  88 struct echo_lock {
  89         struct cl_lock_slice   el_cl;
  90         struct list_head             el_chain;
  91         struct echo_object    *el_object;
  92         __u64             el_cookie;
  93         atomic_t           el_refcount;
  94 };
  95
  96 static int echo_client_setup(const struct lu_env *env,
  97                              struct obd_device *obddev,
  98                              struct lustre_cfg *lcfg);
  99 static int echo_client_cleanup(struct obd_device *obddev);
 100
 101
 102 /** \defgroup echo_helpers Helper functions
 103  * @{
 104  */
 105 static inline struct echo_device *cl2echo_dev(const struct cl_device *dev)
 106 {
 107         return container_of0(dev, struct echo_device, ed_cl);
 108 }
 109
 110 static inline struct cl_device *echo_dev2cl(struct echo_device *d)
 111 {
 112         return &d->ed_cl;
 113 }
 114
 115 static inline struct echo_device *obd2echo_dev(const struct obd_device *obd)
 116 {
 117         return cl2echo_dev(lu2cl_dev(obd->obd_lu_dev));
 118 }
 119
 120 static inline struct cl_object *echo_obj2cl(struct echo_object *eco)
 121 {
 122         return &eco->eo_cl;
 123 }
 124
 125 static inline struct echo_object *cl2echo_obj(const struct cl_object *o)
 126 {
 127         return container_of(o, struct echo_object, eo_cl);
 128 }
 129
 130 static inline struct echo_page *cl2echo_page(const struct cl_page_slice *s)
 131 {
 132         return container_of(s, struct echo_page, ep_cl);
 133 }
 134
 135 static inline struct echo_lock *cl2echo_lock(const struct cl_lock_slice *s)
 136 {
 137         return container_of(s, struct echo_lock, el_cl);
 138 }
 139
 140 static inline struct cl_lock *echo_lock2cl(const struct echo_lock *ecl)
 141 {
 142         return ecl->el_cl.cls_lock;
 143 }
 144
 145 static struct lu_context_key echo_thread_key;
 146 static inline struct echo_thread_info *echo_env_info(const struct lu_env *env)
 147 {
 148         struct echo_thread_info *info;
 149
 150         info = lu_context_key_get(&env->le_ctx, &echo_thread_key);
 151         LASSERT(info != NULL);
 152         return info;
 153 }
 154
 155 static inline
 156 struct echo_object_conf *cl2echo_conf(const struct cl_object_conf *c)
 157 {
 158         return container_of(c, struct echo_object_conf, eoc_cl);
 159 }
 160
 161 /** @} echo_helpers */
 162
 163 static struct echo_object *cl_echo_object_find(struct echo_device *d,
 164                                                struct lov_stripe_md **lsm);
 165 static int cl_echo_object_put(struct echo_object *eco);
 166 static int cl_echo_enqueue(struct echo_object *eco, u64 start,
 167                            u64 end, int mode, __u64 *cookie);
 168 static int cl_echo_cancel(struct echo_device *d, __u64 cookie);
 169 static int cl_echo_object_brw(struct echo_object *eco, int rw, u64 offset,
 170                               struct page **pages, int npages, int async);
 171
 172 static struct echo_thread_info *echo_env_info(const struct lu_env *env);
 173
 174 struct echo_thread_info {
 175         struct echo_object_conf eti_conf;
 176         struct lustre_md        eti_md;
 177
 178         struct cl_2queue        eti_queue;
 179         struct cl_io        eti_io;
 180         struct cl_lock_descr    eti_descr;
 181         struct lu_fid      eti_fid;
 182         struct lu_fid           eti_fid2;
 183 };
 184
 185 /* No session used right now */
 186 struct echo_session_info {
 187         unsigned long dummy;
 188 };
 189
 190 static struct kmem_cache *echo_lock_kmem;
 191 static struct kmem_cache *echo_object_kmem;
 192 static struct kmem_cache *echo_thread_kmem;
 193 static struct kmem_cache *echo_session_kmem;
 194
 195 static struct lu_kmem_descr echo_caches[] = {
 196         {
 197                 .ckd_cache = &echo_lock_kmem,
 198                 .ckd_name  = "echo_lock_kmem",
 199                 .ckd_size  = sizeof(struct echo_lock)
 200         },
 201         {
 202                 .ckd_cache = &echo_object_kmem,
 203                 .ckd_name  = "echo_object_kmem",
 204                 .ckd_size  = sizeof(struct echo_object)
 205         },
 206         {
 207                 .ckd_cache = &echo_thread_kmem,
 208                 .ckd_name  = "echo_thread_kmem",
 209                 .ckd_size  = sizeof(struct echo_thread_info)
 210         },
 211         {
 212                 .ckd_cache = &echo_session_kmem,
 213                 .ckd_name  = "echo_session_kmem",
 214                 .ckd_size  = sizeof(struct echo_session_info)
 215         },
 216         {
 217                 .ckd_cache = NULL
 218         }
 219 };
 220
 221 /** \defgroup echo_page Page operations
 222  *
 223  * Echo page operations.
 224  *
 225  * @{
 226  */
 227 static struct page *echo_page_vmpage(const struct lu_env *env,
 228                                     const struct cl_page_slice *slice)
 229 {
 230         return cl2echo_page(slice)->ep_vmpage;
 231 }
 232
 233 static int echo_page_own(const struct lu_env *env,
 234                          const struct cl_page_slice *slice,
 235                          struct cl_io *io, int nonblock)
 236 {
 237         struct echo_page *ep = cl2echo_page(slice);
 238
 239         if (!nonblock)
 240                 mutex_lock(&ep->ep_lock);
 241         else if (!mutex_trylock(&ep->ep_lock))
 242                 return -EAGAIN;
 243         return 0;
 244 }
 245
 246 static void echo_page_disown(const struct lu_env *env,
 247                              const struct cl_page_slice *slice,
 248                              struct cl_io *io)
 249 {
 250         struct echo_page *ep = cl2echo_page(slice);
 251
 252         LASSERT(mutex_is_locked(&ep->ep_lock));
 253         mutex_unlock(&ep->ep_lock);
 254 }
 255
 256 static void echo_page_discard(const struct lu_env *env,
 257                               const struct cl_page_slice *slice,
 258                               struct cl_io *unused)
 259 {
 260         cl_page_delete(env, slice->cpl_page);
 261 }
 262
 263 static int echo_page_is_vmlocked(const struct lu_env *env,
 264                                  const struct cl_page_slice *slice)
 265 {
 266         if (mutex_is_locked(&cl2echo_page(slice)->ep_lock))
 267                 return -EBUSY;
 268         return -ENODATA;
 269 }
 270
 271 static void echo_page_completion(const struct lu_env *env,
 272                                  const struct cl_page_slice *slice,
 273                                  int ioret)
 274 {
 275         LASSERT(slice->cpl_page->cp_sync_io != NULL);
 276 }
 277
 278 static void echo_page_fini(const struct lu_env *env,
 279                            struct cl_page_slice *slice)
 280 {
 281         struct echo_page *ep    = cl2echo_page(slice);
 282         struct echo_object *eco = cl2echo_obj(slice->cpl_obj);
 283         struct page *vmpage      = ep->ep_vmpage;
 284
 285         atomic_dec(&eco->eo_npages);
 286         page_cache_release(vmpage);
 287 }
 288
 289 static int echo_page_prep(const struct lu_env *env,
 290                           const struct cl_page_slice *slice,
 291                           struct cl_io *unused)
 292 {
 293         return 0;
 294 }
 295
 296 static int echo_page_print(const struct lu_env *env,
 297                            const struct cl_page_slice *slice,
 298                            void *cookie, lu_printer_t printer)
 299 {
 300         struct echo_page *ep = cl2echo_page(slice);
 301
 302         (*printer)(env, cookie, LUSTRE_ECHO_CLIENT_NAME"-page@%p %d vm@%p\n",
 303                    ep, mutex_is_locked(&ep->ep_lock), ep->ep_vmpage);
 304         return 0;
 305 }
 306
 307 static const struct cl_page_operations echo_page_ops = {
 308         .cpo_own           = echo_page_own,
 309         .cpo_disown     = echo_page_disown,
 310         .cpo_discard       = echo_page_discard,
 311         .cpo_vmpage     = echo_page_vmpage,
 312         .cpo_fini         = echo_page_fini,
 313         .cpo_print       = echo_page_print,
 314         .cpo_is_vmlocked   = echo_page_is_vmlocked,
 315         .io = {
 316                 [CRT_READ] = {
 317                         .cpo_prep       = echo_page_prep,
 318                         .cpo_completion  = echo_page_completion,
 319                 },
 320                 [CRT_WRITE] = {
 321                         .cpo_prep       = echo_page_prep,
 322                         .cpo_completion  = echo_page_completion,
 323                 }
 324         }
 325 };
 326 /** @} echo_page */
 327
 328 /** \defgroup echo_lock Locking
 329  *
 330  * echo lock operations
 331  *
 332  * @{
 333  */
 334 static void echo_lock_fini(const struct lu_env *env,
 335                            struct cl_lock_slice *slice)
 336 {
 337         struct echo_lock *ecl = cl2echo_lock(slice);
 338
 339         LASSERT(list_empty(&ecl->el_chain));
 340         OBD_SLAB_FREE_PTR(ecl, echo_lock_kmem);
 341 }
 342
 343 static void echo_lock_delete(const struct lu_env *env,
 344                              const struct cl_lock_slice *slice)
 345 {
 346         struct echo_lock *ecl      = cl2echo_lock(slice);
 347
 348         LASSERT(list_empty(&ecl->el_chain));
 349 }
 350
 351 static int echo_lock_fits_into(const struct lu_env *env,
 352                                const struct cl_lock_slice *slice,
 353                                const struct cl_lock_descr *need,
 354                                const struct cl_io *unused)
 355 {
 356         return 1;
 357 }
 358
 359 static struct cl_lock_operations echo_lock_ops = {
 360         .clo_fini      = echo_lock_fini,
 361         .clo_delete    = echo_lock_delete,
 362         .clo_fits_into = echo_lock_fits_into
 363 };
 364
 365 /** @} echo_lock */
 366
 367 /** \defgroup echo_cl_ops cl_object operations
 368  *
 369  * operations for cl_object
 370  *
 371  * @{
 372  */
 373 static int echo_page_init(const struct lu_env *env, struct cl_object *obj,
 374                         struct cl_page *page, struct page *vmpage)
 375 {
 376         struct echo_page *ep = cl_object_page_slice(obj, page);
 377         struct echo_object *eco = cl2echo_obj(obj);
 378
 379         ep->ep_vmpage = vmpage;
 380         page_cache_get(vmpage);
 381         mutex_init(&ep->ep_lock);
 382         cl_page_slice_add(page, &ep->ep_cl, obj, &echo_page_ops);
 383         atomic_inc(&eco->eo_npages);
 384         return 0;
 385 }
 386
 387 static int echo_io_init(const struct lu_env *env, struct cl_object *obj,
 388                         struct cl_io *io)
 389 {
 390         return 0;
 391 }
 392
 393 static int echo_lock_init(const struct lu_env *env,
 394                           struct cl_object *obj, struct cl_lock *lock,
 395                           const struct cl_io *unused)
 396 {
 397         struct echo_lock *el;
 398
 399         OBD_SLAB_ALLOC_PTR_GFP(el, echo_lock_kmem, GFP_NOFS);
 400         if (el != NULL) {
 401                 cl_lock_slice_add(lock, &el->el_cl, obj, &echo_lock_ops);
 402                 el->el_object = cl2echo_obj(obj);
 403                 INIT_LIST_HEAD(&el->el_chain);
 404                 atomic_set(&el->el_refcount, 0);
 405         }
 406         return el == NULL ? -ENOMEM : 0;
 407 }
 408
 409 static int echo_conf_set(const struct lu_env *env, struct cl_object *obj,
 410                          const struct cl_object_conf *conf)
 411 {
 412         return 0;
 413 }
 414
 415 static const struct cl_object_operations echo_cl_obj_ops = {
 416         .coo_page_init = echo_page_init,
 417         .coo_lock_init = echo_lock_init,
 418         .coo_io_init   = echo_io_init,
 419         .coo_conf_set  = echo_conf_set
 420 };
 421 /** @} echo_cl_ops */
 422
 423 /** \defgroup echo_lu_ops lu_object operations
 424  *
 425  * operations for echo lu object.
 426  *
 427  * @{
 428  */
 429 static int echo_object_init(const struct lu_env *env, struct lu_object *obj,
 430                             const struct lu_object_conf *conf)
 431 {
 432         struct echo_device *ed   = cl2echo_dev(lu2cl_dev(obj->lo_dev));
 433         struct echo_client_obd *ec     = ed->ed_ec;
 434         struct echo_object *eco = cl2echo_obj(lu2cl(obj));
 435         const struct cl_object_conf *cconf;
 436         struct echo_object_conf *econf;
 437
 438         if (ed->ed_next) {
 439                 struct lu_object  *below;
 440                 struct lu_device  *under;
 441
 442                 under = ed->ed_next;
 443                 below = under->ld_ops->ldo_object_alloc(env, obj->lo_header,
 444                                                         under);
 445                 if (below == NULL)
 446                         return -ENOMEM;
 447                 lu_object_add(obj, below);
 448         }
 449
 450         cconf = lu2cl_conf(conf);
 451         econf = cl2echo_conf(cconf);
 452
 453         LASSERT(econf->eoc_md);
 454         eco->eo_lsm = *econf->eoc_md;
 455         /* clear the lsm pointer so that it won't get freed. */
 456         *econf->eoc_md = NULL;
 457
 458         eco->eo_dev = ed;
 459         atomic_set(&eco->eo_npages, 0);
 460         cl_object_page_init(lu2cl(obj), sizeof(struct echo_page));
 461
 462         spin_lock(&ec->ec_lock);
 463         list_add_tail(&eco->eo_obj_chain, &ec->ec_objects);
 464         spin_unlock(&ec->ec_lock);
 465
 466         return 0;
 467 }
 468
 469 /* taken from osc_unpackmd() */
 470 static int echo_alloc_memmd(struct echo_device *ed,
 471                             struct lov_stripe_md **lsmp)
 472 {
 473         int lsm_size;
 474
 475         /* If export is lov/osc then use their obd method */
 476         if (ed->ed_next != NULL)
 477                 return obd_alloc_memmd(ed->ed_ec->ec_exp, lsmp);
 478         /* OFD has no unpackmd method, do everything here */
 479         lsm_size = lov_stripe_md_size(1);
 480
 481         LASSERT(*lsmp == NULL);
 482         OBD_ALLOC(*lsmp, lsm_size);
 483         if (*lsmp == NULL)
 484                 return -ENOMEM;
 485
 486         OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
 487         if ((*lsmp)->lsm_oinfo[0] == NULL) {
 488                 OBD_FREE(*lsmp, lsm_size);
 489                 return -ENOMEM;
 490         }
 491
 492         loi_init((*lsmp)->lsm_oinfo[0]);
 493         (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
 494         ostid_set_seq_echo(&(*lsmp)->lsm_oi);
 495
 496         return lsm_size;
 497 }
 498
 499 static int echo_free_memmd(struct echo_device *ed, struct lov_stripe_md **lsmp)
 500 {
 501         int lsm_size;
 502
 503         /* If export is lov/osc then use their obd method */
 504         if (ed->ed_next != NULL)
 505                 return obd_free_memmd(ed->ed_ec->ec_exp, lsmp);
 506         /* OFD has no unpackmd method, do everything here */
 507         lsm_size = lov_stripe_md_size(1);
 508
 509         LASSERT(*lsmp != NULL);
 510         OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
 511         OBD_FREE(*lsmp, lsm_size);
 512         *lsmp = NULL;
 513         return 0;
 514 }
 515
 516 static void echo_object_free(const struct lu_env *env, struct lu_object *obj)
 517 {
 518         struct echo_object *eco    = cl2echo_obj(lu2cl(obj));
 519         struct echo_client_obd *ec = eco->eo_dev->ed_ec;
 520
 521         LASSERT(atomic_read(&eco->eo_npages) == 0);
 522
 523         spin_lock(&ec->ec_lock);
 524         list_del_init(&eco->eo_obj_chain);
 525         spin_unlock(&ec->ec_lock);
 526
 527         lu_object_fini(obj);
 528         lu_object_header_fini(obj->lo_header);
 529
 530         if (eco->eo_lsm)
 531                 echo_free_memmd(eco->eo_dev, &eco->eo_lsm);
 532         OBD_SLAB_FREE_PTR(eco, echo_object_kmem);
 533 }
 534
 535 static int echo_object_print(const struct lu_env *env, void *cookie,
 536                             lu_printer_t p, const struct lu_object *o)
 537 {
 538         struct echo_object *obj = cl2echo_obj(lu2cl(o));
 539
 540         return (*p)(env, cookie, "echoclient-object@%p", obj);
 541 }
 542
 543 static const struct lu_object_operations echo_lu_obj_ops = {
 544         .loo_object_init      = echo_object_init,
 545         .loo_object_delete    = NULL,
 546         .loo_object_release   = NULL,
 547         .loo_object_free      = echo_object_free,
 548         .loo_object_print     = echo_object_print,
 549         .loo_object_invariant = NULL
 550 };
 551 /** @} echo_lu_ops */
 552
 553 /** \defgroup echo_lu_dev_ops  lu_device operations
 554  *
 555  * Operations for echo lu device.
 556  *
 557  * @{
 558  */
 559 static struct lu_object *echo_object_alloc(const struct lu_env *env,
 560                                            const struct lu_object_header *hdr,
 561                                            struct lu_device *dev)
 562 {
 563         struct echo_object *eco;
 564         struct lu_object *obj = NULL;
 565
 566         /* we're the top dev. */
 567         LASSERT(hdr == NULL);
 568         OBD_SLAB_ALLOC_PTR_GFP(eco, echo_object_kmem, GFP_NOFS);
 569         if (eco != NULL) {
 570                 struct cl_object_header *hdr = &eco->eo_hdr;
 571
 572                 obj = &echo_obj2cl(eco)->co_lu;
 573                 cl_object_header_init(hdr);
 574                 lu_object_init(obj, &hdr->coh_lu, dev);
 575                 lu_object_add_top(&hdr->coh_lu, obj);
 576
 577                 eco->eo_cl.co_ops = &echo_cl_obj_ops;
 578                 obj->lo_ops       = &echo_lu_obj_ops;
 579         }
 580         return obj;
 581 }
 582
 583 static struct lu_device_operations echo_device_lu_ops = {
 584         .ldo_object_alloc   = echo_object_alloc,
 585 };
 586
 587 /** @} echo_lu_dev_ops */
 588
 589 static struct cl_device_operations echo_device_cl_ops = {
 590 };
 591
 592 /** \defgroup echo_init Setup and teardown
 593  *
 594  * Init and fini functions for echo client.
 595  *
 596  * @{
 597  */
 598 static int echo_site_init(const struct lu_env *env, struct echo_device *ed)
 599 {
 600         struct cl_site *site = &ed->ed_site_myself;
 601         int rc;
 602
 603         /* initialize site */
 604         rc = cl_site_init(site, &ed->ed_cl);
 605         if (rc) {
 606                 CERROR("Cannot initialize site for echo client(%d)\n", rc);
 607                 return rc;
 608         }
 609
 610         rc = lu_site_init_finish(&site->cs_lu);
 611         if (rc)
 612                 return rc;
 613
 614         ed->ed_site = site;
 615         return 0;
 616 }
 617
 618 static void echo_site_fini(const struct lu_env *env, struct echo_device *ed)
 619 {
 620         if (ed->ed_site) {
 621                 cl_site_fini(ed->ed_site);
 622                 ed->ed_site = NULL;
 623         }
 624 }
 625
 626 static void *echo_thread_key_init(const struct lu_context *ctx,
 627                           struct lu_context_key *key)
 628 {
 629         struct echo_thread_info *info;
 630
 631         OBD_SLAB_ALLOC_PTR_GFP(info, echo_thread_kmem, GFP_NOFS);
 632         if (info == NULL)
 633                 info = ERR_PTR(-ENOMEM);
 634         return info;
 635 }
 636
 637 static void echo_thread_key_fini(const struct lu_context *ctx,
 638                          struct lu_context_key *key, void *data)
 639 {
 640         struct echo_thread_info *info = data;
 641
 642         OBD_SLAB_FREE_PTR(info, echo_thread_kmem);
 643 }
 644
 645 static void echo_thread_key_exit(const struct lu_context *ctx,
 646                          struct lu_context_key *key, void *data)
 647 {
 648 }
 649
 650 static struct lu_context_key echo_thread_key = {
 651         .lct_tags = LCT_CL_THREAD,
 652         .lct_init = echo_thread_key_init,
 653         .lct_fini = echo_thread_key_fini,
 654         .lct_exit = echo_thread_key_exit
 655 };
 656
 657 static void *echo_session_key_init(const struct lu_context *ctx,
 658                                   struct lu_context_key *key)
 659 {
 660         struct echo_session_info *session;
 661
 662         OBD_SLAB_ALLOC_PTR_GFP(session, echo_session_kmem, GFP_NOFS);
 663         if (session == NULL)
 664                 session = ERR_PTR(-ENOMEM);
 665         return session;
 666 }
 667
 668 static void echo_session_key_fini(const struct lu_context *ctx,
 669                                  struct lu_context_key *key, void *data)
 670 {
 671         struct echo_session_info *session = data;
 672
 673         OBD_SLAB_FREE_PTR(session, echo_session_kmem);
 674 }
 675
 676 static void echo_session_key_exit(const struct lu_context *ctx,
 677                                  struct lu_context_key *key, void *data)
 678 {
 679 }
 680
 681 static struct lu_context_key echo_session_key = {
 682         .lct_tags = LCT_SESSION,
 683         .lct_init = echo_session_key_init,
 684         .lct_fini = echo_session_key_fini,
 685         .lct_exit = echo_session_key_exit
 686 };
 687
 688 LU_TYPE_INIT_FINI(echo, &echo_thread_key, &echo_session_key);
 689
 690 static struct lu_device *echo_device_alloc(const struct lu_env *env,
 691                                            struct lu_device_type *t,
 692                                            struct lustre_cfg *cfg)
 693 {
 694         struct lu_device   *next;
 695         struct echo_device *ed;
 696         struct cl_device   *cd;
 697         struct obd_device  *obd = NULL; /* to keep compiler happy */
 698         struct obd_device  *tgt;
 699         const char *tgt_type_name;
 700         int rc;
 701         int cleanup = 0;
 702
 703         OBD_ALLOC_PTR(ed);
 704         if (ed == NULL) {
 705                 rc = -ENOMEM;
 706                 goto out;
 707         }
 708
 709         cleanup = 1;
 710         cd = &ed->ed_cl;
 711         rc = cl_device_init(cd, t);
 712         if (rc)
 713                 goto out;
 714
 715         cd->cd_lu_dev.ld_ops = &echo_device_lu_ops;
 716         cd->cd_ops = &echo_device_cl_ops;
 717
 718         cleanup = 2;
 719         obd = class_name2obd(lustre_cfg_string(cfg, 0));
 720         LASSERT(obd != NULL);
 721         LASSERT(env != NULL);
 722
 723         tgt = class_name2obd(lustre_cfg_string(cfg, 1));
 724         if (tgt == NULL) {
 725                 CERROR("Can not find tgt device %s\n",
 726                         lustre_cfg_string(cfg, 1));
 727                 rc = -ENODEV;
 728                 goto out;
 729         }
 730
 731         next = tgt->obd_lu_dev;
 732         if (!strcmp(tgt->obd_type->typ_name, LUSTRE_MDT_NAME)) {
 733                 CERROR("echo MDT client must be run on server\n");
 734                 rc = -EOPNOTSUPP;
 735                 goto out;
 736         }
 737
 738         rc = echo_site_init(env, ed);
 739         if (rc)
 740                 goto out;
 741
 742         cleanup = 3;
 743
 744         rc = echo_client_setup(env, obd, cfg);
 745         if (rc)
 746                 goto out;
 747
 748         ed->ed_ec = &obd->u.echo_client;
 749         cleanup = 4;
 750
 751         /* if echo client is to be stacked upon ost device, the next is
 752          * NULL since ost is not a clio device so far */
 753         if (next != NULL && !lu_device_is_cl(next))
 754                 next = NULL;
 755
 756         tgt_type_name = tgt->obd_type->typ_name;
 757         if (next != NULL) {
 758                 LASSERT(next != NULL);
 759                 if (next->ld_site != NULL) {
 760                         rc = -EBUSY;
 761                         goto out;
 762                 }
 763
 764                 next->ld_site = &ed->ed_site->cs_lu;
 765                 rc = next->ld_type->ldt_ops->ldto_device_init(env, next,
 766                                                 next->ld_type->ldt_name,
 767                                                               NULL);
 768                 if (rc)
 769                         goto out;
 770
 771                 /* Tricky case, I have to determine the obd type since
 772                  * CLIO uses the different parameters to initialize
 773                  * objects for lov & osc. */
 774                 if (strcmp(tgt_type_name, LUSTRE_LOV_NAME) == 0)
 775                         ed->ed_next_islov = 1;
 776                 else
 777                         LASSERT(strcmp(tgt_type_name,
 778                                        LUSTRE_OSC_NAME) == 0);
 779         } else {
 780                 LASSERT(strcmp(tgt_type_name, LUSTRE_OST_NAME) == 0);
 781         }
 782
 783         ed->ed_next = next;
 784         return &cd->cd_lu_dev;
 785 out:
 786         switch (cleanup) {
 787         case 4: {
 788                 int rc2;
 789
 790                 rc2 = echo_client_cleanup(obd);
 791                 if (rc2)
 792                         CERROR("Cleanup obd device %s error(%d)\n",
 793                                obd->obd_name, rc2);
 794         }
 795
 796         case 3:
 797                 echo_site_fini(env, ed);
 798         case 2:
 799                 cl_device_fini(&ed->ed_cl);
 800         case 1:
 801                 OBD_FREE_PTR(ed);
 802         case 0:
 803         default:
 804                 break;
 805         }
 806         return ERR_PTR(rc);
 807 }
 808
 809 static int echo_device_init(const struct lu_env *env, struct lu_device *d,
 810                           const char *name, struct lu_device *next)
 811 {
 812         LBUG();
 813         return 0;
 814 }
 815
 816 static struct lu_device *echo_device_fini(const struct lu_env *env,
 817                                           struct lu_device *d)
 818 {
 819         struct echo_device *ed = cl2echo_dev(lu2cl_dev(d));
 820         struct lu_device *next = ed->ed_next;
 821
 822         while (next)
 823                 next = next->ld_type->ldt_ops->ldto_device_fini(env, next);
 824         return NULL;
 825 }
 826
 827 static void echo_lock_release(const struct lu_env *env,
 828                               struct echo_lock *ecl,
 829                               int still_used)
 830 {
 831         struct cl_lock *clk = echo_lock2cl(ecl);
 832
 833         cl_lock_get(clk);
 834         cl_unuse(env, clk);
 835         cl_lock_release(env, clk, "ec enqueue", ecl->el_object);
 836         if (!still_used) {
 837                 cl_lock_mutex_get(env, clk);
 838                 cl_lock_cancel(env, clk);
 839                 cl_lock_delete(env, clk);
 840                 cl_lock_mutex_put(env, clk);
 841         }
 842         cl_lock_put(env, clk);
 843 }
 844
 845 static struct lu_device *echo_device_free(const struct lu_env *env,
 846                                           struct lu_device *d)
 847 {
 848         struct echo_device     *ed   = cl2echo_dev(lu2cl_dev(d));
 849         struct echo_client_obd *ec   = ed->ed_ec;
 850         struct echo_object     *eco;
 851         struct lu_device       *next = ed->ed_next;
 852
 853         CDEBUG(D_INFO, "echo device:%p is going to be freed, next = %p\n",
 854                ed, next);
 855
 856         lu_site_purge(env, &ed->ed_site->cs_lu, -1);
 857
 858         /* check if there are objects still alive.
 859          * It shouldn't have any object because lu_site_purge would cleanup
 860          * all of cached objects. Anyway, probably the echo device is being
 861          * parallelly accessed.
 862          */
 863         spin_lock(&ec->ec_lock);
 864         list_for_each_entry(eco, &ec->ec_objects, eo_obj_chain)
 865                 eco->eo_deleted = 1;
 866         spin_unlock(&ec->ec_lock);
 867
 868         /* purge again */
 869         lu_site_purge(env, &ed->ed_site->cs_lu, -1);
 870
 871         CDEBUG(D_INFO,
 872                "Waiting for the reference of echo object to be dropped\n");
 873
 874         /* Wait for the last reference to be dropped. */
 875         spin_lock(&ec->ec_lock);
 876         while (!list_empty(&ec->ec_objects)) {
 877                 spin_unlock(&ec->ec_lock);
 878                 CERROR("echo_client still has objects at cleanup time, wait for 1 second\n");
 879                 set_current_state(TASK_UNINTERRUPTIBLE);
 880                 schedule_timeout(cfs_time_seconds(1));
 881                 lu_site_purge(env, &ed->ed_site->cs_lu, -1);
 882                 spin_lock(&ec->ec_lock);
 883         }
 884         spin_unlock(&ec->ec_lock);
 885
 886         LASSERT(list_empty(&ec->ec_locks));
 887
 888         CDEBUG(D_INFO, "No object exists, exiting...\n");
 889
 890         echo_client_cleanup(d->ld_obd);
 891
 892         while (next)
 893                 next = next->ld_type->ldt_ops->ldto_device_free(env, next);
 894
 895         LASSERT(ed->ed_site == lu2cl_site(d->ld_site));
 896         echo_site_fini(env, ed);
 897         cl_device_fini(&ed->ed_cl);
 898         OBD_FREE_PTR(ed);
 899
 900         return NULL;
 901 }
 902
 903 static const struct lu_device_type_operations echo_device_type_ops = {
 904         .ldto_init = echo_type_init,
 905         .ldto_fini = echo_type_fini,
 906
 907         .ldto_start = echo_type_start,
 908         .ldto_stop  = echo_type_stop,
 909
 910         .ldto_device_alloc = echo_device_alloc,
 911         .ldto_device_free  = echo_device_free,
 912         .ldto_device_init  = echo_device_init,
 913         .ldto_device_fini  = echo_device_fini
 914 };
 915
 916 static struct lu_device_type echo_device_type = {
 917         .ldt_tags     = LU_DEVICE_CL,
 918         .ldt_name     = LUSTRE_ECHO_CLIENT_NAME,
 919         .ldt_ops      = &echo_device_type_ops,
 920         .ldt_ctx_tags = LCT_CL_THREAD,
 921 };
 922 /** @} echo_init */
 923
 924 /** \defgroup echo_exports Exported operations
 925  *
 926  * exporting functions to echo client
 927  *
 928  * @{
 929  */
 930
 931 /* Interfaces to echo client obd device */
 932 static struct echo_object *cl_echo_object_find(struct echo_device *d,
 933                                                struct lov_stripe_md **lsmp)
 934 {
 935         struct lu_env *env;
 936         struct echo_thread_info *info;
 937         struct echo_object_conf *conf;
 938         struct lov_stripe_md    *lsm;
 939         struct echo_object *eco;
 940         struct cl_object   *obj;
 941         struct lu_fid *fid;
 942         int refcheck;
 943         int rc;
 944
 945         LASSERT(lsmp);
 946         lsm = *lsmp;
 947         LASSERT(lsm);
 948         LASSERTF(ostid_id(&lsm->lsm_oi) != 0, DOSTID"\n", POSTID(&lsm->lsm_oi));
 949         LASSERTF(ostid_seq(&lsm->lsm_oi) == FID_SEQ_ECHO, DOSTID"\n",
 950                  POSTID(&lsm->lsm_oi));
 951
 952         /* Never return an object if the obd is to be freed. */
 953         if (echo_dev2cl(d)->cd_lu_dev.ld_obd->obd_stopping)
 954                 return ERR_PTR(-ENODEV);
 955
 956         env = cl_env_get(&refcheck);
 957         if (IS_ERR(env))
 958                 return (void *)env;
 959
 960         info = echo_env_info(env);
 961         conf = &info->eti_conf;
 962         if (d->ed_next) {
 963                 if (!d->ed_next_islov) {
 964                         struct lov_oinfo *oinfo = lsm->lsm_oinfo[0];
 965
 966                         LASSERT(oinfo != NULL);
 967                         oinfo->loi_oi = lsm->lsm_oi;
 968                         conf->eoc_cl.u.coc_oinfo = oinfo;
 969                 } else {
 970                         struct lustre_md *md;
 971
 972                         md = &info->eti_md;
 973                         memset(md, 0, sizeof(*md));
 974                         md->lsm = lsm;
 975                         conf->eoc_cl.u.coc_md = md;
 976                 }
 977         }
 978         conf->eoc_md = lsmp;
 979
 980         fid  = &info->eti_fid;
 981         rc = ostid_to_fid(fid, &lsm->lsm_oi, 0);
 982         if (rc != 0) {
 983                 eco = ERR_PTR(rc);
 984                 goto out;
 985         }
 986
 987         /* In the function below, .hs_keycmp resolves to
 988          * lu_obj_hop_keycmp() */
 989         /* coverity[overrun-buffer-val] */
 990         obj = cl_object_find(env, echo_dev2cl(d), fid, &conf->eoc_cl);
 991         if (IS_ERR(obj)) {
 992                 eco = (void *)obj;
 993                 goto out;
 994         }
 995
 996         eco = cl2echo_obj(obj);
 997         if (eco->eo_deleted) {
 998                 cl_object_put(env, obj);
 999                 eco = ERR_PTR(-EAGAIN);
1000         }
1001
1002 out:
1003         cl_env_put(env, &refcheck);
1004         return eco;
1005 }
1006
1007 static int cl_echo_object_put(struct echo_object *eco)
1008 {
1009         struct lu_env *env;
1010         struct cl_object *obj = echo_obj2cl(eco);
1011         int refcheck;
1012
1013         env = cl_env_get(&refcheck);
1014         if (IS_ERR(env))
1015                 return PTR_ERR(env);
1016
1017         /* an external function to kill an object? */
1018         if (eco->eo_deleted) {
1019                 struct lu_object_header *loh = obj->co_lu.lo_header;
1020
1021                 LASSERT(&eco->eo_hdr == luh2coh(loh));
1022                 set_bit(LU_OBJECT_HEARD_BANSHEE, &loh->loh_flags);
1023         }
1024
1025         cl_object_put(env, obj);
1026         cl_env_put(env, &refcheck);
1027         return 0;
1028 }
1029
1030 static int cl_echo_enqueue0(struct lu_env *env, struct echo_object *eco,
1031                             u64 start, u64 end, int mode,
1032                             __u64 *cookie, __u32 enqflags)
1033 {
1034         struct cl_io *io;
1035         struct cl_lock *lck;
1036         struct cl_object *obj;
1037         struct cl_lock_descr *descr;
1038         struct echo_thread_info *info;
1039         int rc = -ENOMEM;
1040
1041         info = echo_env_info(env);
1042         io = &info->eti_io;
1043         descr = &info->eti_descr;
1044         obj = echo_obj2cl(eco);
1045
1046         descr->cld_obj   = obj;
1047         descr->cld_start = cl_index(obj, start);
1048         descr->cld_end   = cl_index(obj, end);
1049         descr->cld_mode  = mode == LCK_PW ? CLM_WRITE : CLM_READ;
1050         descr->cld_enq_flags = enqflags;
1051         io->ci_obj = obj;
1052
1053         lck = cl_lock_request(env, io, descr, "ec enqueue", eco);
1054         if (lck) {
1055                 struct echo_client_obd *ec = eco->eo_dev->ed_ec;
1056                 struct echo_lock *el;
1057
1058                 rc = cl_wait(env, lck);
1059                 if (rc == 0) {
1060                         el = cl2echo_lock(cl_lock_at(lck, &echo_device_type));
1061                         spin_lock(&ec->ec_lock);
1062                         if (list_empty(&el->el_chain)) {
1063                                 list_add(&el->el_chain, &ec->ec_locks);
1064                                 el->el_cookie = ++ec->ec_unique;
1065                         }
1066                         atomic_inc(&el->el_refcount);
1067                         *cookie = el->el_cookie;
1068                         spin_unlock(&ec->ec_lock);
1069                 } else {
1070                         cl_lock_release(env, lck, "ec enqueue", current);
1071                 }
1072         }
1073         return rc;
1074 }
1075
1076 static int cl_echo_enqueue(struct echo_object *eco, u64 start, u64 end,
1077                            int mode, __u64 *cookie)
1078 {
1079         struct echo_thread_info *info;
1080         struct lu_env *env;
1081         struct cl_io *io;
1082         int refcheck;
1083         int result;
1084
1085         env = cl_env_get(&refcheck);
1086         if (IS_ERR(env))
1087                 return PTR_ERR(env);
1088
1089         info = echo_env_info(env);
1090         io = &info->eti_io;
1091
1092         io->ci_ignore_layout = 1;
1093         result = cl_io_init(env, io, CIT_MISC, echo_obj2cl(eco));
1094         if (result < 0)
1095                 goto out;
1096         LASSERT(result == 0);
1097
1098         result = cl_echo_enqueue0(env, eco, start, end, mode, cookie, 0);
1099         cl_io_fini(env, io);
1100
1101 out:
1102         cl_env_put(env, &refcheck);
1103         return result;
1104 }
1105
1106 static int cl_echo_cancel0(struct lu_env *env, struct echo_device *ed,
1107                            __u64 cookie)
1108 {
1109         struct echo_client_obd *ec = ed->ed_ec;
1110         struct echo_lock       *ecl = NULL;
1111         struct list_head             *el;
1112         int found = 0, still_used = 0;
1113
1114         LASSERT(ec != NULL);
1115         spin_lock(&ec->ec_lock);
1116         list_for_each(el, &ec->ec_locks) {
1117                 ecl = list_entry(el, struct echo_lock, el_chain);
1118                 CDEBUG(D_INFO, "ecl: %p, cookie: %#llx\n", ecl, ecl->el_cookie);
1119                 found = (ecl->el_cookie == cookie);
1120                 if (found) {
1121                         if (atomic_dec_and_test(&ecl->el_refcount))
1122                                 list_del_init(&ecl->el_chain);
1123                         else
1124                                 still_used = 1;
1125                         break;
1126                 }
1127         }
1128         spin_unlock(&ec->ec_lock);
1129
1130         if (!found)
1131                 return -ENOENT;
1132
1133         echo_lock_release(env, ecl, still_used);
1134         return 0;
1135 }
1136
1137 static int cl_echo_cancel(struct echo_device *ed, __u64 cookie)
1138 {
1139         struct lu_env *env;
1140         int refcheck;
1141         int rc;
1142
1143         env = cl_env_get(&refcheck);
1144         if (IS_ERR(env))
1145                 return PTR_ERR(env);
1146
1147         rc = cl_echo_cancel0(env, ed, cookie);
1148
1149         cl_env_put(env, &refcheck);
1150         return rc;
1151 }
1152
1153 static int cl_echo_async_brw(const struct lu_env *env, struct cl_io *io,
1154                              enum cl_req_type unused, struct cl_2queue *queue)
1155 {
1156         struct cl_page *clp;
1157         struct cl_page *temp;
1158         int result = 0;
1159
1160         cl_page_list_for_each_safe(clp, temp, &queue->c2_qin) {
1161                 int rc;
1162
1163                 rc = cl_page_cache_add(env, io, clp, CRT_WRITE);
1164                 if (rc == 0)
1165                         continue;
1166                 result = result ?: rc;
1167         }
1168         return result;
1169 }
1170
1171 static int cl_echo_object_brw(struct echo_object *eco, int rw, u64 offset,
1172                               struct page **pages, int npages, int async)
1173 {
1174         struct lu_env      *env;
1175         struct echo_thread_info *info;
1176         struct cl_object        *obj = echo_obj2cl(eco);
1177         struct echo_device      *ed  = eco->eo_dev;
1178         struct cl_2queue        *queue;
1179         struct cl_io        *io;
1180         struct cl_page    *clp;
1181         struct lustre_handle    lh = { 0 };
1182         int page_size = cl_page_size(obj);
1183         int refcheck;
1184         int rc;
1185         int i;
1186
1187         LASSERT((offset & ~CFS_PAGE_MASK) == 0);
1188         LASSERT(ed->ed_next != NULL);
1189         env = cl_env_get(&refcheck);
1190         if (IS_ERR(env))
1191                 return PTR_ERR(env);
1192
1193         info    = echo_env_info(env);
1194         io      = &info->eti_io;
1195         queue   = &info->eti_queue;
1196
1197         cl_2queue_init(queue);
1198
1199         io->ci_ignore_layout = 1;
1200         rc = cl_io_init(env, io, CIT_MISC, obj);
1201         if (rc < 0)
1202                 goto out;
1203         LASSERT(rc == 0);
1204
1205
1206         rc = cl_echo_enqueue0(env, eco, offset,
1207                               offset + npages * PAGE_CACHE_SIZE - 1,
1208                               rw == READ ? LCK_PR : LCK_PW, &lh.cookie,
1209                               CEF_NEVER);
1210         if (rc < 0)
1211                 goto error_lock;
1212
1213         for (i = 0; i < npages; i++) {
1214                 LASSERT(pages[i]);
1215                 clp = cl_page_find(env, obj, cl_index(obj, offset),
1216                                    pages[i], CPT_TRANSIENT);
1217                 if (IS_ERR(clp)) {
1218                         rc = PTR_ERR(clp);
1219                         break;
1220                 }
1221                 LASSERT(clp->cp_type == CPT_TRANSIENT);
1222
1223                 rc = cl_page_own(env, io, clp);
1224                 if (rc) {
1225                         LASSERT(clp->cp_state == CPS_FREEING);
1226                         cl_page_put(env, clp);
1227                         break;
1228                 }
1229
1230                 cl_2queue_add(queue, clp);
1231
1232                 /* drop the reference count for cl_page_find, so that the page
1233                  * will be freed in cl_2queue_fini. */
1234                 cl_page_put(env, clp);
1235                 cl_page_clip(env, clp, 0, page_size);
1236
1237                 offset += page_size;
1238         }
1239
1240         if (rc == 0) {
1241                 enum cl_req_type typ = rw == READ ? CRT_READ : CRT_WRITE;
1242
1243                 async = async && (typ == CRT_WRITE);
1244                 if (async)
1245                         rc = cl_echo_async_brw(env, io, typ, queue);
1246                 else
1247                         rc = cl_io_submit_sync(env, io, typ, queue, 0);
1248                 CDEBUG(D_INFO, "echo_client %s write returns %d\n",
1249                        async ? "async" : "sync", rc);
1250         }
1251
1252         cl_echo_cancel0(env, ed, lh.cookie);
1253 error_lock:
1254         cl_2queue_discard(env, io, queue);
1255         cl_2queue_disown(env, io, queue);
1256         cl_2queue_fini(env, queue);
1257         cl_io_fini(env, io);
1258 out:
1259         cl_env_put(env, &refcheck);
1260         return rc;
1261 }
1262 /** @} echo_exports */
1263
1264
1265 static u64 last_object_id;
1266
1267 static int
1268 echo_copyout_lsm(struct lov_stripe_md *lsm, void *_ulsm, int ulsm_nob)
1269 {
1270         struct lov_stripe_md *ulsm = _ulsm;
1271         int nob, i;
1272
1273         nob = offsetof(struct lov_stripe_md, lsm_oinfo[lsm->lsm_stripe_count]);
1274         if (nob > ulsm_nob)
1275                 return -EINVAL;
1276
1277         if (copy_to_user(ulsm, lsm, sizeof(*ulsm)))
1278                 return -EFAULT;
1279
1280         for (i = 0; i < lsm->lsm_stripe_count; i++) {
1281                 if (copy_to_user(ulsm->lsm_oinfo[i], lsm->lsm_oinfo[i],
1282                                       sizeof(lsm->lsm_oinfo[0])))
1283                         return -EFAULT;
1284         }
1285         return 0;
1286 }
1287
1288 static int
1289 echo_copyin_lsm(struct echo_device *ed, struct lov_stripe_md *lsm,
1290                  void *ulsm, int ulsm_nob)
1291 {
1292         struct echo_client_obd *ec = ed->ed_ec;
1293         int                  i;
1294
1295         if (ulsm_nob < sizeof(*lsm))
1296                 return -EINVAL;
1297
1298         if (copy_from_user(lsm, ulsm, sizeof(*lsm)))
1299                 return -EFAULT;
1300
1301         if (lsm->lsm_stripe_count > ec->ec_nstripes ||
1302             lsm->lsm_magic != LOV_MAGIC ||
1303             (lsm->lsm_stripe_size & (~CFS_PAGE_MASK)) != 0 ||
1304             ((__u64)lsm->lsm_stripe_size * lsm->lsm_stripe_count > ~0UL))
1305                 return -EINVAL;
1306
1307
1308         for (i = 0; i < lsm->lsm_stripe_count; i++) {
1309                 if (copy_from_user(lsm->lsm_oinfo[i],
1310                                        ((struct lov_stripe_md *)ulsm)-> \
1311                                        lsm_oinfo[i],
1312                                        sizeof(lsm->lsm_oinfo[0])))
1313                         return -EFAULT;
1314         }
1315         return 0;
1316 }
1317
1318 static int echo_create_object(const struct lu_env *env, struct echo_device *ed,
1319                               int on_target, struct obdo *oa, void *ulsm,
1320                               int ulsm_nob, struct obd_trans_info *oti)
1321 {
1322         struct echo_object     *eco;
1323         struct echo_client_obd *ec = ed->ed_ec;
1324         struct lov_stripe_md   *lsm = NULL;
1325         int                  rc;
1326         int                  created = 0;
1327
1328         if ((oa->o_valid & OBD_MD_FLID) == 0 && /* no obj id */
1329             (on_target ||                      /* set_stripe */
1330              ec->ec_nstripes != 0)) {      /* LOV */
1331                 CERROR("No valid oid\n");
1332                 return -EINVAL;
1333         }
1334
1335         rc = echo_alloc_memmd(ed, &lsm);
1336         if (rc < 0) {
1337                 CERROR("Cannot allocate md: rc = %d\n", rc);
1338                 goto failed;
1339         }
1340
1341         if (ulsm != NULL) {
1342                 int i, idx;
1343
1344                 rc = echo_copyin_lsm(ed, lsm, ulsm, ulsm_nob);
1345                 if (rc != 0)
1346                         goto failed;
1347
1348                 if (lsm->lsm_stripe_count == 0)
1349                         lsm->lsm_stripe_count = ec->ec_nstripes;
1350
1351                 if (lsm->lsm_stripe_size == 0)
1352                         lsm->lsm_stripe_size = PAGE_CACHE_SIZE;
1353
1354                 idx = cfs_rand();
1355
1356                 /* setup stripes: indices + default ids if required */
1357                 for (i = 0; i < lsm->lsm_stripe_count; i++) {
1358                         if (ostid_id(&lsm->lsm_oinfo[i]->loi_oi) == 0)
1359                                 lsm->lsm_oinfo[i]->loi_oi = lsm->lsm_oi;
1360
1361                         lsm->lsm_oinfo[i]->loi_ost_idx =
1362                                 (idx + i) % ec->ec_nstripes;
1363                 }
1364         }
1365
1366         /* setup object ID here for !on_target and LOV hint */
1367         if (oa->o_valid & OBD_MD_FLID) {
1368                 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
1369                 lsm->lsm_oi = oa->o_oi;
1370         }
1371
1372         if (ostid_id(&lsm->lsm_oi) == 0)
1373                 ostid_set_id(&lsm->lsm_oi, ++last_object_id);
1374
1375         rc = 0;
1376         if (on_target) {
1377                 /* Only echo objects are allowed to be created */
1378                 LASSERT((oa->o_valid & OBD_MD_FLGROUP) &&
1379                         (ostid_seq(&oa->o_oi) == FID_SEQ_ECHO));
1380                 rc = obd_create(env, ec->ec_exp, oa, &lsm, oti);
1381                 if (rc != 0) {
1382                         CERROR("Cannot create objects: rc = %d\n", rc);
1383                         goto failed;
1384                 }
1385                 created = 1;
1386         }
1387
1388         /* See what object ID we were given */
1389         oa->o_oi = lsm->lsm_oi;
1390         oa->o_valid |= OBD_MD_FLID;
1391
1392         eco = cl_echo_object_find(ed, &lsm);
1393         if (IS_ERR(eco)) {
1394                 rc = PTR_ERR(eco);
1395                 goto failed;
1396         }
1397         cl_echo_object_put(eco);
1398
1399         CDEBUG(D_INFO, "oa oid "DOSTID"\n", POSTID(&oa->o_oi));
1400
1401  failed:
1402         if (created && rc)
1403                 obd_destroy(env, ec->ec_exp, oa, lsm, oti, NULL, NULL);
1404         if (lsm)
1405                 echo_free_memmd(ed, &lsm);
1406         if (rc)
1407                 CERROR("create object failed with: rc = %d\n", rc);
1408         return rc;
1409 }
1410
1411 static int echo_get_object(struct echo_object **ecop, struct echo_device *ed,
1412                            struct obdo *oa)
1413 {
1414         struct lov_stripe_md   *lsm = NULL;
1415         struct echo_object     *eco;
1416         int                  rc;
1417
1418         if ((oa->o_valid & OBD_MD_FLID) == 0 || ostid_id(&oa->o_oi) == 0) {
1419                 /* disallow use of object id 0 */
1420                 CERROR("No valid oid\n");
1421                 return -EINVAL;
1422         }
1423
1424         rc = echo_alloc_memmd(ed, &lsm);
1425         if (rc < 0)
1426                 return rc;
1427
1428         lsm->lsm_oi = oa->o_oi;
1429         if (!(oa->o_valid & OBD_MD_FLGROUP))
1430                 ostid_set_seq_echo(&lsm->lsm_oi);
1431
1432         rc = 0;
1433         eco = cl_echo_object_find(ed, &lsm);
1434         if (!IS_ERR(eco))
1435                 *ecop = eco;
1436         else
1437                 rc = PTR_ERR(eco);
1438         if (lsm)
1439                 echo_free_memmd(ed, &lsm);
1440         return rc;
1441 }
1442
1443 static void echo_put_object(struct echo_object *eco)
1444 {
1445         if (cl_echo_object_put(eco))
1446                 CERROR("echo client: drop an object failed");
1447 }
1448
1449 static void
1450 echo_get_stripe_off_id(struct lov_stripe_md *lsm, u64 *offp, u64 *idp)
1451 {
1452         unsigned long stripe_count;
1453         unsigned long stripe_size;
1454         unsigned long width;
1455         unsigned long woffset;
1456         int        stripe_index;
1457         u64       offset;
1458
1459         if (lsm->lsm_stripe_count <= 1)
1460                 return;
1461
1462         offset       = *offp;
1463         stripe_size  = lsm->lsm_stripe_size;
1464         stripe_count = lsm->lsm_stripe_count;
1465
1466         /* width = # bytes in all stripes */
1467         width = stripe_size * stripe_count;
1468
1469         /* woffset = offset within a width; offset = whole number of widths */
1470         woffset = do_div(offset, width);
1471
1472         stripe_index = woffset / stripe_size;
1473
1474         *idp = ostid_id(&lsm->lsm_oinfo[stripe_index]->loi_oi);
1475         *offp = offset * stripe_size + woffset % stripe_size;
1476 }
1477
1478 static void
1479 echo_client_page_debug_setup(struct lov_stripe_md *lsm,
1480                              struct page *page, int rw, u64 id,
1481                              u64 offset, u64 count)
1482 {
1483         char    *addr;
1484         u64      stripe_off;
1485         u64      stripe_id;
1486         int      delta;
1487
1488         /* no partial pages on the client */
1489         LASSERT(count == PAGE_CACHE_SIZE);
1490
1491         addr = kmap(page);
1492
1493         for (delta = 0; delta < PAGE_CACHE_SIZE; delta += OBD_ECHO_BLOCK_SIZE) {
1494                 if (rw == OBD_BRW_WRITE) {
1495                         stripe_off = offset + delta;
1496                         stripe_id = id;
1497                         echo_get_stripe_off_id(lsm, &stripe_off, &stripe_id);
1498                 } else {
1499                         stripe_off = 0xdeadbeef00c0ffeeULL;
1500                         stripe_id = 0xdeadbeef00c0ffeeULL;
1501                 }
1502                 block_debug_setup(addr + delta, OBD_ECHO_BLOCK_SIZE,
1503                                   stripe_off, stripe_id);
1504         }
1505
1506         kunmap(page);
1507 }
1508
1509 static int echo_client_page_debug_check(struct lov_stripe_md *lsm,
1510                                         struct page *page, u64 id,
1511                                         u64 offset, u64 count)
1512 {
1513         u64     stripe_off;
1514         u64     stripe_id;
1515         char   *addr;
1516         int     delta;
1517         int     rc;
1518         int     rc2;
1519
1520         /* no partial pages on the client */
1521         LASSERT(count == PAGE_CACHE_SIZE);
1522
1523         addr = kmap(page);
1524
1525         for (rc = delta = 0; delta < PAGE_CACHE_SIZE; delta += OBD_ECHO_BLOCK_SIZE) {
1526                 stripe_off = offset + delta;
1527                 stripe_id = id;
1528                 echo_get_stripe_off_id(lsm, &stripe_off, &stripe_id);
1529
1530                 rc2 = block_debug_check("test_brw",
1531                                         addr + delta, OBD_ECHO_BLOCK_SIZE,
1532                                         stripe_off, stripe_id);
1533                 if (rc2 != 0) {
1534                         CERROR("Error in echo object %#llx\n", id);
1535                         rc = rc2;
1536                 }
1537         }
1538
1539         kunmap(page);
1540         return rc;
1541 }
1542
1543 static int echo_client_kbrw(struct echo_device *ed, int rw, struct obdo *oa,
1544                             struct echo_object *eco, u64 offset,
1545                             u64 count, int async,
1546                             struct obd_trans_info *oti)
1547 {
1548         struct lov_stripe_md   *lsm = eco->eo_lsm;
1549         u32            npages;
1550         struct brw_page *pga;
1551         struct brw_page *pgp;
1552         struct page         **pages;
1553         u64              off;
1554         int                  i;
1555         int                  rc;
1556         int                  verify;
1557         gfp_t                gfp_mask;
1558         int                  brw_flags = 0;
1559
1560         verify = (ostid_id(&oa->o_oi) != ECHO_PERSISTENT_OBJID &&
1561                   (oa->o_valid & OBD_MD_FLFLAGS) != 0 &&
1562                   (oa->o_flags & OBD_FL_DEBUG_CHECK) != 0);
1563
1564         gfp_mask = ((ostid_id(&oa->o_oi) & 2) == 0) ? GFP_IOFS : GFP_HIGHUSER;
1565
1566         LASSERT(rw == OBD_BRW_WRITE || rw == OBD_BRW_READ);
1567         LASSERT(lsm != NULL);
1568         LASSERT(ostid_id(&lsm->lsm_oi) == ostid_id(&oa->o_oi));
1569
1570         if (count <= 0 ||
1571             (count & (~CFS_PAGE_MASK)) != 0)
1572                 return -EINVAL;
1573
1574         /* XXX think again with misaligned I/O */
1575         npages = count >> PAGE_CACHE_SHIFT;
1576
1577         if (rw == OBD_BRW_WRITE)
1578                 brw_flags = OBD_BRW_ASYNC;
1579
1580         OBD_ALLOC(pga, npages * sizeof(*pga));
1581         if (pga == NULL)
1582                 return -ENOMEM;
1583
1584         OBD_ALLOC(pages, npages * sizeof(*pages));
1585         if (pages == NULL) {
1586                 OBD_FREE(pga, npages * sizeof(*pga));
1587                 return -ENOMEM;
1588         }
1589
1590         for (i = 0, pgp = pga, off = offset;
1591              i < npages;
1592              i++, pgp++, off += PAGE_CACHE_SIZE) {
1593
1594                 LASSERT(pgp->pg == NULL);      /* for cleanup */
1595
1596                 rc = -ENOMEM;
1597                 OBD_PAGE_ALLOC(pgp->pg, gfp_mask);
1598                 if (pgp->pg == NULL)
1599                         goto out;
1600
1601                 pages[i] = pgp->pg;
1602                 pgp->count = PAGE_CACHE_SIZE;
1603                 pgp->off = off;
1604                 pgp->flag = brw_flags;
1605
1606                 if (verify)
1607                         echo_client_page_debug_setup(lsm, pgp->pg, rw,
1608                                                      ostid_id(&oa->o_oi), off,
1609                                                      pgp->count);
1610         }
1611
1612         /* brw mode can only be used at client */
1613         LASSERT(ed->ed_next != NULL);
1614         rc = cl_echo_object_brw(eco, rw, offset, pages, npages, async);
1615
1616  out:
1617         if (rc != 0 || rw != OBD_BRW_READ)
1618                 verify = 0;
1619
1620         for (i = 0, pgp = pga; i < npages; i++, pgp++) {
1621                 if (pgp->pg == NULL)
1622                         continue;
1623
1624                 if (verify) {
1625                         int vrc;
1626
1627                         vrc = echo_client_page_debug_check(lsm, pgp->pg,
1628                                                            ostid_id(&oa->o_oi),
1629                                                            pgp->off, pgp->count);
1630                         if (vrc != 0 && rc == 0)
1631                                 rc = vrc;
1632                 }
1633                 OBD_PAGE_FREE(pgp->pg);
1634         }
1635         OBD_FREE(pga, npages * sizeof(*pga));
1636         OBD_FREE(pages, npages * sizeof(*pages));
1637         return rc;
1638 }
1639
1640 static int echo_client_prep_commit(const struct lu_env *env,
1641                                    struct obd_export *exp, int rw,
1642                                    struct obdo *oa, struct echo_object *eco,
1643                                    u64 offset, u64 count,
1644                                    u64 batch, struct obd_trans_info *oti,
1645                                    int async)
1646 {
1647         struct lov_stripe_md *lsm = eco->eo_lsm;
1648         struct obd_ioobj ioo;
1649         struct niobuf_local *lnb;
1650         struct niobuf_remote *rnb;
1651         u64 off;
1652         u64 npages, tot_pages;
1653         int i, ret = 0, brw_flags = 0;
1654
1655         if (count <= 0 || (count & (~CFS_PAGE_MASK)) != 0 ||
1656             (lsm != NULL && ostid_id(&lsm->lsm_oi) != ostid_id(&oa->o_oi)))
1657                 return -EINVAL;
1658
1659         npages = batch >> PAGE_CACHE_SHIFT;
1660         tot_pages = count >> PAGE_CACHE_SHIFT;
1661
1662         OBD_ALLOC(lnb, npages * sizeof(struct niobuf_local));
1663         OBD_ALLOC(rnb, npages * sizeof(struct niobuf_remote));
1664
1665         if (lnb == NULL || rnb == NULL) {
1666                 ret = -ENOMEM;
1667                 goto out;
1668         }
1669
1670         if (rw == OBD_BRW_WRITE && async)
1671                 brw_flags |= OBD_BRW_ASYNC;
1672
1673         obdo_to_ioobj(oa, &ioo);
1674
1675         off = offset;
1676
1677         for (; tot_pages; tot_pages -= npages) {
1678                 int lpages;
1679
1680                 if (tot_pages < npages)
1681                         npages = tot_pages;
1682
1683                 for (i = 0; i < npages; i++, off += PAGE_CACHE_SIZE) {
1684                         rnb[i].offset = off;
1685                         rnb[i].len = PAGE_CACHE_SIZE;
1686                         rnb[i].flags = brw_flags;
1687                 }
1688
1689                 ioo.ioo_bufcnt = npages;
1690                 oti->oti_transno = 0;
1691
1692                 lpages = npages;
1693                 ret = obd_preprw(env, rw, exp, oa, 1, &ioo, rnb, &lpages,
1694                                  lnb, oti, NULL);
1695                 if (ret != 0)
1696                         goto out;
1697                 LASSERT(lpages == npages);
1698
1699                 for (i = 0; i < lpages; i++) {
1700                         struct page *page = lnb[i].page;
1701
1702                         /* read past eof? */
1703                         if (page == NULL && lnb[i].rc == 0)
1704                                 continue;
1705
1706                         if (async)
1707                                 lnb[i].flags |= OBD_BRW_ASYNC;
1708
1709                         if (ostid_id(&oa->o_oi) == ECHO_PERSISTENT_OBJID ||
1710                             (oa->o_valid & OBD_MD_FLFLAGS) == 0 ||
1711                             (oa->o_flags & OBD_FL_DEBUG_CHECK) == 0)
1712                                 continue;
1713
1714                         if (rw == OBD_BRW_WRITE)
1715                                 echo_client_page_debug_setup(lsm, page, rw,
1716                                                             ostid_id(&oa->o_oi),
1717                                                              rnb[i].offset,
1718                                                              rnb[i].len);
1719                         else
1720                                 echo_client_page_debug_check(lsm, page,
1721                                                             ostid_id(&oa->o_oi),
1722                                                              rnb[i].offset,
1723                                                              rnb[i].len);
1724                 }
1725
1726                 ret = obd_commitrw(env, rw, exp, oa, 1, &ioo,
1727                                    rnb, npages, lnb, oti, ret);
1728                 if (ret != 0)
1729                         goto out;
1730
1731                 /* Reset oti otherwise it would confuse ldiskfs. */
1732                 memset(oti, 0, sizeof(*oti));
1733
1734                 /* Reuse env context. */
1735                 lu_context_exit((struct lu_context *)&env->le_ctx);
1736                 lu_context_enter((struct lu_context *)&env->le_ctx);
1737         }
1738
1739 out:
1740         if (lnb)
1741                 OBD_FREE(lnb, npages * sizeof(struct niobuf_local));
1742         if (rnb)
1743                 OBD_FREE(rnb, npages * sizeof(struct niobuf_remote));
1744         return ret;
1745 }
1746
1747 static int echo_client_brw_ioctl(const struct lu_env *env, int rw,
1748                                  struct obd_export *exp,
1749                                  struct obd_ioctl_data *data,
1750                                  struct obd_trans_info *dummy_oti)
1751 {
1752         struct obd_device *obd = class_exp2obd(exp);
1753         struct echo_device *ed = obd2echo_dev(obd);
1754         struct echo_client_obd *ec = ed->ed_ec;
1755         struct obdo *oa = &data->ioc_obdo1;
1756         struct echo_object *eco;
1757         int rc;
1758         int async = 1;
1759         long test_mode;
1760
1761         LASSERT(oa->o_valid & OBD_MD_FLGROUP);
1762
1763         rc = echo_get_object(&eco, ed, oa);
1764         if (rc)
1765                 return rc;
1766
1767         oa->o_valid &= ~OBD_MD_FLHANDLE;
1768
1769         /* OFD/obdfilter works only via prep/commit */
1770         test_mode = (long)data->ioc_pbuf1;
1771         if (test_mode == 1)
1772                 async = 0;
1773
1774         if (ed->ed_next == NULL && test_mode != 3) {
1775                 test_mode = 3;
1776                 data->ioc_plen1 = data->ioc_count;
1777         }
1778
1779         /* Truncate batch size to maximum */
1780         if (data->ioc_plen1 > PTLRPC_MAX_BRW_SIZE)
1781                 data->ioc_plen1 = PTLRPC_MAX_BRW_SIZE;
1782
1783         switch (test_mode) {
1784         case 1:
1785                 /* fall through */
1786         case 2:
1787                 rc = echo_client_kbrw(ed, rw, oa,
1788                                       eco, data->ioc_offset,
1789                                       data->ioc_count, async, dummy_oti);
1790                 break;
1791         case 3:
1792                 rc = echo_client_prep_commit(env, ec->ec_exp, rw, oa,
1793                                              eco, data->ioc_offset,
1794                                              data->ioc_count, data->ioc_plen1,
1795                                              dummy_oti, async);
1796                 break;
1797         default:
1798                 rc = -EINVAL;
1799         }
1800         echo_put_object(eco);
1801         return rc;
1802 }
1803
1804 static int
1805 echo_client_enqueue(struct obd_export *exp, struct obdo *oa,
1806                     int mode, u64 offset, u64 nob)
1807 {
1808         struct echo_device     *ed = obd2echo_dev(exp->exp_obd);
1809         struct lustre_handle   *ulh = &oa->o_handle;
1810         struct echo_object     *eco;
1811         u64              end;
1812         int                  rc;
1813
1814         if (ed->ed_next == NULL)
1815                 return -EOPNOTSUPP;
1816
1817         if (!(mode == LCK_PR || mode == LCK_PW))
1818                 return -EINVAL;
1819
1820         if ((offset & (~CFS_PAGE_MASK)) != 0 ||
1821             (nob & (~CFS_PAGE_MASK)) != 0)
1822                 return -EINVAL;
1823
1824         rc = echo_get_object(&eco, ed, oa);
1825         if (rc != 0)
1826                 return rc;
1827
1828         end = (nob == 0) ? ((u64) -1) : (offset + nob - 1);
1829         rc = cl_echo_enqueue(eco, offset, end, mode, &ulh->cookie);
1830         if (rc == 0) {
1831                 oa->o_valid |= OBD_MD_FLHANDLE;
1832                 CDEBUG(D_INFO, "Cookie is %#llx\n", ulh->cookie);
1833         }
1834         echo_put_object(eco);
1835         return rc;
1836 }
1837
1838 static int
1839 echo_client_cancel(struct obd_export *exp, struct obdo *oa)
1840 {
1841         struct echo_device *ed     = obd2echo_dev(exp->exp_obd);
1842         __u64          cookie = oa->o_handle.cookie;
1843
1844         if ((oa->o_valid & OBD_MD_FLHANDLE) == 0)
1845                 return -EINVAL;
1846
1847         CDEBUG(D_INFO, "Cookie is %#llx\n", cookie);
1848         return cl_echo_cancel(ed, cookie);
1849 }
1850
1851 static int
1852 echo_client_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
1853                       void *karg, void *uarg)
1854 {
1855         struct obd_device      *obd = exp->exp_obd;
1856         struct echo_device     *ed = obd2echo_dev(obd);
1857         struct echo_client_obd *ec = ed->ed_ec;
1858         struct echo_object     *eco;
1859         struct obd_ioctl_data  *data = karg;
1860         struct obd_trans_info   dummy_oti;
1861         struct lu_env     *env;
1862         struct oti_req_ack_lock *ack_lock;
1863         struct obdo         *oa;
1864         struct lu_fid      fid;
1865         int                  rw = OBD_BRW_READ;
1866         int                  rc = 0;
1867         int                  i;
1868
1869         memset(&dummy_oti, 0, sizeof(dummy_oti));
1870
1871         oa = &data->ioc_obdo1;
1872         if (!(oa->o_valid & OBD_MD_FLGROUP)) {
1873                 oa->o_valid |= OBD_MD_FLGROUP;
1874                 ostid_set_seq_echo(&oa->o_oi);
1875         }
1876
1877         /* This FID is unpacked just for validation at this point */
1878         rc = ostid_to_fid(&fid, &oa->o_oi, 0);
1879         if (rc < 0)
1880                 return rc;
1881
1882         OBD_ALLOC_PTR(env);
1883         if (env == NULL)
1884                 return -ENOMEM;
1885
1886         rc = lu_env_init(env, LCT_DT_THREAD);
1887         if (rc) {
1888                 rc = -ENOMEM;
1889                 goto out;
1890         }
1891
1892         switch (cmd) {
1893         case OBD_IOC_CREATE:                /* may create echo object */
1894                 if (!capable(CFS_CAP_SYS_ADMIN)) {
1895                         rc = -EPERM;
1896                         goto out;
1897                 }
1898
1899                 rc = echo_create_object(env, ed, 1, oa, data->ioc_pbuf1,
1900                                         data->ioc_plen1, &dummy_oti);
1901                 goto out;
1902
1903         case OBD_IOC_DESTROY:
1904                 if (!capable(CFS_CAP_SYS_ADMIN)) {
1905                         rc = -EPERM;
1906                         goto out;
1907                 }
1908
1909                 rc = echo_get_object(&eco, ed, oa);
1910                 if (rc == 0) {
1911                         rc = obd_destroy(env, ec->ec_exp, oa, eco->eo_lsm,
1912                                          &dummy_oti, NULL, NULL);
1913                         if (rc == 0)
1914                                 eco->eo_deleted = 1;
1915                         echo_put_object(eco);
1916                 }
1917                 goto out;
1918
1919         case OBD_IOC_GETATTR:
1920                 rc = echo_get_object(&eco, ed, oa);
1921                 if (rc == 0) {
1922                         struct obd_info oinfo = { { { 0 } } };
1923
1924                         oinfo.oi_md = eco->eo_lsm;
1925                         oinfo.oi_oa = oa;
1926                         rc = obd_getattr(env, ec->ec_exp, &oinfo);
1927                         echo_put_object(eco);
1928                 }
1929                 goto out;
1930
1931         case OBD_IOC_SETATTR:
1932                 if (!capable(CFS_CAP_SYS_ADMIN)) {
1933                         rc = -EPERM;
1934                         goto out;
1935                 }
1936
1937                 rc = echo_get_object(&eco, ed, oa);
1938                 if (rc == 0) {
1939                         struct obd_info oinfo = { { { 0 } } };
1940
1941                         oinfo.oi_oa = oa;
1942                         oinfo.oi_md = eco->eo_lsm;
1943
1944                         rc = obd_setattr(env, ec->ec_exp, &oinfo, NULL);
1945                         echo_put_object(eco);
1946                 }
1947                 goto out;
1948
1949         case OBD_IOC_BRW_WRITE:
1950                 if (!capable(CFS_CAP_SYS_ADMIN)) {
1951                         rc = -EPERM;
1952                         goto out;
1953                 }
1954
1955                 rw = OBD_BRW_WRITE;
1956                 /* fall through */
1957         case OBD_IOC_BRW_READ:
1958                 rc = echo_client_brw_ioctl(env, rw, exp, data, &dummy_oti);
1959                 goto out;
1960
1961         case ECHO_IOC_GET_STRIPE:
1962                 rc = echo_get_object(&eco, ed, oa);
1963                 if (rc == 0) {
1964                         rc = echo_copyout_lsm(eco->eo_lsm, data->ioc_pbuf1,
1965                                               data->ioc_plen1);
1966                         echo_put_object(eco);
1967                 }
1968                 goto out;
1969
1970         case ECHO_IOC_SET_STRIPE:
1971                 if (!capable(CFS_CAP_SYS_ADMIN)) {
1972                         rc = -EPERM;
1973                         goto out;
1974                 }
1975
1976                 if (data->ioc_pbuf1 == NULL) {  /* unset */
1977                         rc = echo_get_object(&eco, ed, oa);
1978                         if (rc == 0) {
1979                                 eco->eo_deleted = 1;
1980                                 echo_put_object(eco);
1981                         }
1982                 } else {
1983                         rc = echo_create_object(env, ed, 0, oa,
1984                                                 data->ioc_pbuf1,
1985                                                 data->ioc_plen1, &dummy_oti);
1986                 }
1987                 goto out;
1988
1989         case ECHO_IOC_ENQUEUE:
1990                 if (!capable(CFS_CAP_SYS_ADMIN)) {
1991                         rc = -EPERM;
1992                         goto out;
1993                 }
1994
1995                 rc = echo_client_enqueue(exp, oa,
1996                                          data->ioc_conn1, /* lock mode */
1997                                          data->ioc_offset,
1998                                          data->ioc_count);/*extent*/
1999                 goto out;
2000
2001         case ECHO_IOC_CANCEL:
2002                 rc = echo_client_cancel(exp, oa);
2003                 goto out;
2004
2005         default:
2006                 CERROR("echo_ioctl(): unrecognised ioctl %#x\n", cmd);
2007                 rc = -ENOTTY;
2008                 goto out;
2009         }
2010
2011 out:
2012         lu_env_fini(env);
2013         OBD_FREE_PTR(env);
2014
2015         /* XXX this should be in a helper also called by target_send_reply */
2016         for (ack_lock = dummy_oti.oti_ack_locks, i = 0; i < 4;
2017              i++, ack_lock++) {
2018                 if (!ack_lock->mode)
2019                         break;
2020                 ldlm_lock_decref(&ack_lock->lock, ack_lock->mode);
2021         }
2022
2023         return rc;
2024 }
2025
2026 static int echo_client_setup(const struct lu_env *env,
2027                              struct obd_device *obddev, struct lustre_cfg *lcfg)
2028 {
2029         struct echo_client_obd *ec = &obddev->u.echo_client;
2030         struct obd_device *tgt;
2031         struct obd_uuid echo_uuid = { "ECHO_UUID" };
2032         struct obd_connect_data *ocd = NULL;
2033         int rc;
2034
2035         if (lcfg->lcfg_bufcount < 2 || LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
2036                 CERROR("requires a TARGET OBD name\n");
2037                 return -EINVAL;
2038         }
2039
2040         tgt = class_name2obd(lustre_cfg_string(lcfg, 1));
2041         if (!tgt || !tgt->obd_attached || !tgt->obd_set_up) {
2042                 CERROR("device not attached or not set up (%s)\n",
2043                        lustre_cfg_string(lcfg, 1));
2044                 return -EINVAL;
2045         }
2046
2047         spin_lock_init(&ec->ec_lock);
2048         INIT_LIST_HEAD(&ec->ec_objects);
2049         INIT_LIST_HEAD(&ec->ec_locks);
2050         ec->ec_unique = 0;
2051         ec->ec_nstripes = 0;
2052
2053         OBD_ALLOC(ocd, sizeof(*ocd));
2054         if (ocd == NULL) {
2055                 CERROR("Can't alloc ocd connecting to %s\n",
2056                        lustre_cfg_string(lcfg, 1));
2057                 return -ENOMEM;
2058         }
2059
2060         ocd->ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_REQPORTAL |
2061                                  OBD_CONNECT_BRW_SIZE |
2062                                  OBD_CONNECT_GRANT | OBD_CONNECT_FULL20 |
2063                                  OBD_CONNECT_64BITHASH | OBD_CONNECT_LVB_TYPE |
2064                                  OBD_CONNECT_FID;
2065         ocd->ocd_brw_size = DT_MAX_BRW_SIZE;
2066         ocd->ocd_version = LUSTRE_VERSION_CODE;
2067         ocd->ocd_group = FID_SEQ_ECHO;
2068
2069         rc = obd_connect(env, &ec->ec_exp, tgt, &echo_uuid, ocd, NULL);
2070         if (rc == 0) {
2071                 /* Turn off pinger because it connects to tgt obd directly. */
2072                 spin_lock(&tgt->obd_dev_lock);
2073                 list_del_init(&ec->ec_exp->exp_obd_chain_timed);
2074                 spin_unlock(&tgt->obd_dev_lock);
2075         }
2076
2077         OBD_FREE(ocd, sizeof(*ocd));
2078
2079         if (rc != 0) {
2080                 CERROR("fail to connect to device %s\n",
2081                        lustre_cfg_string(lcfg, 1));
2082                 return rc;
2083         }
2084
2085         return rc;
2086 }
2087
2088 static int echo_client_cleanup(struct obd_device *obddev)
2089 {
2090         struct echo_client_obd *ec = &obddev->u.echo_client;
2091         int rc;
2092
2093         if (!list_empty(&obddev->obd_exports)) {
2094                 CERROR("still has clients!\n");
2095                 return -EBUSY;
2096         }
2097
2098         LASSERT(atomic_read(&ec->ec_exp->exp_refcount) > 0);
2099         rc = obd_disconnect(ec->ec_exp);
2100         if (rc != 0)
2101                 CERROR("fail to disconnect device: %d\n", rc);
2102
2103         return rc;
2104 }
2105
2106 static int echo_client_connect(const struct lu_env *env,
2107                                struct obd_export **exp,
2108                                struct obd_device *src, struct obd_uuid *cluuid,
2109                                struct obd_connect_data *data, void *localdata)
2110 {
2111         int             rc;
2112         struct lustre_handle conn = { 0 };
2113
2114         rc = class_connect(&conn, src, cluuid);
2115         if (rc == 0) {
2116                 *exp = class_conn2export(&conn);
2117         }
2118
2119         return rc;
2120 }
2121
2122 static int echo_client_disconnect(struct obd_export *exp)
2123 {
2124         int                  rc;
2125
2126         if (exp == NULL) {
2127                 rc = -EINVAL;
2128                 goto out;
2129         }
2130
2131         rc = class_disconnect(exp);
2132         goto out;
2133  out:
2134         return rc;
2135 }
2136
2137 static struct obd_ops echo_client_obd_ops = {
2138         .o_owner       = THIS_MODULE,
2139         .o_iocontrol   = echo_client_iocontrol,
2140         .o_connect     = echo_client_connect,
2141         .o_disconnect  = echo_client_disconnect
2142 };
2143
2144 int echo_client_init(void)
2145 {
2146         struct lprocfs_static_vars lvars = { NULL };
2147         int rc;
2148
2149         lprocfs_echo_init_vars(&lvars);
2150
2151         rc = lu_kmem_init(echo_caches);
2152         if (rc == 0) {
2153                 rc = class_register_type(&echo_client_obd_ops, NULL,
2154                                          lvars.module_vars,
2155                                          LUSTRE_ECHO_CLIENT_NAME,
2156                                          &echo_device_type);
2157                 if (rc)
2158                         lu_kmem_fini(echo_caches);
2159         }
2160         return rc;
2161 }
2162
2163 void echo_client_exit(void)
2164 {
2165         class_unregister_type(LUSTRE_ECHO_CLIENT_NAME);
2166         lu_kmem_fini(echo_caches);
2167 }
2168
2169 static int __init obdecho_init(void)
2170 {
2171         struct lprocfs_static_vars lvars;
2172
2173         LCONSOLE_INFO("Echo OBD driver; http://www.lustre.org/\n");
2174
2175         LASSERT(PAGE_CACHE_SIZE % OBD_ECHO_BLOCK_SIZE == 0);
2176
2177         lprocfs_echo_init_vars(&lvars);
2178
2179
2180         return echo_client_init();
2181 }
2182
2183 static void /*__exit*/ obdecho_exit(void)
2184 {
2185         echo_client_exit();
2186
2187 }
2188
2189 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
2190 MODULE_DESCRIPTION("Lustre Testing Echo OBD driver");
2191 MODULE_LICENSE("GPL");
2192 MODULE_VERSION(LUSTRE_VERSION_STRING);
2193
2194 module_init(obdecho_init);
2195 module_exit(obdecho_exit);
2196
2197 /** @} echo_client */