drivers/staging/lustre/lustre/lmv/lmv_obd.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) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2011, 2015, 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_LMV
  38 #include <linux/slab.h>
  39 #include <linux/module.h>
  40 #include <linux/init.h>
  41 #include <linux/pagemap.h>
  42 #include <linux/mm.h>
  43 #include <asm/div64.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/namei.h>
  46 #include <linux/uaccess.h>
  47
  48 #include "../include/lustre/lustre_idl.h"
  49 #include "../include/obd_support.h"
  50 #include "../include/lustre_lib.h"
  51 #include "../include/lustre_net.h"
  52 #include "../include/obd_class.h"
  53 #include "../include/lprocfs_status.h"
  54 #include "../include/lustre_lite.h"
  55 #include "../include/lustre_fid.h"
  56 #include "../include/lustre_kernelcomm.h"
  57 #include "lmv_internal.h"
  58
  59 static void lmv_activate_target(struct lmv_obd *lmv,
  60                                 struct lmv_tgt_desc *tgt,
  61                                 int activate)
  62 {
  63         if (tgt->ltd_active == activate)
  64                 return;
  65
  66         tgt->ltd_active = activate;
  67         lmv->desc.ld_active_tgt_count += (activate ? 1 : -1);
  68 }
  69
  70 /**
  71  * Error codes:
  72  *
  73  *  -EINVAL  : UUID can't be found in the LMV's target list
  74  *  -ENOTCONN: The UUID is found, but the target connection is bad (!)
  75  *  -EBADF   : The UUID is found, but the OBD of the wrong type (!)
  76  */
  77 static int lmv_set_mdc_active(struct lmv_obd *lmv, struct obd_uuid *uuid,
  78                               int activate)
  79 {
  80         struct lmv_tgt_desc    *uninitialized_var(tgt);
  81         struct obd_device      *obd;
  82         int                  i;
  83         int                  rc = 0;
  84
  85         CDEBUG(D_INFO, "Searching in lmv %p for uuid %s (activate=%d)\n",
  86                lmv, uuid->uuid, activate);
  87
  88         spin_lock(&lmv->lmv_lock);
  89         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
  90                 tgt = lmv->tgts[i];
  91                 if (!tgt || !tgt->ltd_exp)
  92                         continue;
  93
  94                 CDEBUG(D_INFO, "Target idx %d is %s conn %#llx\n", i,
  95                        tgt->ltd_uuid.uuid, tgt->ltd_exp->exp_handle.h_cookie);
  96
  97                 if (obd_uuid_equals(uuid, &tgt->ltd_uuid))
  98                         break;
  99         }
 100
 101         if (i == lmv->desc.ld_tgt_count) {
 102                 rc = -EINVAL;
 103                 goto out_lmv_lock;
 104         }
 105
 106         obd = class_exp2obd(tgt->ltd_exp);
 107         if (!obd) {
 108                 rc = -ENOTCONN;
 109                 goto out_lmv_lock;
 110         }
 111
 112         CDEBUG(D_INFO, "Found OBD %s=%s device %d (%p) type %s at LMV idx %d\n",
 113                obd->obd_name, obd->obd_uuid.uuid, obd->obd_minor, obd,
 114                obd->obd_type->typ_name, i);
 115         LASSERT(strcmp(obd->obd_type->typ_name, LUSTRE_MDC_NAME) == 0);
 116
 117         if (tgt->ltd_active == activate) {
 118                 CDEBUG(D_INFO, "OBD %p already %sactive!\n", obd,
 119                        activate ? "" : "in");
 120                 goto out_lmv_lock;
 121         }
 122
 123         CDEBUG(D_INFO, "Marking OBD %p %sactive\n", obd,
 124                activate ? "" : "in");
 125         lmv_activate_target(lmv, tgt, activate);
 126
 127  out_lmv_lock:
 128         spin_unlock(&lmv->lmv_lock);
 129         return rc;
 130 }
 131
 132 static struct obd_uuid *lmv_get_uuid(struct obd_export *exp)
 133 {
 134         struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
 135
 136         return obd_get_uuid(lmv->tgts[0]->ltd_exp);
 137 }
 138
 139 static int lmv_notify(struct obd_device *obd, struct obd_device *watched,
 140                       enum obd_notify_event ev, void *data)
 141 {
 142         struct obd_connect_data *conn_data;
 143         struct lmv_obd    *lmv = &obd->u.lmv;
 144         struct obd_uuid  *uuid;
 145         int                   rc = 0;
 146
 147         if (strcmp(watched->obd_type->typ_name, LUSTRE_MDC_NAME)) {
 148                 CERROR("unexpected notification of %s %s!\n",
 149                        watched->obd_type->typ_name,
 150                        watched->obd_name);
 151                 return -EINVAL;
 152         }
 153
 154         uuid = &watched->u.cli.cl_target_uuid;
 155         if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE) {
 156                 /*
 157                  * Set MDC as active before notifying the observer, so the
 158                  * observer can use the MDC normally.
 159                  */
 160                 rc = lmv_set_mdc_active(lmv, uuid,
 161                                         ev == OBD_NOTIFY_ACTIVE);
 162                 if (rc) {
 163                         CERROR("%sactivation of %s failed: %d\n",
 164                                ev == OBD_NOTIFY_ACTIVE ? "" : "de",
 165                                uuid->uuid, rc);
 166                         return rc;
 167                 }
 168         } else if (ev == OBD_NOTIFY_OCD) {
 169                 conn_data = &watched->u.cli.cl_import->imp_connect_data;
 170                 /*
 171                  * XXX: Make sure that ocd_connect_flags from all targets are
 172                  * the same. Otherwise one of MDTs runs wrong version or
 173                  * something like this.  --umka
 174                  */
 175                 obd->obd_self_export->exp_connect_data = *conn_data;
 176         }
 177 #if 0
 178         else if (ev == OBD_NOTIFY_DISCON) {
 179                 /*
 180                  * For disconnect event, flush fld cache for failout MDS case.
 181                  */
 182                 fld_client_flush(&lmv->lmv_fld);
 183         }
 184 #endif
 185         /*
 186          * Pass the notification up the chain.
 187          */
 188         if (obd->obd_observer)
 189                 rc = obd_notify(obd->obd_observer, watched, ev, data);
 190
 191         return rc;
 192 }
 193
 194 /**
 195  * This is fake connect function. Its purpose is to initialize lmv and say
 196  * caller that everything is okay. Real connection will be performed later.
 197  */
 198 static int lmv_connect(const struct lu_env *env,
 199                        struct obd_export **exp, struct obd_device *obd,
 200                        struct obd_uuid *cluuid, struct obd_connect_data *data,
 201                        void *localdata)
 202 {
 203         struct lmv_obd  *lmv = &obd->u.lmv;
 204         struct lustre_handle  conn = { 0 };
 205         int                 rc = 0;
 206
 207         /*
 208          * We don't want to actually do the underlying connections more than
 209          * once, so keep track.
 210          */
 211         lmv->refcount++;
 212         if (lmv->refcount > 1) {
 213                 *exp = NULL;
 214                 return 0;
 215         }
 216
 217         rc = class_connect(&conn, obd, cluuid);
 218         if (rc) {
 219                 CERROR("class_connection() returned %d\n", rc);
 220                 return rc;
 221         }
 222
 223         *exp = class_conn2export(&conn);
 224         class_export_get(*exp);
 225
 226         lmv->exp = *exp;
 227         lmv->connected = 0;
 228         lmv->cluuid = *cluuid;
 229
 230         if (data)
 231                 lmv->conn_data = *data;
 232
 233         lmv->lmv_tgts_kobj = kobject_create_and_add("target_obds",
 234                                                     &obd->obd_kobj);
 235         /*
 236          * All real clients should perform actual connection right away, because
 237          * it is possible, that LMV will not have opportunity to connect targets
 238          * and MDC stuff will be called directly, for instance while reading
 239          * ../mdc/../kbytesfree procfs file, etc.
 240          */
 241         if (data && data->ocd_connect_flags & OBD_CONNECT_REAL)
 242                 rc = lmv_check_connect(obd);
 243
 244         if (rc && lmv->lmv_tgts_kobj)
 245                 kobject_put(lmv->lmv_tgts_kobj);
 246
 247         return rc;
 248 }
 249
 250 static void lmv_set_timeouts(struct obd_device *obd)
 251 {
 252         struct lmv_tgt_desc   *tgt;
 253         struct lmv_obd  *lmv;
 254         int                 i;
 255
 256         lmv = &obd->u.lmv;
 257         if (lmv->server_timeout == 0)
 258                 return;
 259
 260         if (lmv->connected == 0)
 261                 return;
 262
 263         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
 264                 tgt = lmv->tgts[i];
 265                 if (!tgt || !tgt->ltd_exp || tgt->ltd_active == 0)
 266                         continue;
 267
 268                 obd_set_info_async(NULL, tgt->ltd_exp, sizeof(KEY_INTERMDS),
 269                                    KEY_INTERMDS, 0, NULL, NULL);
 270         }
 271 }
 272
 273 static int lmv_init_ea_size(struct obd_export *exp, int easize,
 274                             int def_easize, int cookiesize, int def_cookiesize)
 275 {
 276         struct obd_device   *obd = exp->exp_obd;
 277         struct lmv_obd      *lmv = &obd->u.lmv;
 278         int               i;
 279         int               rc = 0;
 280         int               change = 0;
 281
 282         if (lmv->max_easize < easize) {
 283                 lmv->max_easize = easize;
 284                 change = 1;
 285         }
 286         if (lmv->max_def_easize < def_easize) {
 287                 lmv->max_def_easize = def_easize;
 288                 change = 1;
 289         }
 290         if (lmv->max_cookiesize < cookiesize) {
 291                 lmv->max_cookiesize = cookiesize;
 292                 change = 1;
 293         }
 294         if (lmv->max_def_cookiesize < def_cookiesize) {
 295                 lmv->max_def_cookiesize = def_cookiesize;
 296                 change = 1;
 297         }
 298         if (change == 0)
 299                 return 0;
 300
 301         if (lmv->connected == 0)
 302                 return 0;
 303
 304         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
 305                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp ||
 306                     lmv->tgts[i]->ltd_active == 0) {
 307                         CWARN("%s: NULL export for %d\n", obd->obd_name, i);
 308                         continue;
 309                 }
 310
 311                 rc = md_init_ea_size(lmv->tgts[i]->ltd_exp, easize, def_easize,
 312                                      cookiesize, def_cookiesize);
 313                 if (rc) {
 314                         CERROR("%s: obd_init_ea_size() failed on MDT target %d: rc = %d\n",
 315                                obd->obd_name, i, rc);
 316                         break;
 317                 }
 318         }
 319         return rc;
 320 }
 321
 322 #define MAX_STRING_SIZE 128
 323
 324 static int lmv_connect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
 325 {
 326         struct lmv_obd    *lmv = &obd->u.lmv;
 327         struct obd_uuid  *cluuid = &lmv->cluuid;
 328         struct obd_uuid   lmv_mdc_uuid = { "LMV_MDC_UUID" };
 329         struct obd_device       *mdc_obd;
 330         struct obd_export       *mdc_exp;
 331         struct lu_fld_target     target;
 332         int                   rc;
 333
 334         mdc_obd = class_find_client_obd(&tgt->ltd_uuid, LUSTRE_MDC_NAME,
 335                                         &obd->obd_uuid);
 336         if (!mdc_obd) {
 337                 CERROR("target %s not attached\n", tgt->ltd_uuid.uuid);
 338                 return -EINVAL;
 339         }
 340
 341         CDEBUG(D_CONFIG, "connect to %s(%s) - %s, %s FOR %s\n",
 342                mdc_obd->obd_name, mdc_obd->obd_uuid.uuid,
 343                tgt->ltd_uuid.uuid, obd->obd_uuid.uuid, cluuid->uuid);
 344
 345         if (!mdc_obd->obd_set_up) {
 346                 CERROR("target %s is not set up\n", tgt->ltd_uuid.uuid);
 347                 return -EINVAL;
 348         }
 349
 350         rc = obd_connect(NULL, &mdc_exp, mdc_obd, &lmv_mdc_uuid,
 351                          &lmv->conn_data, NULL);
 352         if (rc) {
 353                 CERROR("target %s connect error %d\n", tgt->ltd_uuid.uuid, rc);
 354                 return rc;
 355         }
 356
 357         /*
 358          * Init fid sequence client for this mdc and add new fld target.
 359          */
 360         rc = obd_fid_init(mdc_obd, mdc_exp, LUSTRE_SEQ_METADATA);
 361         if (rc)
 362                 return rc;
 363
 364         target.ft_srv = NULL;
 365         target.ft_exp = mdc_exp;
 366         target.ft_idx = tgt->ltd_idx;
 367
 368         fld_client_add_target(&lmv->lmv_fld, &target);
 369
 370         rc = obd_register_observer(mdc_obd, obd);
 371         if (rc) {
 372                 obd_disconnect(mdc_exp);
 373                 CERROR("target %s register_observer error %d\n",
 374                        tgt->ltd_uuid.uuid, rc);
 375                 return rc;
 376         }
 377
 378         if (obd->obd_observer) {
 379                 /*
 380                  * Tell the observer about the new target.
 381                  */
 382                 rc = obd_notify(obd->obd_observer, mdc_exp->exp_obd,
 383                                 OBD_NOTIFY_ACTIVE,
 384                                 (void *)(tgt - lmv->tgts[0]));
 385                 if (rc) {
 386                         obd_disconnect(mdc_exp);
 387                         return rc;
 388                 }
 389         }
 390
 391         tgt->ltd_active = 1;
 392         tgt->ltd_exp = mdc_exp;
 393         lmv->desc.ld_active_tgt_count++;
 394
 395         md_init_ea_size(tgt->ltd_exp, lmv->max_easize, lmv->max_def_easize,
 396                         lmv->max_cookiesize, lmv->max_def_cookiesize);
 397
 398         CDEBUG(D_CONFIG, "Connected to %s(%s) successfully (%d)\n",
 399                mdc_obd->obd_name, mdc_obd->obd_uuid.uuid,
 400                atomic_read(&obd->obd_refcount));
 401
 402         if (lmv->lmv_tgts_kobj)
 403                 /* Even if we failed to create the link, that's fine */
 404                 rc = sysfs_create_link(lmv->lmv_tgts_kobj, &mdc_obd->obd_kobj,
 405                                        mdc_obd->obd_name);
 406         return 0;
 407 }
 408
 409 static void lmv_del_target(struct lmv_obd *lmv, int index)
 410 {
 411         if (!lmv->tgts[index])
 412                 return;
 413
 414         kfree(lmv->tgts[index]);
 415         lmv->tgts[index] = NULL;
 416         return;
 417 }
 418
 419 static int lmv_add_target(struct obd_device *obd, struct obd_uuid *uuidp,
 420                           __u32 index, int gen)
 421 {
 422         struct lmv_obd      *lmv = &obd->u.lmv;
 423         struct lmv_tgt_desc *tgt;
 424         int               rc = 0;
 425
 426         CDEBUG(D_CONFIG, "Target uuid: %s. index %d\n", uuidp->uuid, index);
 427
 428         lmv_init_lock(lmv);
 429
 430         if (lmv->desc.ld_tgt_count == 0) {
 431                 struct obd_device *mdc_obd;
 432
 433                 mdc_obd = class_find_client_obd(uuidp, LUSTRE_MDC_NAME,
 434                                                 &obd->obd_uuid);
 435                 if (!mdc_obd) {
 436                         lmv_init_unlock(lmv);
 437                         CERROR("%s: Target %s not attached: rc = %d\n",
 438                                obd->obd_name, uuidp->uuid, -EINVAL);
 439                         return -EINVAL;
 440                 }
 441         }
 442
 443         if ((index < lmv->tgts_size) && lmv->tgts[index]) {
 444                 tgt = lmv->tgts[index];
 445                 CERROR("%s: UUID %s already assigned at LOV target index %d: rc = %d\n",
 446                        obd->obd_name,
 447                        obd_uuid2str(&tgt->ltd_uuid), index, -EEXIST);
 448                 lmv_init_unlock(lmv);
 449                 return -EEXIST;
 450         }
 451
 452         if (index >= lmv->tgts_size) {
 453                 /* We need to reallocate the lmv target array. */
 454                 struct lmv_tgt_desc **newtgts, **old = NULL;
 455                 __u32 newsize = 1;
 456                 __u32 oldsize = 0;
 457
 458                 while (newsize < index + 1)
 459                         newsize <<= 1;
 460                 newtgts = kcalloc(newsize, sizeof(*newtgts), GFP_NOFS);
 461                 if (!newtgts) {
 462                         lmv_init_unlock(lmv);
 463                         return -ENOMEM;
 464                 }
 465
 466                 if (lmv->tgts_size) {
 467                         memcpy(newtgts, lmv->tgts,
 468                                sizeof(*newtgts) * lmv->tgts_size);
 469                         old = lmv->tgts;
 470                         oldsize = lmv->tgts_size;
 471                 }
 472
 473                 lmv->tgts = newtgts;
 474                 lmv->tgts_size = newsize;
 475                 smp_rmb();
 476                 kfree(old);
 477
 478                 CDEBUG(D_CONFIG, "tgts: %p size: %d\n", lmv->tgts,
 479                        lmv->tgts_size);
 480         }
 481
 482         tgt = kzalloc(sizeof(*tgt), GFP_NOFS);
 483         if (!tgt) {
 484                 lmv_init_unlock(lmv);
 485                 return -ENOMEM;
 486         }
 487
 488         mutex_init(&tgt->ltd_fid_mutex);
 489         tgt->ltd_idx = index;
 490         tgt->ltd_uuid = *uuidp;
 491         tgt->ltd_active = 0;
 492         lmv->tgts[index] = tgt;
 493         if (index >= lmv->desc.ld_tgt_count)
 494                 lmv->desc.ld_tgt_count = index + 1;
 495
 496         if (lmv->connected) {
 497                 rc = lmv_connect_mdc(obd, tgt);
 498                 if (rc) {
 499                         spin_lock(&lmv->lmv_lock);
 500                         lmv->desc.ld_tgt_count--;
 501                         memset(tgt, 0, sizeof(*tgt));
 502                         spin_unlock(&lmv->lmv_lock);
 503                 } else {
 504                         int easize = sizeof(struct lmv_stripe_md) +
 505                                 lmv->desc.ld_tgt_count * sizeof(struct lu_fid);
 506                         lmv_init_ea_size(obd->obd_self_export, easize, 0, 0, 0);
 507                 }
 508         }
 509
 510         lmv_init_unlock(lmv);
 511         return rc;
 512 }
 513
 514 int lmv_check_connect(struct obd_device *obd)
 515 {
 516         struct lmv_obd       *lmv = &obd->u.lmv;
 517         struct lmv_tgt_desc  *tgt;
 518         int                i;
 519         int                rc;
 520         int                easize;
 521
 522         if (lmv->connected)
 523                 return 0;
 524
 525         lmv_init_lock(lmv);
 526         if (lmv->connected) {
 527                 lmv_init_unlock(lmv);
 528                 return 0;
 529         }
 530
 531         if (lmv->desc.ld_tgt_count == 0) {
 532                 lmv_init_unlock(lmv);
 533                 CERROR("%s: no targets configured.\n", obd->obd_name);
 534                 return -EINVAL;
 535         }
 536
 537         CDEBUG(D_CONFIG, "Time to connect %s to %s\n",
 538                lmv->cluuid.uuid, obd->obd_name);
 539
 540         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
 541                 tgt = lmv->tgts[i];
 542                 if (!tgt)
 543                         continue;
 544                 rc = lmv_connect_mdc(obd, tgt);
 545                 if (rc)
 546                         goto out_disc;
 547         }
 548
 549         lmv_set_timeouts(obd);
 550         class_export_put(lmv->exp);
 551         lmv->connected = 1;
 552         easize = lmv_get_easize(lmv);
 553         lmv_init_ea_size(obd->obd_self_export, easize, 0, 0, 0);
 554         lmv_init_unlock(lmv);
 555         return 0;
 556
 557  out_disc:
 558         while (i-- > 0) {
 559                 int rc2;
 560
 561                 tgt = lmv->tgts[i];
 562                 if (!tgt)
 563                         continue;
 564                 tgt->ltd_active = 0;
 565                 if (tgt->ltd_exp) {
 566                         --lmv->desc.ld_active_tgt_count;
 567                         rc2 = obd_disconnect(tgt->ltd_exp);
 568                         if (rc2) {
 569                                 CERROR("LMV target %s disconnect on MDC idx %d: error %d\n",
 570                                        tgt->ltd_uuid.uuid, i, rc2);
 571                         }
 572                 }
 573         }
 574         class_disconnect(lmv->exp);
 575         lmv_init_unlock(lmv);
 576         return rc;
 577 }
 578
 579 static int lmv_disconnect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
 580 {
 581         struct lmv_obd   *lmv = &obd->u.lmv;
 582         struct obd_device      *mdc_obd;
 583         int                  rc;
 584
 585         mdc_obd = class_exp2obd(tgt->ltd_exp);
 586
 587         if (mdc_obd) {
 588                 mdc_obd->obd_force = obd->obd_force;
 589                 mdc_obd->obd_fail = obd->obd_fail;
 590                 mdc_obd->obd_no_recov = obd->obd_no_recov;
 591
 592                 if (lmv->lmv_tgts_kobj)
 593                         sysfs_remove_link(lmv->lmv_tgts_kobj,
 594                                           mdc_obd->obd_name);
 595         }
 596
 597         rc = obd_fid_fini(tgt->ltd_exp->exp_obd);
 598         if (rc)
 599                 CERROR("Can't finalize fids factory\n");
 600
 601         CDEBUG(D_INFO, "Disconnected from %s(%s) successfully\n",
 602                tgt->ltd_exp->exp_obd->obd_name,
 603                tgt->ltd_exp->exp_obd->obd_uuid.uuid);
 604
 605         obd_register_observer(tgt->ltd_exp->exp_obd, NULL);
 606         rc = obd_disconnect(tgt->ltd_exp);
 607         if (rc) {
 608                 if (tgt->ltd_active) {
 609                         CERROR("Target %s disconnect error %d\n",
 610                                tgt->ltd_uuid.uuid, rc);
 611                 }
 612         }
 613
 614         lmv_activate_target(lmv, tgt, 0);
 615         tgt->ltd_exp = NULL;
 616         return 0;
 617 }
 618
 619 static int lmv_disconnect(struct obd_export *exp)
 620 {
 621         struct obd_device     *obd = class_exp2obd(exp);
 622         struct lmv_obd  *lmv = &obd->u.lmv;
 623         int                 rc;
 624         int                 i;
 625
 626         if (!lmv->tgts)
 627                 goto out_local;
 628
 629         /*
 630          * Only disconnect the underlying layers on the final disconnect.
 631          */
 632         lmv->refcount--;
 633         if (lmv->refcount != 0)
 634                 goto out_local;
 635
 636         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
 637                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp)
 638                         continue;
 639
 640                 lmv_disconnect_mdc(obd, lmv->tgts[i]);
 641         }
 642
 643         if (lmv->lmv_tgts_kobj)
 644                 kobject_put(lmv->lmv_tgts_kobj);
 645
 646 out_local:
 647         /*
 648          * This is the case when no real connection is established by
 649          * lmv_check_connect().
 650          */
 651         if (!lmv->connected)
 652                 class_export_put(exp);
 653         rc = class_disconnect(exp);
 654         if (lmv->refcount == 0)
 655                 lmv->connected = 0;
 656         return rc;
 657 }
 658
 659 static int lmv_fid2path(struct obd_export *exp, int len, void *karg,
 660                         void __user *uarg)
 661 {
 662         struct obd_device       *obddev = class_exp2obd(exp);
 663         struct lmv_obd          *lmv = &obddev->u.lmv;
 664         struct getinfo_fid2path *gf;
 665         struct lmv_tgt_desc     *tgt;
 666         struct getinfo_fid2path *remote_gf = NULL;
 667         int                     remote_gf_size = 0;
 668         int                     rc;
 669
 670         gf = (struct getinfo_fid2path *)karg;
 671         tgt = lmv_find_target(lmv, &gf->gf_fid);
 672         if (IS_ERR(tgt))
 673                 return PTR_ERR(tgt);
 674
 675 repeat_fid2path:
 676         rc = obd_iocontrol(OBD_IOC_FID2PATH, tgt->ltd_exp, len, gf, uarg);
 677         if (rc != 0 && rc != -EREMOTE)
 678                 goto out_fid2path;
 679
 680         /* If remote_gf != NULL, it means just building the
 681          * path on the remote MDT, copy this path segment to gf
 682          */
 683         if (remote_gf) {
 684                 struct getinfo_fid2path *ori_gf;
 685                 char *ptr;
 686
 687                 ori_gf = (struct getinfo_fid2path *)karg;
 688                 if (strlen(ori_gf->gf_path) +
 689                     strlen(gf->gf_path) > ori_gf->gf_pathlen) {
 690                         rc = -EOVERFLOW;
 691                         goto out_fid2path;
 692                 }
 693
 694                 ptr = ori_gf->gf_path;
 695
 696                 memmove(ptr + strlen(gf->gf_path) + 1, ptr,
 697                         strlen(ori_gf->gf_path));
 698
 699                 strncpy(ptr, gf->gf_path, strlen(gf->gf_path));
 700                 ptr += strlen(gf->gf_path);
 701                 *ptr = '/';
 702         }
 703
 704         CDEBUG(D_INFO, "%s: get path %s "DFID" rec: %llu ln: %u\n",
 705                tgt->ltd_exp->exp_obd->obd_name,
 706                gf->gf_path, PFID(&gf->gf_fid), gf->gf_recno,
 707                gf->gf_linkno);
 708
 709         if (rc == 0)
 710                 goto out_fid2path;
 711
 712         /* sigh, has to go to another MDT to do path building further */
 713         if (!remote_gf) {
 714                 remote_gf_size = sizeof(*remote_gf) + PATH_MAX;
 715                 remote_gf = kzalloc(remote_gf_size, GFP_NOFS);
 716                 if (!remote_gf) {
 717                         rc = -ENOMEM;
 718                         goto out_fid2path;
 719                 }
 720                 remote_gf->gf_pathlen = PATH_MAX;
 721         }
 722
 723         if (!fid_is_sane(&gf->gf_fid)) {
 724                 CERROR("%s: invalid FID "DFID": rc = %d\n",
 725                        tgt->ltd_exp->exp_obd->obd_name,
 726                        PFID(&gf->gf_fid), -EINVAL);
 727                 rc = -EINVAL;
 728                 goto out_fid2path;
 729         }
 730
 731         tgt = lmv_find_target(lmv, &gf->gf_fid);
 732         if (IS_ERR(tgt)) {
 733                 rc = -EINVAL;
 734                 goto out_fid2path;
 735         }
 736
 737         remote_gf->gf_fid = gf->gf_fid;
 738         remote_gf->gf_recno = -1;
 739         remote_gf->gf_linkno = -1;
 740         memset(remote_gf->gf_path, 0, remote_gf->gf_pathlen);
 741         gf = remote_gf;
 742         goto repeat_fid2path;
 743
 744 out_fid2path:
 745         kfree(remote_gf);
 746         return rc;
 747 }
 748
 749 static int lmv_hsm_req_count(struct lmv_obd *lmv,
 750                              const struct hsm_user_request *hur,
 751                              const struct lmv_tgt_desc *tgt_mds)
 752 {
 753         int                     i, nr = 0;
 754         struct lmv_tgt_desc    *curr_tgt;
 755
 756         /* count how many requests must be sent to the given target */
 757         for (i = 0; i < hur->hur_request.hr_itemcount; i++) {
 758                 curr_tgt = lmv_find_target(lmv, &hur->hur_user_item[i].hui_fid);
 759                 if (obd_uuid_equals(&curr_tgt->ltd_uuid, &tgt_mds->ltd_uuid))
 760                         nr++;
 761         }
 762         return nr;
 763 }
 764
 765 static void lmv_hsm_req_build(struct lmv_obd *lmv,
 766                               struct hsm_user_request *hur_in,
 767                               const struct lmv_tgt_desc *tgt_mds,
 768                               struct hsm_user_request *hur_out)
 769 {
 770         int                     i, nr_out;
 771         struct lmv_tgt_desc    *curr_tgt;
 772
 773         /* build the hsm_user_request for the given target */
 774         hur_out->hur_request = hur_in->hur_request;
 775         nr_out = 0;
 776         for (i = 0; i < hur_in->hur_request.hr_itemcount; i++) {
 777                 curr_tgt = lmv_find_target(lmv,
 778                                            &hur_in->hur_user_item[i].hui_fid);
 779                 if (obd_uuid_equals(&curr_tgt->ltd_uuid, &tgt_mds->ltd_uuid)) {
 780                         hur_out->hur_user_item[nr_out] =
 781                                 hur_in->hur_user_item[i];
 782                         nr_out++;
 783                 }
 784         }
 785         hur_out->hur_request.hr_itemcount = nr_out;
 786         memcpy(hur_data(hur_out), hur_data(hur_in),
 787                hur_in->hur_request.hr_data_len);
 788 }
 789
 790 static int lmv_hsm_ct_unregister(struct lmv_obd *lmv, unsigned int cmd, int len,
 791                                  struct lustre_kernelcomm *lk,
 792                                  void __user *uarg)
 793 {
 794         int rc = 0;
 795         __u32 i;
 796
 797         /* unregister request (call from llapi_hsm_copytool_fini) */
 798         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
 799                 /* best effort: try to clean as much as possible
 800                  * (continue on error)
 801                  */
 802                 obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp, len, lk, uarg);
 803         }
 804
 805         /* Whatever the result, remove copytool from kuc groups.
 806          * Unreached coordinators will get EPIPE on next requests
 807          * and will unregister automatically.
 808          */
 809         rc = libcfs_kkuc_group_rem(lk->lk_uid, lk->lk_group);
 810
 811         return rc;
 812 }
 813
 814 static int lmv_hsm_ct_register(struct lmv_obd *lmv, unsigned int cmd, int len,
 815                                struct lustre_kernelcomm *lk, void __user *uarg)
 816 {
 817         struct file *filp;
 818         __u32 i, j;
 819         int err, rc = 0;
 820         bool any_set = false;
 821         struct kkuc_ct_data kcd = { 0 };
 822
 823         /* All or nothing: try to register to all MDS.
 824          * In case of failure, unregister from previous MDS,
 825          * except if it because of inactive target.
 826          */
 827         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
 828                 err = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp, len, lk, uarg);
 829                 if (err) {
 830                         if (lmv->tgts[i]->ltd_active) {
 831                                 /* permanent error */
 832                                 CERROR("error: iocontrol MDC %s on MDTidx %d cmd %x: err = %d\n",
 833                                        lmv->tgts[i]->ltd_uuid.uuid,
 834                                        i, cmd, err);
 835                                 rc = err;
 836                                 lk->lk_flags |= LK_FLG_STOP;
 837                                 /* unregister from previous MDS */
 838                                 for (j = 0; j < i; j++)
 839                                         obd_iocontrol(cmd,
 840                                                       lmv->tgts[j]->ltd_exp,
 841                                                       len, lk, uarg);
 842                                 return rc;
 843                         }
 844                         /* else: transient error.
 845                          * kuc will register to the missing MDT when it is back
 846                          */
 847                 } else {
 848                         any_set = true;
 849                 }
 850         }
 851
 852         if (!any_set)
 853                 /* no registration done: return error */
 854                 return -ENOTCONN;
 855
 856         /* at least one registration done, with no failure */
 857         filp = fget(lk->lk_wfd);
 858         if (!filp)
 859                 return -EBADF;
 860
 861         kcd.kcd_magic = KKUC_CT_DATA_MAGIC;
 862         kcd.kcd_uuid = lmv->cluuid;
 863         kcd.kcd_archive = lk->lk_data;
 864
 865         rc = libcfs_kkuc_group_add(filp, lk->lk_uid, lk->lk_group,
 866                                    &kcd, sizeof(kcd));
 867         if (rc) {
 868                 if (filp)
 869                         fput(filp);
 870         }
 871
 872         return rc;
 873 }
 874
 875 static int lmv_iocontrol(unsigned int cmd, struct obd_export *exp,
 876                          int len, void *karg, void __user *uarg)
 877 {
 878         struct obd_device    *obddev = class_exp2obd(exp);
 879         struct lmv_obd       *lmv = &obddev->u.lmv;
 880         int                i = 0;
 881         int                rc = 0;
 882         int                set = 0;
 883         int                count = lmv->desc.ld_tgt_count;
 884
 885         if (count == 0)
 886                 return -ENOTTY;
 887
 888         switch (cmd) {
 889         case IOC_OBD_STATFS: {
 890                 struct obd_ioctl_data *data = karg;
 891                 struct obd_device *mdc_obd;
 892                 struct obd_statfs stat_buf = {0};
 893                 __u32 index;
 894
 895                 memcpy(&index, data->ioc_inlbuf2, sizeof(__u32));
 896                 if (index >= count)
 897                         return -ENODEV;
 898
 899                 if (!lmv->tgts[index] || lmv->tgts[index]->ltd_active == 0)
 900                         return -ENODATA;
 901
 902                 mdc_obd = class_exp2obd(lmv->tgts[index]->ltd_exp);
 903                 if (!mdc_obd)
 904                         return -EINVAL;
 905
 906                 /* copy UUID */
 907                 if (copy_to_user(data->ioc_pbuf2, obd2cli_tgt(mdc_obd),
 908                                  min((int)data->ioc_plen2,
 909                                      (int)sizeof(struct obd_uuid))))
 910                         return -EFAULT;
 911
 912                 rc = obd_statfs(NULL, lmv->tgts[index]->ltd_exp, &stat_buf,
 913                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
 914                                 0);
 915                 if (rc)
 916                         return rc;
 917                 if (copy_to_user(data->ioc_pbuf1, &stat_buf,
 918                                  min((int)data->ioc_plen1,
 919                                      (int)sizeof(stat_buf))))
 920                         return -EFAULT;
 921                 break;
 922         }
 923         case OBD_IOC_QUOTACTL: {
 924                 struct if_quotactl *qctl = karg;
 925                 struct lmv_tgt_desc *tgt = NULL;
 926                 struct obd_quotactl *oqctl;
 927
 928                 if (qctl->qc_valid == QC_MDTIDX) {
 929                         if (qctl->qc_idx < 0 || count <= qctl->qc_idx)
 930                                 return -EINVAL;
 931
 932                         tgt = lmv->tgts[qctl->qc_idx];
 933                         if (!tgt || !tgt->ltd_exp)
 934                                 return -EINVAL;
 935                 } else if (qctl->qc_valid == QC_UUID) {
 936                         for (i = 0; i < count; i++) {
 937                                 tgt = lmv->tgts[i];
 938                                 if (!tgt)
 939                                         continue;
 940                                 if (!obd_uuid_equals(&tgt->ltd_uuid,
 941                                                      &qctl->obd_uuid))
 942                                         continue;
 943
 944                                 if (!tgt->ltd_exp)
 945                                         return -EINVAL;
 946
 947                                 break;
 948                         }
 949                 } else {
 950                         return -EINVAL;
 951                 }
 952
 953                 if (i >= count)
 954                         return -EAGAIN;
 955
 956                 LASSERT(tgt && tgt->ltd_exp);
 957                 oqctl = kzalloc(sizeof(*oqctl), GFP_NOFS);
 958                 if (!oqctl)
 959                         return -ENOMEM;
 960
 961                 QCTL_COPY(oqctl, qctl);
 962                 rc = obd_quotactl(tgt->ltd_exp, oqctl);
 963                 if (rc == 0) {
 964                         QCTL_COPY(qctl, oqctl);
 965                         qctl->qc_valid = QC_MDTIDX;
 966                         qctl->obd_uuid = tgt->ltd_uuid;
 967                 }
 968                 kfree(oqctl);
 969                 break;
 970         }
 971         case OBD_IOC_CHANGELOG_SEND:
 972         case OBD_IOC_CHANGELOG_CLEAR: {
 973                 struct ioc_changelog *icc = karg;
 974
 975                 if (icc->icc_mdtindex >= count)
 976                         return -ENODEV;
 977
 978                 if (!lmv->tgts[icc->icc_mdtindex] ||
 979                     !lmv->tgts[icc->icc_mdtindex]->ltd_exp ||
 980                     lmv->tgts[icc->icc_mdtindex]->ltd_active == 0)
 981                         return -ENODEV;
 982                 rc = obd_iocontrol(cmd, lmv->tgts[icc->icc_mdtindex]->ltd_exp,
 983                                    sizeof(*icc), icc, NULL);
 984                 break;
 985         }
 986         case LL_IOC_GET_CONNECT_FLAGS: {
 987                 if (!lmv->tgts[0])
 988                         return -ENODATA;
 989                 rc = obd_iocontrol(cmd, lmv->tgts[0]->ltd_exp, len, karg, uarg);
 990                 break;
 991         }
 992         case OBD_IOC_FID2PATH: {
 993                 rc = lmv_fid2path(exp, len, karg, uarg);
 994                 break;
 995         }
 996         case LL_IOC_HSM_STATE_GET:
 997         case LL_IOC_HSM_STATE_SET:
 998         case LL_IOC_HSM_ACTION: {
 999                 struct md_op_data       *op_data = karg;
1000                 struct lmv_tgt_desc     *tgt;
1001
1002                 tgt = lmv_find_target(lmv, &op_data->op_fid1);
1003                 if (IS_ERR(tgt))
1004                         return PTR_ERR(tgt);
1005
1006                 if (!tgt->ltd_exp)
1007                         return -EINVAL;
1008
1009                 rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
1010                 break;
1011         }
1012         case LL_IOC_HSM_PROGRESS: {
1013                 const struct hsm_progress_kernel *hpk = karg;
1014                 struct lmv_tgt_desc     *tgt;
1015
1016                 tgt = lmv_find_target(lmv, &hpk->hpk_fid);
1017                 if (IS_ERR(tgt))
1018                         return PTR_ERR(tgt);
1019                 rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
1020                 break;
1021         }
1022         case LL_IOC_HSM_REQUEST: {
1023                 struct hsm_user_request *hur = karg;
1024                 struct lmv_tgt_desc     *tgt;
1025                 unsigned int reqcount = hur->hur_request.hr_itemcount;
1026
1027                 if (reqcount == 0)
1028                         return 0;
1029
1030                 /* if the request is about a single fid
1031                  * or if there is a single MDS, no need to split
1032                  * the request.
1033                  */
1034                 if (reqcount == 1 || count == 1) {
1035                         tgt = lmv_find_target(lmv,
1036                                               &hur->hur_user_item[0].hui_fid);
1037                         if (IS_ERR(tgt))
1038                                 return PTR_ERR(tgt);
1039                         rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
1040                 } else {
1041                         /* split fid list to their respective MDS */
1042                         for (i = 0; i < count; i++) {
1043                                 unsigned int            nr, reqlen;
1044                                 int                     rc1;
1045                                 struct hsm_user_request *req;
1046
1047                                 nr = lmv_hsm_req_count(lmv, hur, lmv->tgts[i]);
1048                                 if (nr == 0) /* nothing for this MDS */
1049                                         continue;
1050
1051                                 /* build a request with fids for this MDS */
1052                                 reqlen = offsetof(typeof(*hur),
1053                                                   hur_user_item[nr])
1054                                          + hur->hur_request.hr_data_len;
1055                                 req = libcfs_kvzalloc(reqlen, GFP_NOFS);
1056                                 if (!req)
1057                                         return -ENOMEM;
1058
1059                                 lmv_hsm_req_build(lmv, hur, lmv->tgts[i], req);
1060
1061                                 rc1 = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp,
1062                                                     reqlen, req, uarg);
1063                                 if (rc1 != 0 && rc == 0)
1064                                         rc = rc1;
1065                                 kvfree(req);
1066                         }
1067                 }
1068                 break;
1069         }
1070         case LL_IOC_LOV_SWAP_LAYOUTS: {
1071                 struct md_op_data       *op_data = karg;
1072                 struct lmv_tgt_desc     *tgt1, *tgt2;
1073
1074                 tgt1 = lmv_find_target(lmv, &op_data->op_fid1);
1075                 if (IS_ERR(tgt1))
1076                         return PTR_ERR(tgt1);
1077
1078                 tgt2 = lmv_find_target(lmv, &op_data->op_fid2);
1079                 if (IS_ERR(tgt2))
1080                         return PTR_ERR(tgt2);
1081
1082                 if (!tgt1->ltd_exp || !tgt2->ltd_exp)
1083                         return -EINVAL;
1084
1085                 /* only files on same MDT can have their layouts swapped */
1086                 if (tgt1->ltd_idx != tgt2->ltd_idx)
1087                         return -EPERM;
1088
1089                 rc = obd_iocontrol(cmd, tgt1->ltd_exp, len, karg, uarg);
1090                 break;
1091         }
1092         case LL_IOC_HSM_CT_START: {
1093                 struct lustre_kernelcomm *lk = karg;
1094
1095                 if (lk->lk_flags & LK_FLG_STOP)
1096                         rc = lmv_hsm_ct_unregister(lmv, cmd, len, lk, uarg);
1097                 else
1098                         rc = lmv_hsm_ct_register(lmv, cmd, len, lk, uarg);
1099                 break;
1100         }
1101         default:
1102                 for (i = 0; i < count; i++) {
1103                         struct obd_device *mdc_obd;
1104                         int err;
1105
1106                         if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp)
1107                                 continue;
1108                         /* ll_umount_begin() sets force flag but for lmv, not
1109                          * mdc. Let's pass it through
1110                          */
1111                         mdc_obd = class_exp2obd(lmv->tgts[i]->ltd_exp);
1112                         mdc_obd->obd_force = obddev->obd_force;
1113                         err = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp, len,
1114                                             karg, uarg);
1115                         if (err == -ENODATA && cmd == OBD_IOC_POLL_QUOTACHECK) {
1116                                 return err;
1117                         } else if (err) {
1118                                 if (lmv->tgts[i]->ltd_active) {
1119                                         CERROR("error: iocontrol MDC %s on MDTidx %d cmd %x: err = %d\n",
1120                                                lmv->tgts[i]->ltd_uuid.uuid,
1121                                                i, cmd, err);
1122                                         if (!rc)
1123                                                 rc = err;
1124                                 }
1125                         } else
1126                                 set = 1;
1127                 }
1128                 if (!set && !rc)
1129                         rc = -EIO;
1130         }
1131         return rc;
1132 }
1133
1134 /**
1135  * This is _inode_ placement policy function (not name).
1136  */
1137 static int lmv_placement_policy(struct obd_device *obd,
1138                                 struct md_op_data *op_data, u32 *mds)
1139 {
1140         struct lmv_obd    *lmv = &obd->u.lmv;
1141
1142         LASSERT(mds);
1143
1144         if (lmv->desc.ld_tgt_count == 1) {
1145                 *mds = 0;
1146                 return 0;
1147         }
1148
1149         /**
1150          * If stripe_offset is provided during setdirstripe
1151          * (setdirstripe -i xx), xx MDS will be chosen.
1152          */
1153         if (op_data->op_cli_flags & CLI_SET_MEA) {
1154                 struct lmv_user_md *lum;
1155
1156                 lum = (struct lmv_user_md *)op_data->op_data;
1157                 if (lum->lum_type == LMV_STRIPE_TYPE &&
1158                     lum->lum_stripe_offset != -1) {
1159                         if (lum->lum_stripe_offset >= lmv->desc.ld_tgt_count) {
1160                                 CERROR("%s: Stripe_offset %d > MDT count %d: rc = %d\n",
1161                                        obd->obd_name,
1162                                        lum->lum_stripe_offset,
1163                                        lmv->desc.ld_tgt_count, -ERANGE);
1164                                 return -ERANGE;
1165                         }
1166                         *mds = lum->lum_stripe_offset;
1167                         return 0;
1168                 }
1169         }
1170
1171         /* Allocate new fid on target according to operation type and parent
1172          * home mds.
1173          */
1174         *mds = op_data->op_mds;
1175         return 0;
1176 }
1177
1178 int __lmv_fid_alloc(struct lmv_obd *lmv, struct lu_fid *fid, u32 mds)
1179 {
1180         struct lmv_tgt_desc     *tgt;
1181         int                      rc;
1182
1183         tgt = lmv_get_target(lmv, mds);
1184         if (IS_ERR(tgt))
1185                 return PTR_ERR(tgt);
1186
1187         /*
1188          * New seq alloc and FLD setup should be atomic. Otherwise we may find
1189          * on server that seq in new allocated fid is not yet known.
1190          */
1191         mutex_lock(&tgt->ltd_fid_mutex);
1192
1193         if (tgt->ltd_active == 0 || !tgt->ltd_exp) {
1194                 rc = -ENODEV;
1195                 goto out;
1196         }
1197
1198         /*
1199          * Asking underlaying tgt layer to allocate new fid.
1200          */
1201         rc = obd_fid_alloc(tgt->ltd_exp, fid, NULL);
1202         if (rc > 0) {
1203                 LASSERT(fid_is_sane(fid));
1204                 rc = 0;
1205         }
1206
1207 out:
1208         mutex_unlock(&tgt->ltd_fid_mutex);
1209         return rc;
1210 }
1211
1212 int lmv_fid_alloc(struct obd_export *exp, struct lu_fid *fid,
1213                   struct md_op_data *op_data)
1214 {
1215         struct obd_device     *obd = class_exp2obd(exp);
1216         struct lmv_obd  *lmv = &obd->u.lmv;
1217         u32                    mds = 0;
1218         int                 rc;
1219
1220         LASSERT(op_data);
1221         LASSERT(fid);
1222
1223         rc = lmv_placement_policy(obd, op_data, &mds);
1224         if (rc) {
1225                 CERROR("Can't get target for allocating fid, rc %d\n",
1226                        rc);
1227                 return rc;
1228         }
1229
1230         rc = __lmv_fid_alloc(lmv, fid, mds);
1231         if (rc) {
1232                 CERROR("Can't alloc new fid, rc %d\n", rc);
1233                 return rc;
1234         }
1235
1236         return rc;
1237 }
1238
1239 static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
1240 {
1241         struct lmv_obd       *lmv = &obd->u.lmv;
1242         struct lprocfs_static_vars  lvars = { NULL };
1243         struct lmv_desc     *desc;
1244         int                      rc;
1245
1246         if (LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
1247                 CERROR("LMV setup requires a descriptor\n");
1248                 return -EINVAL;
1249         }
1250
1251         desc = (struct lmv_desc *)lustre_cfg_buf(lcfg, 1);
1252         if (sizeof(*desc) > LUSTRE_CFG_BUFLEN(lcfg, 1)) {
1253                 CERROR("Lmv descriptor size wrong: %d > %d\n",
1254                        (int)sizeof(*desc), LUSTRE_CFG_BUFLEN(lcfg, 1));
1255                 return -EINVAL;
1256         }
1257
1258         lmv->tgts = kcalloc(32, sizeof(*lmv->tgts), GFP_NOFS);
1259         if (!lmv->tgts)
1260                 return -ENOMEM;
1261         lmv->tgts_size = 32;
1262
1263         obd_str2uuid(&lmv->desc.ld_uuid, desc->ld_uuid.uuid);
1264         lmv->desc.ld_tgt_count = 0;
1265         lmv->desc.ld_active_tgt_count = 0;
1266         lmv->max_cookiesize = 0;
1267         lmv->max_def_easize = 0;
1268         lmv->max_easize = 0;
1269         lmv->lmv_placement = PLACEMENT_CHAR_POLICY;
1270
1271         spin_lock_init(&lmv->lmv_lock);
1272         mutex_init(&lmv->init_mutex);
1273
1274         lprocfs_lmv_init_vars(&lvars);
1275
1276         lprocfs_obd_setup(obd, lvars.obd_vars, lvars.sysfs_vars);
1277         rc = ldebugfs_seq_create(obd->obd_debugfs_entry, "target_obd",
1278                                  0444, &lmv_proc_target_fops, obd);
1279         if (rc)
1280                 CWARN("%s: error adding LMV target_obd file: rc = %d\n",
1281                       obd->obd_name, rc);
1282         rc = fld_client_init(&lmv->lmv_fld, obd->obd_name,
1283                              LUSTRE_CLI_FLD_HASH_DHT);
1284         if (rc) {
1285                 CERROR("Can't init FLD, err %d\n", rc);
1286                 goto out;
1287         }
1288
1289         return 0;
1290
1291 out:
1292         return rc;
1293 }
1294
1295 static int lmv_cleanup(struct obd_device *obd)
1296 {
1297         struct lmv_obd   *lmv = &obd->u.lmv;
1298
1299         fld_client_fini(&lmv->lmv_fld);
1300         if (lmv->tgts) {
1301                 int i;
1302
1303                 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1304                         if (!lmv->tgts[i])
1305                                 continue;
1306                         lmv_del_target(lmv, i);
1307                 }
1308                 kfree(lmv->tgts);
1309                 lmv->tgts_size = 0;
1310         }
1311         return 0;
1312 }
1313
1314 static int lmv_process_config(struct obd_device *obd, u32 len, void *buf)
1315 {
1316         struct lustre_cfg       *lcfg = buf;
1317         struct obd_uuid         obd_uuid;
1318         int                     gen;
1319         __u32                   index;
1320         int                     rc;
1321
1322         switch (lcfg->lcfg_command) {
1323         case LCFG_ADD_MDC:
1324                 /* modify_mdc_tgts add 0:lustre-clilmv  1:lustre-MDT0000_UUID
1325                  * 2:0  3:1  4:lustre-MDT0000-mdc_UUID
1326                  */
1327                 if (LUSTRE_CFG_BUFLEN(lcfg, 1) > sizeof(obd_uuid.uuid)) {
1328                         rc = -EINVAL;
1329                         goto out;
1330                 }
1331
1332                 obd_str2uuid(&obd_uuid,  lustre_cfg_buf(lcfg, 1));
1333
1334                 if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", &index) != 1) {
1335                         rc = -EINVAL;
1336                         goto out;
1337                 }
1338                 if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", &gen) != 1) {
1339                         rc = -EINVAL;
1340                         goto out;
1341                 }
1342                 rc = lmv_add_target(obd, &obd_uuid, index, gen);
1343                 goto out;
1344         default:
1345                 CERROR("Unknown command: %d\n", lcfg->lcfg_command);
1346                 rc = -EINVAL;
1347                 goto out;
1348         }
1349 out:
1350         return rc;
1351 }
1352
1353 static int lmv_statfs(const struct lu_env *env, struct obd_export *exp,
1354                       struct obd_statfs *osfs, __u64 max_age, __u32 flags)
1355 {
1356         struct obd_device     *obd = class_exp2obd(exp);
1357         struct lmv_obd  *lmv = &obd->u.lmv;
1358         struct obd_statfs     *temp;
1359         int                 rc = 0;
1360         int                 i;
1361
1362         rc = lmv_check_connect(obd);
1363         if (rc)
1364                 return rc;
1365
1366         temp = kzalloc(sizeof(*temp), GFP_NOFS);
1367         if (!temp)
1368                 return -ENOMEM;
1369
1370         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1371                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp)
1372                         continue;
1373
1374                 rc = obd_statfs(env, lmv->tgts[i]->ltd_exp, temp,
1375                                 max_age, flags);
1376                 if (rc) {
1377                         CERROR("can't stat MDS #%d (%s), error %d\n", i,
1378                                lmv->tgts[i]->ltd_exp->exp_obd->obd_name,
1379                                rc);
1380                         goto out_free_temp;
1381                 }
1382
1383                 if (i == 0) {
1384                         *osfs = *temp;
1385                         /* If the statfs is from mount, it will needs
1386                          * retrieve necessary information from MDT0.
1387                          * i.e. mount does not need the merged osfs
1388                          * from all of MDT.
1389                          * And also clients can be mounted as long as
1390                          * MDT0 is in service
1391                          */
1392                         if (flags & OBD_STATFS_FOR_MDT0)
1393                                 goto out_free_temp;
1394                 } else {
1395                         osfs->os_bavail += temp->os_bavail;
1396                         osfs->os_blocks += temp->os_blocks;
1397                         osfs->os_ffree += temp->os_ffree;
1398                         osfs->os_files += temp->os_files;
1399                 }
1400         }
1401
1402 out_free_temp:
1403         kfree(temp);
1404         return rc;
1405 }
1406
1407 static int lmv_getstatus(struct obd_export *exp,
1408                          struct lu_fid *fid)
1409 {
1410         struct obd_device    *obd = exp->exp_obd;
1411         struct lmv_obd       *lmv = &obd->u.lmv;
1412         int                rc;
1413
1414         rc = lmv_check_connect(obd);
1415         if (rc)
1416                 return rc;
1417
1418         rc = md_getstatus(lmv->tgts[0]->ltd_exp, fid);
1419         return rc;
1420 }
1421
1422 static int lmv_getxattr(struct obd_export *exp, const struct lu_fid *fid,
1423                         u64 valid, const char *name,
1424                         const char *input, int input_size, int output_size,
1425                         int flags, struct ptlrpc_request **request)
1426 {
1427         struct obd_device      *obd = exp->exp_obd;
1428         struct lmv_obd   *lmv = &obd->u.lmv;
1429         struct lmv_tgt_desc    *tgt;
1430         int                  rc;
1431
1432         rc = lmv_check_connect(obd);
1433         if (rc)
1434                 return rc;
1435
1436         tgt = lmv_find_target(lmv, fid);
1437         if (IS_ERR(tgt))
1438                 return PTR_ERR(tgt);
1439
1440         rc = md_getxattr(tgt->ltd_exp, fid, valid, name, input,
1441                          input_size, output_size, flags, request);
1442
1443         return rc;
1444 }
1445
1446 static int lmv_setxattr(struct obd_export *exp, const struct lu_fid *fid,
1447                         u64 valid, const char *name,
1448                         const char *input, int input_size, int output_size,
1449                         int flags, __u32 suppgid,
1450                         struct ptlrpc_request **request)
1451 {
1452         struct obd_device      *obd = exp->exp_obd;
1453         struct lmv_obd   *lmv = &obd->u.lmv;
1454         struct lmv_tgt_desc    *tgt;
1455         int                  rc;
1456
1457         rc = lmv_check_connect(obd);
1458         if (rc)
1459                 return rc;
1460
1461         tgt = lmv_find_target(lmv, fid);
1462         if (IS_ERR(tgt))
1463                 return PTR_ERR(tgt);
1464
1465         rc = md_setxattr(tgt->ltd_exp, fid, valid, name, input,
1466                          input_size, output_size, flags, suppgid,
1467                          request);
1468
1469         return rc;
1470 }
1471
1472 static int lmv_getattr(struct obd_export *exp, struct md_op_data *op_data,
1473                        struct ptlrpc_request **request)
1474 {
1475         struct obd_device       *obd = exp->exp_obd;
1476         struct lmv_obd    *lmv = &obd->u.lmv;
1477         struct lmv_tgt_desc     *tgt;
1478         int                   rc;
1479
1480         rc = lmv_check_connect(obd);
1481         if (rc)
1482                 return rc;
1483
1484         tgt = lmv_find_target(lmv, &op_data->op_fid1);
1485         if (IS_ERR(tgt))
1486                 return PTR_ERR(tgt);
1487
1488         if (op_data->op_flags & MF_GET_MDT_IDX) {
1489                 op_data->op_mds = tgt->ltd_idx;
1490                 return 0;
1491         }
1492
1493         rc = md_getattr(tgt->ltd_exp, op_data, request);
1494
1495         return rc;
1496 }
1497
1498 static int lmv_null_inode(struct obd_export *exp, const struct lu_fid *fid)
1499 {
1500         struct obd_device   *obd = exp->exp_obd;
1501         struct lmv_obd      *lmv = &obd->u.lmv;
1502         int               i;
1503         int               rc;
1504
1505         rc = lmv_check_connect(obd);
1506         if (rc)
1507                 return rc;
1508
1509         CDEBUG(D_INODE, "CBDATA for "DFID"\n", PFID(fid));
1510
1511         /*
1512          * With DNE every object can have two locks in different namespaces:
1513          * lookup lock in space of MDT storing direntry and update/open lock in
1514          * space of MDT storing inode.
1515          */
1516         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1517                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp)
1518                         continue;
1519                 md_null_inode(lmv->tgts[i]->ltd_exp, fid);
1520         }
1521
1522         return 0;
1523 }
1524
1525 static int lmv_find_cbdata(struct obd_export *exp, const struct lu_fid *fid,
1526                            ldlm_iterator_t it, void *data)
1527 {
1528         struct obd_device   *obd = exp->exp_obd;
1529         struct lmv_obd      *lmv = &obd->u.lmv;
1530         int               i;
1531         int               rc;
1532
1533         rc = lmv_check_connect(obd);
1534         if (rc)
1535                 return rc;
1536
1537         CDEBUG(D_INODE, "CBDATA for "DFID"\n", PFID(fid));
1538
1539         /*
1540          * With DNE every object can have two locks in different namespaces:
1541          * lookup lock in space of MDT storing direntry and update/open lock in
1542          * space of MDT storing inode.
1543          */
1544         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1545                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp)
1546                         continue;
1547                 rc = md_find_cbdata(lmv->tgts[i]->ltd_exp, fid, it, data);
1548                 if (rc)
1549                         return rc;
1550         }
1551
1552         return rc;
1553 }
1554
1555 static int lmv_close(struct obd_export *exp, struct md_op_data *op_data,
1556                      struct md_open_data *mod, struct ptlrpc_request **request)
1557 {
1558         struct obd_device     *obd = exp->exp_obd;
1559         struct lmv_obd  *lmv = &obd->u.lmv;
1560         struct lmv_tgt_desc   *tgt;
1561         int                 rc;
1562
1563         rc = lmv_check_connect(obd);
1564         if (rc)
1565                 return rc;
1566
1567         tgt = lmv_find_target(lmv, &op_data->op_fid1);
1568         if (IS_ERR(tgt))
1569                 return PTR_ERR(tgt);
1570
1571         CDEBUG(D_INODE, "CLOSE "DFID"\n", PFID(&op_data->op_fid1));
1572         rc = md_close(tgt->ltd_exp, op_data, mod, request);
1573         return rc;
1574 }
1575
1576 struct lmv_tgt_desc
1577 *lmv_locate_mds(struct lmv_obd *lmv, struct md_op_data *op_data,
1578                 struct lu_fid *fid)
1579 {
1580         struct lmv_tgt_desc *tgt;
1581
1582         tgt = lmv_find_target(lmv, fid);
1583         if (IS_ERR(tgt))
1584                 return tgt;
1585
1586         op_data->op_mds = tgt->ltd_idx;
1587
1588         return tgt;
1589 }
1590
1591 static int lmv_create(struct obd_export *exp, struct md_op_data *op_data,
1592                       const void *data, int datalen, int mode, __u32 uid,
1593                       __u32 gid, cfs_cap_t cap_effective, __u64 rdev,
1594                       struct ptlrpc_request **request)
1595 {
1596         struct obd_device       *obd = exp->exp_obd;
1597         struct lmv_obd    *lmv = &obd->u.lmv;
1598         struct lmv_tgt_desc     *tgt;
1599         int                   rc;
1600
1601         rc = lmv_check_connect(obd);
1602         if (rc)
1603                 return rc;
1604
1605         if (!lmv->desc.ld_active_tgt_count)
1606                 return -EIO;
1607
1608         tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1609         if (IS_ERR(tgt))
1610                 return PTR_ERR(tgt);
1611
1612         rc = lmv_fid_alloc(exp, &op_data->op_fid2, op_data);
1613         if (rc)
1614                 return rc;
1615
1616         CDEBUG(D_INODE, "CREATE '%*s' on "DFID" -> mds #%x\n",
1617                op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
1618                op_data->op_mds);
1619
1620         op_data->op_flags |= MF_MDC_CANCEL_FID1;
1621         rc = md_create(tgt->ltd_exp, op_data, data, datalen, mode, uid, gid,
1622                        cap_effective, rdev, request);
1623
1624         if (rc == 0) {
1625                 if (!*request)
1626                         return rc;
1627                 CDEBUG(D_INODE, "Created - "DFID"\n", PFID(&op_data->op_fid2));
1628         }
1629         return rc;
1630 }
1631
1632 static int lmv_done_writing(struct obd_export *exp,
1633                             struct md_op_data *op_data,
1634                             struct md_open_data *mod)
1635 {
1636         struct obd_device     *obd = exp->exp_obd;
1637         struct lmv_obd  *lmv = &obd->u.lmv;
1638         struct lmv_tgt_desc   *tgt;
1639         int                 rc;
1640
1641         rc = lmv_check_connect(obd);
1642         if (rc)
1643                 return rc;
1644
1645         tgt = lmv_find_target(lmv, &op_data->op_fid1);
1646         if (IS_ERR(tgt))
1647                 return PTR_ERR(tgt);
1648
1649         rc = md_done_writing(tgt->ltd_exp, op_data, mod);
1650         return rc;
1651 }
1652
1653 static int
1654 lmv_enqueue_remote(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
1655                    struct lookup_intent *it, struct md_op_data *op_data,
1656                    struct lustre_handle *lockh, void *lmm, int lmmsize,
1657                    __u64 extra_lock_flags)
1658 {
1659         struct ptlrpc_request      *req = it->d.lustre.it_data;
1660         struct obd_device         *obd = exp->exp_obd;
1661         struct lmv_obd       *lmv = &obd->u.lmv;
1662         struct lustre_handle    plock;
1663         struct lmv_tgt_desc     *tgt;
1664         struct md_op_data         *rdata;
1665         struct lu_fid          fid1;
1666         struct mdt_body     *body;
1667         int                      rc = 0;
1668         int                      pmode;
1669
1670         body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
1671
1672         if (!(body->valid & OBD_MD_MDS))
1673                 return 0;
1674
1675         CDEBUG(D_INODE, "REMOTE_ENQUEUE '%s' on "DFID" -> "DFID"\n",
1676                LL_IT2STR(it), PFID(&op_data->op_fid1), PFID(&body->fid1));
1677
1678         /*
1679          * We got LOOKUP lock, but we really need attrs.
1680          */
1681         pmode = it->d.lustre.it_lock_mode;
1682         LASSERT(pmode != 0);
1683         memcpy(&plock, lockh, sizeof(plock));
1684         it->d.lustre.it_lock_mode = 0;
1685         it->d.lustre.it_data = NULL;
1686         fid1 = body->fid1;
1687
1688         ptlrpc_req_finished(req);
1689
1690         tgt = lmv_find_target(lmv, &fid1);
1691         if (IS_ERR(tgt)) {
1692                 rc = PTR_ERR(tgt);
1693                 goto out;
1694         }
1695
1696         rdata = kzalloc(sizeof(*rdata), GFP_NOFS);
1697         if (!rdata) {
1698                 rc = -ENOMEM;
1699                 goto out;
1700         }
1701
1702         rdata->op_fid1 = fid1;
1703         rdata->op_bias = MDS_CROSS_REF;
1704
1705         rc = md_enqueue(tgt->ltd_exp, einfo, it, rdata, lockh,
1706                         lmm, lmmsize, NULL, extra_lock_flags);
1707         kfree(rdata);
1708 out:
1709         ldlm_lock_decref(&plock, pmode);
1710         return rc;
1711 }
1712
1713 static int
1714 lmv_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
1715             struct lookup_intent *it, struct md_op_data *op_data,
1716             struct lustre_handle *lockh, void *lmm, int lmmsize,
1717             struct ptlrpc_request **req, __u64 extra_lock_flags)
1718 {
1719         struct obd_device       *obd = exp->exp_obd;
1720         struct lmv_obd     *lmv = &obd->u.lmv;
1721         struct lmv_tgt_desc      *tgt;
1722         int                    rc;
1723
1724         rc = lmv_check_connect(obd);
1725         if (rc)
1726                 return rc;
1727
1728         CDEBUG(D_INODE, "ENQUEUE '%s' on "DFID"\n",
1729                LL_IT2STR(it), PFID(&op_data->op_fid1));
1730
1731         tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1732         if (IS_ERR(tgt))
1733                 return PTR_ERR(tgt);
1734
1735         CDEBUG(D_INODE, "ENQUEUE '%s' on "DFID" -> mds #%d\n",
1736                LL_IT2STR(it), PFID(&op_data->op_fid1), tgt->ltd_idx);
1737
1738         rc = md_enqueue(tgt->ltd_exp, einfo, it, op_data, lockh,
1739                         lmm, lmmsize, req, extra_lock_flags);
1740
1741         if (rc == 0 && it && it->it_op == IT_OPEN) {
1742                 rc = lmv_enqueue_remote(exp, einfo, it, op_data, lockh,
1743                                         lmm, lmmsize, extra_lock_flags);
1744         }
1745         return rc;
1746 }
1747
1748 static int
1749 lmv_getattr_name(struct obd_export *exp, struct md_op_data *op_data,
1750                  struct ptlrpc_request **request)
1751 {
1752         struct ptlrpc_request   *req = NULL;
1753         struct obd_device       *obd = exp->exp_obd;
1754         struct lmv_obd    *lmv = &obd->u.lmv;
1755         struct lmv_tgt_desc     *tgt;
1756         struct mdt_body  *body;
1757         int                   rc;
1758
1759         rc = lmv_check_connect(obd);
1760         if (rc)
1761                 return rc;
1762
1763         tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1764         if (IS_ERR(tgt))
1765                 return PTR_ERR(tgt);
1766
1767         CDEBUG(D_INODE, "GETATTR_NAME for %*s on "DFID" -> mds #%d\n",
1768                op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
1769                tgt->ltd_idx);
1770
1771         rc = md_getattr_name(tgt->ltd_exp, op_data, request);
1772         if (rc != 0)
1773                 return rc;
1774
1775         body = req_capsule_server_get(&(*request)->rq_pill,
1776                                       &RMF_MDT_BODY);
1777
1778         if (body->valid & OBD_MD_MDS) {
1779                 struct lu_fid rid = body->fid1;
1780
1781                 CDEBUG(D_INODE, "Request attrs for "DFID"\n",
1782                        PFID(&rid));
1783
1784                 tgt = lmv_find_target(lmv, &rid);
1785                 if (IS_ERR(tgt)) {
1786                         ptlrpc_req_finished(*request);
1787                         return PTR_ERR(tgt);
1788                 }
1789
1790                 op_data->op_fid1 = rid;
1791                 op_data->op_valid |= OBD_MD_FLCROSSREF;
1792                 op_data->op_namelen = 0;
1793                 op_data->op_name = NULL;
1794                 rc = md_getattr_name(tgt->ltd_exp, op_data, &req);
1795                 ptlrpc_req_finished(*request);
1796                 *request = req;
1797         }
1798
1799         return rc;
1800 }
1801
1802 #define md_op_data_fid(op_data, fl)                  \
1803         (fl == MF_MDC_CANCEL_FID1 ? &op_data->op_fid1 : \
1804          fl == MF_MDC_CANCEL_FID2 ? &op_data->op_fid2 : \
1805          fl == MF_MDC_CANCEL_FID3 ? &op_data->op_fid3 : \
1806          fl == MF_MDC_CANCEL_FID4 ? &op_data->op_fid4 : \
1807          NULL)
1808
1809 static int lmv_early_cancel(struct obd_export *exp, struct md_op_data *op_data,
1810                             int op_tgt, enum ldlm_mode mode, int bits,
1811                             int flag)
1812 {
1813         struct lu_fid     *fid = md_op_data_fid(op_data, flag);
1814         struct obd_device      *obd = exp->exp_obd;
1815         struct lmv_obd   *lmv = &obd->u.lmv;
1816         struct lmv_tgt_desc    *tgt;
1817         ldlm_policy_data_t      policy = { {0} };
1818         int                  rc = 0;
1819
1820         if (!fid_is_sane(fid))
1821                 return 0;
1822
1823         tgt = lmv_find_target(lmv, fid);
1824         if (IS_ERR(tgt))
1825                 return PTR_ERR(tgt);
1826
1827         if (tgt->ltd_idx != op_tgt) {
1828                 CDEBUG(D_INODE, "EARLY_CANCEL on "DFID"\n", PFID(fid));
1829                 policy.l_inodebits.bits = bits;
1830                 rc = md_cancel_unused(tgt->ltd_exp, fid, &policy,
1831                                       mode, LCF_ASYNC, NULL);
1832         } else {
1833                 CDEBUG(D_INODE,
1834                        "EARLY_CANCEL skip operation target %d on "DFID"\n",
1835                        op_tgt, PFID(fid));
1836                 op_data->op_flags |= flag;
1837                 rc = 0;
1838         }
1839
1840         return rc;
1841 }
1842
1843 /*
1844  * llite passes fid of an target inode in op_data->op_fid1 and id of directory in
1845  * op_data->op_fid2
1846  */
1847 static int lmv_link(struct obd_export *exp, struct md_op_data *op_data,
1848                     struct ptlrpc_request **request)
1849 {
1850         struct obd_device       *obd = exp->exp_obd;
1851         struct lmv_obd    *lmv = &obd->u.lmv;
1852         struct lmv_tgt_desc     *tgt;
1853         int                   rc;
1854
1855         rc = lmv_check_connect(obd);
1856         if (rc)
1857                 return rc;
1858
1859         LASSERT(op_data->op_namelen != 0);
1860
1861         CDEBUG(D_INODE, "LINK "DFID":%*s to "DFID"\n",
1862                PFID(&op_data->op_fid2), op_data->op_namelen,
1863                op_data->op_name, PFID(&op_data->op_fid1));
1864
1865         op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
1866         op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
1867         op_data->op_cap = cfs_curproc_cap_pack();
1868         tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
1869         if (IS_ERR(tgt))
1870                 return PTR_ERR(tgt);
1871
1872         /*
1873          * Cancel UPDATE lock on child (fid1).
1874          */
1875         op_data->op_flags |= MF_MDC_CANCEL_FID2;
1876         rc = lmv_early_cancel(exp, op_data, tgt->ltd_idx, LCK_EX,
1877                               MDS_INODELOCK_UPDATE, MF_MDC_CANCEL_FID1);
1878         if (rc != 0)
1879                 return rc;
1880
1881         rc = md_link(tgt->ltd_exp, op_data, request);
1882
1883         return rc;
1884 }
1885
1886 static int lmv_rename(struct obd_export *exp, struct md_op_data *op_data,
1887                       const char *old, int oldlen, const char *new, int newlen,
1888                       struct ptlrpc_request **request)
1889 {
1890         struct obd_device       *obd = exp->exp_obd;
1891         struct lmv_obd    *lmv = &obd->u.lmv;
1892         struct lmv_tgt_desc     *src_tgt;
1893         struct lmv_tgt_desc     *tgt_tgt;
1894         int                     rc;
1895
1896         LASSERT(oldlen != 0);
1897
1898         CDEBUG(D_INODE, "RENAME %*s in "DFID" to %*s in "DFID"\n",
1899                oldlen, old, PFID(&op_data->op_fid1),
1900                newlen, new, PFID(&op_data->op_fid2));
1901
1902         rc = lmv_check_connect(obd);
1903         if (rc)
1904                 return rc;
1905
1906         op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
1907         op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
1908         op_data->op_cap = cfs_curproc_cap_pack();
1909         src_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1910         if (IS_ERR(src_tgt))
1911                 return PTR_ERR(src_tgt);
1912
1913         tgt_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
1914         if (IS_ERR(tgt_tgt))
1915                 return PTR_ERR(tgt_tgt);
1916         /*
1917          * LOOKUP lock on src child (fid3) should also be cancelled for
1918          * src_tgt in mdc_rename.
1919          */
1920         op_data->op_flags |= MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID3;
1921
1922         /*
1923          * Cancel UPDATE locks on tgt parent (fid2), tgt_tgt is its
1924          * own target.
1925          */
1926         rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
1927                               LCK_EX, MDS_INODELOCK_UPDATE,
1928                               MF_MDC_CANCEL_FID2);
1929
1930         /*
1931          * Cancel LOOKUP locks on tgt child (fid4) for parent tgt_tgt.
1932          */
1933         if (rc == 0) {
1934                 rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
1935                                       LCK_EX, MDS_INODELOCK_LOOKUP,
1936                                       MF_MDC_CANCEL_FID4);
1937         }
1938
1939         /*
1940          * Cancel all the locks on tgt child (fid4).
1941          */
1942         if (rc == 0)
1943                 rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
1944                                       LCK_EX, MDS_INODELOCK_FULL,
1945                                       MF_MDC_CANCEL_FID4);
1946
1947         if (rc == 0)
1948                 rc = md_rename(src_tgt->ltd_exp, op_data, old, oldlen,
1949                                new, newlen, request);
1950         return rc;
1951 }
1952
1953 static int lmv_setattr(struct obd_export *exp, struct md_op_data *op_data,
1954                        void *ea, int ealen, void *ea2, int ea2len,
1955                        struct ptlrpc_request **request,
1956                        struct md_open_data **mod)
1957 {
1958         struct obd_device       *obd = exp->exp_obd;
1959         struct lmv_obd    *lmv = &obd->u.lmv;
1960         struct lmv_tgt_desc     *tgt;
1961         int                   rc;
1962
1963         rc = lmv_check_connect(obd);
1964         if (rc)
1965                 return rc;
1966
1967         CDEBUG(D_INODE, "SETATTR for "DFID", valid 0x%x\n",
1968                PFID(&op_data->op_fid1), op_data->op_attr.ia_valid);
1969
1970         op_data->op_flags |= MF_MDC_CANCEL_FID1;
1971         tgt = lmv_find_target(lmv, &op_data->op_fid1);
1972         if (IS_ERR(tgt))
1973                 return PTR_ERR(tgt);
1974
1975         rc = md_setattr(tgt->ltd_exp, op_data, ea, ealen, ea2,
1976                         ea2len, request, mod);
1977
1978         return rc;
1979 }
1980
1981 static int lmv_sync(struct obd_export *exp, const struct lu_fid *fid,
1982                     struct ptlrpc_request **request)
1983 {
1984         struct obd_device        *obd = exp->exp_obd;
1985         struct lmv_obd      *lmv = &obd->u.lmv;
1986         struct lmv_tgt_desc       *tgt;
1987         int                     rc;
1988
1989         rc = lmv_check_connect(obd);
1990         if (rc)
1991                 return rc;
1992
1993         tgt = lmv_find_target(lmv, fid);
1994         if (IS_ERR(tgt))
1995                 return PTR_ERR(tgt);
1996
1997         rc = md_sync(tgt->ltd_exp, fid, request);
1998         return rc;
1999 }
2000
2001 /*
2002  * Adjust a set of pages, each page containing an array of lu_dirpages,
2003  * so that each page can be used as a single logical lu_dirpage.
2004  *
2005  * A lu_dirpage is laid out as follows, where s = ldp_hash_start,
2006  * e = ldp_hash_end, f = ldp_flags, p = padding, and each "ent" is a
2007  * struct lu_dirent.  It has size up to LU_PAGE_SIZE. The ldp_hash_end
2008  * value is used as a cookie to request the next lu_dirpage in a
2009  * directory listing that spans multiple pages (two in this example):
2010  *   ________
2011  *  |   |
2012  * .|--------v-------   -----.
2013  * |s|e|f|p|ent|ent| ... |ent|
2014  * '--|--------------   -----'   Each CFS_PAGE contains a single
2015  *    '------.             lu_dirpage.
2016  * .---------v-------   -----.
2017  * |s|e|f|p|ent| 0 | ... | 0 |
2018  * '-----------------   -----'
2019  *
2020  * However, on hosts where the native VM page size (PAGE_CACHE_SIZE) is
2021  * larger than LU_PAGE_SIZE, a single host page may contain multiple
2022  * lu_dirpages. After reading the lu_dirpages from the MDS, the
2023  * ldp_hash_end of the first lu_dirpage refers to the one immediately
2024  * after it in the same CFS_PAGE (arrows simplified for brevity, but
2025  * in general e0==s1, e1==s2, etc.):
2026  *
2027  * .--------------------   -----.
2028  * |s0|e0|f0|p|ent|ent| ... |ent|
2029  * |---v----------------   -----|
2030  * |s1|e1|f1|p|ent|ent| ... |ent|
2031  * |---v----------------   -----|  Here, each CFS_PAGE contains
2032  *           ...                 multiple lu_dirpages.
2033  * |---v----------------   -----|
2034  * |s'|e'|f'|p|ent|ent| ... |ent|
2035  * '---|----------------   -----'
2036  *     v
2037  * .----------------------------.
2038  * |    next CFS_PAGE       |
2039  *
2040  * This structure is transformed into a single logical lu_dirpage as follows:
2041  *
2042  * - Replace e0 with e' so the request for the next lu_dirpage gets the page
2043  *   labeled 'next CFS_PAGE'.
2044  *
2045  * - Copy the LDF_COLLIDE flag from f' to f0 to correctly reflect whether
2046  *   a hash collision with the next page exists.
2047  *
2048  * - Adjust the lde_reclen of the ending entry of each lu_dirpage to span
2049  *   to the first entry of the next lu_dirpage.
2050  */
2051 #if PAGE_CACHE_SIZE > LU_PAGE_SIZE
2052 static void lmv_adjust_dirpages(struct page **pages, int ncfspgs, int nlupgs)
2053 {
2054         int i;
2055
2056         for (i = 0; i < ncfspgs; i++) {
2057                 struct lu_dirpage       *dp = kmap(pages[i]);
2058                 struct lu_dirpage       *first = dp;
2059                 struct lu_dirent        *end_dirent = NULL;
2060                 struct lu_dirent        *ent;
2061                 __u64                   hash_end = dp->ldp_hash_end;
2062                 __u32                   flags = dp->ldp_flags;
2063
2064                 while (--nlupgs > 0) {
2065                         ent = lu_dirent_start(dp);
2066                         for (end_dirent = ent; ent;
2067                              end_dirent = ent, ent = lu_dirent_next(ent))
2068                                 ;
2069
2070                         /* Advance dp to next lu_dirpage. */
2071                         dp = (struct lu_dirpage *)((char *)dp + LU_PAGE_SIZE);
2072
2073                         /* Check if we've reached the end of the CFS_PAGE. */
2074                         if (!((unsigned long)dp & ~CFS_PAGE_MASK))
2075                                 break;
2076
2077                         /* Save the hash and flags of this lu_dirpage. */
2078                         hash_end = dp->ldp_hash_end;
2079                         flags = dp->ldp_flags;
2080
2081                         /* Check if lu_dirpage contains no entries. */
2082                         if (!end_dirent)
2083                                 break;
2084
2085                         /* Enlarge the end entry lde_reclen from 0 to
2086                          * first entry of next lu_dirpage.
2087                          */
2088                         LASSERT(le16_to_cpu(end_dirent->lde_reclen) == 0);
2089                         end_dirent->lde_reclen =
2090                                 cpu_to_le16((char *)(dp->ldp_entries) -
2091                                             (char *)end_dirent);
2092                 }
2093
2094                 first->ldp_hash_end = hash_end;
2095                 first->ldp_flags &= ~cpu_to_le32(LDF_COLLIDE);
2096                 first->ldp_flags |= flags & cpu_to_le32(LDF_COLLIDE);
2097
2098                 kunmap(pages[i]);
2099         }
2100         LASSERTF(nlupgs == 0, "left = %d", nlupgs);
2101 }
2102 #else
2103 #define lmv_adjust_dirpages(pages, ncfspgs, nlupgs) do {} while (0)
2104 #endif  /* PAGE_CACHE_SIZE > LU_PAGE_SIZE */
2105
2106 static int lmv_readpage(struct obd_export *exp, struct md_op_data *op_data,
2107                         struct page **pages, struct ptlrpc_request **request)
2108 {
2109         struct obd_device       *obd = exp->exp_obd;
2110         struct lmv_obd          *lmv = &obd->u.lmv;
2111         __u64                   offset = op_data->op_offset;
2112         int                     rc;
2113         int                     ncfspgs; /* pages read in PAGE_CACHE_SIZE */
2114         int                     nlupgs; /* pages read in LU_PAGE_SIZE */
2115         struct lmv_tgt_desc     *tgt;
2116
2117         rc = lmv_check_connect(obd);
2118         if (rc)
2119                 return rc;
2120
2121         CDEBUG(D_INODE, "READPAGE at %#llx from "DFID"\n",
2122                offset, PFID(&op_data->op_fid1));
2123
2124         tgt = lmv_find_target(lmv, &op_data->op_fid1);
2125         if (IS_ERR(tgt))
2126                 return PTR_ERR(tgt);
2127
2128         rc = md_readpage(tgt->ltd_exp, op_data, pages, request);
2129         if (rc != 0)
2130                 return rc;
2131
2132         ncfspgs = ((*request)->rq_bulk->bd_nob_transferred + PAGE_CACHE_SIZE - 1)
2133                  >> PAGE_CACHE_SHIFT;
2134         nlupgs = (*request)->rq_bulk->bd_nob_transferred >> LU_PAGE_SHIFT;
2135         LASSERT(!((*request)->rq_bulk->bd_nob_transferred & ~LU_PAGE_MASK));
2136         LASSERT(ncfspgs > 0 && ncfspgs <= op_data->op_npages);
2137
2138         CDEBUG(D_INODE, "read %d(%d)/%d pages\n", ncfspgs, nlupgs,
2139                op_data->op_npages);
2140
2141         lmv_adjust_dirpages(pages, ncfspgs, nlupgs);
2142
2143         return rc;
2144 }
2145
2146 static int lmv_unlink(struct obd_export *exp, struct md_op_data *op_data,
2147                       struct ptlrpc_request **request)
2148 {
2149         struct obd_device       *obd = exp->exp_obd;
2150         struct lmv_obd    *lmv = &obd->u.lmv;
2151         struct lmv_tgt_desc     *tgt = NULL;
2152         struct mdt_body         *body;
2153         int                  rc;
2154
2155         rc = lmv_check_connect(obd);
2156         if (rc)
2157                 return rc;
2158 retry:
2159         /* Send unlink requests to the MDT where the child is located */
2160         if (likely(!fid_is_zero(&op_data->op_fid2)))
2161                 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
2162         else
2163                 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
2164         if (IS_ERR(tgt))
2165                 return PTR_ERR(tgt);
2166
2167         op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
2168         op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
2169         op_data->op_cap = cfs_curproc_cap_pack();
2170
2171         /*
2172          * If child's fid is given, cancel unused locks for it if it is from
2173          * another export than parent.
2174          *
2175          * LOOKUP lock for child (fid3) should also be cancelled on parent
2176          * tgt_tgt in mdc_unlink().
2177          */
2178         op_data->op_flags |= MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID3;
2179
2180         /*
2181          * Cancel FULL locks on child (fid3).
2182          */
2183         rc = lmv_early_cancel(exp, op_data, tgt->ltd_idx, LCK_EX,
2184                               MDS_INODELOCK_FULL, MF_MDC_CANCEL_FID3);
2185
2186         if (rc != 0)
2187                 return rc;
2188
2189         CDEBUG(D_INODE, "unlink with fid="DFID"/"DFID" -> mds #%d\n",
2190                PFID(&op_data->op_fid1), PFID(&op_data->op_fid2), tgt->ltd_idx);
2191
2192         rc = md_unlink(tgt->ltd_exp, op_data, request);
2193         if (rc != 0 && rc != -EREMOTE)
2194                 return rc;
2195
2196         body = req_capsule_server_get(&(*request)->rq_pill, &RMF_MDT_BODY);
2197         if (!body)
2198                 return -EPROTO;
2199
2200         /* Not cross-ref case, just get out of here. */
2201         if (likely(!(body->valid & OBD_MD_MDS)))
2202                 return 0;
2203
2204         CDEBUG(D_INODE, "%s: try unlink to another MDT for "DFID"\n",
2205                exp->exp_obd->obd_name, PFID(&body->fid1));
2206
2207         /* This is a remote object, try remote MDT, Note: it may
2208          * try more than 1 time here, Considering following case
2209          * /mnt/lustre is root on MDT0, remote1 is on MDT1
2210          * 1. Initially A does not know where remote1 is, it send
2211          *    unlink RPC to MDT0, MDT0 return -EREMOTE, it will
2212          *    resend unlink RPC to MDT1 (retry 1st time).
2213          *
2214          * 2. During the unlink RPC in flight,
2215          *    client B mv /mnt/lustre/remote1 /mnt/lustre/remote2
2216          *    and create new remote1, but on MDT0
2217          *
2218          * 3. MDT1 get unlink RPC(from A), then do remote lock on
2219          *    /mnt/lustre, then lookup get fid of remote1, and find
2220          *    it is remote dir again, and replay -EREMOTE again.
2221          *
2222          * 4. Then A will resend unlink RPC to MDT0. (retry 2nd times).
2223          *
2224          * In theory, it might try unlimited time here, but it should
2225          * be very rare case.
2226          */
2227         op_data->op_fid2 = body->fid1;
2228         ptlrpc_req_finished(*request);
2229         *request = NULL;
2230
2231         goto retry;
2232 }
2233
2234 static int lmv_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
2235 {
2236         struct lmv_obd *lmv = &obd->u.lmv;
2237
2238         switch (stage) {
2239         case OBD_CLEANUP_EARLY:
2240                 /* XXX: here should be calling obd_precleanup() down to
2241                  * stack.
2242                  */
2243                 break;
2244         case OBD_CLEANUP_EXPORTS:
2245                 fld_client_debugfs_fini(&lmv->lmv_fld);
2246                 lprocfs_obd_cleanup(obd);
2247                 break;
2248         default:
2249                 break;
2250         }
2251         return 0;
2252 }
2253
2254 static int lmv_get_info(const struct lu_env *env, struct obd_export *exp,
2255                         __u32 keylen, void *key, __u32 *vallen, void *val,
2256                         struct lov_stripe_md *lsm)
2257 {
2258         struct obd_device       *obd;
2259         struct lmv_obd    *lmv;
2260         int                   rc = 0;
2261
2262         obd = class_exp2obd(exp);
2263         if (!obd) {
2264                 CDEBUG(D_IOCTL, "Invalid client cookie %#llx\n",
2265                        exp->exp_handle.h_cookie);
2266                 return -EINVAL;
2267         }
2268
2269         lmv = &obd->u.lmv;
2270         if (keylen >= strlen("remote_flag") && !strcmp(key, "remote_flag")) {
2271                 struct lmv_tgt_desc *tgt;
2272                 int i;
2273
2274                 rc = lmv_check_connect(obd);
2275                 if (rc)
2276                         return rc;
2277
2278                 LASSERT(*vallen == sizeof(__u32));
2279                 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2280                         tgt = lmv->tgts[i];
2281                         /*
2282                          * All tgts should be connected when this gets called.
2283                          */
2284                         if (!tgt || !tgt->ltd_exp)
2285                                 continue;
2286
2287                         if (!obd_get_info(env, tgt->ltd_exp, keylen, key,
2288                                           vallen, val, NULL))
2289                                 return 0;
2290                 }
2291                 return -EINVAL;
2292         } else if (KEY_IS(KEY_MAX_EASIZE) ||
2293                    KEY_IS(KEY_DEFAULT_EASIZE) ||
2294                    KEY_IS(KEY_CONN_DATA)) {
2295                 rc = lmv_check_connect(obd);
2296                 if (rc)
2297                         return rc;
2298
2299                 /*
2300                  * Forwarding this request to first MDS, it should know LOV
2301                  * desc.
2302                  */
2303                 rc = obd_get_info(env, lmv->tgts[0]->ltd_exp, keylen, key,
2304                                   vallen, val, NULL);
2305                 if (!rc && KEY_IS(KEY_CONN_DATA))
2306                         exp->exp_connect_data = *(struct obd_connect_data *)val;
2307                 return rc;
2308         } else if (KEY_IS(KEY_TGT_COUNT)) {
2309                 *((int *)val) = lmv->desc.ld_tgt_count;
2310                 return 0;
2311         }
2312
2313         CDEBUG(D_IOCTL, "Invalid key\n");
2314         return -EINVAL;
2315 }
2316
2317 static int lmv_set_info_async(const struct lu_env *env, struct obd_export *exp,
2318                               u32 keylen, void *key, u32 vallen,
2319                               void *val, struct ptlrpc_request_set *set)
2320 {
2321         struct lmv_tgt_desc    *tgt;
2322         struct obd_device      *obd;
2323         struct lmv_obd   *lmv;
2324         int rc = 0;
2325
2326         obd = class_exp2obd(exp);
2327         if (!obd) {
2328                 CDEBUG(D_IOCTL, "Invalid client cookie %#llx\n",
2329                        exp->exp_handle.h_cookie);
2330                 return -EINVAL;
2331         }
2332         lmv = &obd->u.lmv;
2333
2334         if (KEY_IS(KEY_READ_ONLY) || KEY_IS(KEY_FLUSH_CTX)) {
2335                 int i, err = 0;
2336
2337                 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2338                         tgt = lmv->tgts[i];
2339
2340                         if (!tgt || !tgt->ltd_exp)
2341                                 continue;
2342
2343                         err = obd_set_info_async(env, tgt->ltd_exp,
2344                                                  keylen, key, vallen, val, set);
2345                         if (err && rc == 0)
2346                                 rc = err;
2347                 }
2348
2349                 return rc;
2350         }
2351
2352         return -EINVAL;
2353 }
2354
2355 static int lmv_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
2356                       struct lov_stripe_md *lsm)
2357 {
2358         struct obd_device        *obd = class_exp2obd(exp);
2359         struct lmv_obd      *lmv = &obd->u.lmv;
2360         struct lmv_stripe_md      *meap;
2361         struct lmv_stripe_md      *lsmp;
2362         int                     mea_size;
2363         int                     i;
2364
2365         mea_size = lmv_get_easize(lmv);
2366         if (!lmmp)
2367                 return mea_size;
2368
2369         if (*lmmp && !lsm) {
2370                 kvfree(*lmmp);
2371                 *lmmp = NULL;
2372                 return 0;
2373         }
2374
2375         if (!*lmmp) {
2376                 *lmmp = libcfs_kvzalloc(mea_size, GFP_NOFS);
2377                 if (!*lmmp)
2378                         return -ENOMEM;
2379         }
2380
2381         if (!lsm)
2382                 return mea_size;
2383
2384         lsmp = (struct lmv_stripe_md *)lsm;
2385         meap = (struct lmv_stripe_md *)*lmmp;
2386
2387         if (lsmp->mea_magic != MEA_MAGIC_LAST_CHAR &&
2388             lsmp->mea_magic != MEA_MAGIC_ALL_CHARS)
2389                 return -EINVAL;
2390
2391         meap->mea_magic = cpu_to_le32(lsmp->mea_magic);
2392         meap->mea_count = cpu_to_le32(lsmp->mea_count);
2393         meap->mea_master = cpu_to_le32(lsmp->mea_master);
2394
2395         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2396                 meap->mea_ids[i] = lsmp->mea_ids[i];
2397                 fid_cpu_to_le(&meap->mea_ids[i], &lsmp->mea_ids[i]);
2398         }
2399
2400         return mea_size;
2401 }
2402
2403 static int lmv_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
2404                         struct lov_mds_md *lmm, int lmm_size)
2405 {
2406         struct obd_device         *obd = class_exp2obd(exp);
2407         struct lmv_stripe_md      **tmea = (struct lmv_stripe_md **)lsmp;
2408         struct lmv_stripe_md       *mea = (struct lmv_stripe_md *)lmm;
2409         struct lmv_obd       *lmv = &obd->u.lmv;
2410         int                      mea_size;
2411         int                      i;
2412         __u32                  magic;
2413
2414         mea_size = lmv_get_easize(lmv);
2415         if (!lsmp)
2416                 return mea_size;
2417
2418         if (*lsmp && !lmm) {
2419                 kvfree(*tmea);
2420                 *lsmp = NULL;
2421                 return 0;
2422         }
2423
2424         LASSERT(mea_size == lmm_size);
2425
2426         *tmea = libcfs_kvzalloc(mea_size, GFP_NOFS);
2427         if (!*tmea)
2428                 return -ENOMEM;
2429
2430         if (!lmm)
2431                 return mea_size;
2432
2433         if (mea->mea_magic == MEA_MAGIC_LAST_CHAR ||
2434             mea->mea_magic == MEA_MAGIC_ALL_CHARS ||
2435             mea->mea_magic == MEA_MAGIC_HASH_SEGMENT) {
2436                 magic = le32_to_cpu(mea->mea_magic);
2437         } else {
2438                 /*
2439                  * Old mea is not handled here.
2440                  */
2441                 CERROR("Old not supportable EA is found\n");
2442                 LBUG();
2443         }
2444
2445         (*tmea)->mea_magic = magic;
2446         (*tmea)->mea_count = le32_to_cpu(mea->mea_count);
2447         (*tmea)->mea_master = le32_to_cpu(mea->mea_master);
2448
2449         for (i = 0; i < (*tmea)->mea_count; i++) {
2450                 (*tmea)->mea_ids[i] = mea->mea_ids[i];
2451                 fid_le_to_cpu(&(*tmea)->mea_ids[i], &(*tmea)->mea_ids[i]);
2452         }
2453         return mea_size;
2454 }
2455
2456 static int lmv_cancel_unused(struct obd_export *exp, const struct lu_fid *fid,
2457                              ldlm_policy_data_t *policy, enum ldlm_mode mode,
2458                              enum ldlm_cancel_flags flags, void *opaque)
2459 {
2460         struct obd_device       *obd = exp->exp_obd;
2461         struct lmv_obd    *lmv = &obd->u.lmv;
2462         int                   rc = 0;
2463         int                   err;
2464         int                   i;
2465
2466         LASSERT(fid);
2467
2468         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2469                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp ||
2470                     lmv->tgts[i]->ltd_active == 0)
2471                         continue;
2472
2473                 err = md_cancel_unused(lmv->tgts[i]->ltd_exp, fid,
2474                                        policy, mode, flags, opaque);
2475                 if (!rc)
2476                         rc = err;
2477         }
2478         return rc;
2479 }
2480
2481 static int lmv_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
2482                              __u64 *bits)
2483 {
2484         struct lmv_obd    *lmv = &exp->exp_obd->u.lmv;
2485         int                   rc;
2486
2487         rc =  md_set_lock_data(lmv->tgts[0]->ltd_exp, lockh, data, bits);
2488         return rc;
2489 }
2490
2491 static enum ldlm_mode lmv_lock_match(struct obd_export *exp, __u64 flags,
2492                                      const struct lu_fid *fid,
2493                                      enum ldlm_type type,
2494                                      ldlm_policy_data_t *policy,
2495                                      enum ldlm_mode mode,
2496                                      struct lustre_handle *lockh)
2497 {
2498         struct obd_device       *obd = exp->exp_obd;
2499         struct lmv_obd    *lmv = &obd->u.lmv;
2500         enum ldlm_mode        rc;
2501         int                   i;
2502
2503         CDEBUG(D_INODE, "Lock match for "DFID"\n", PFID(fid));
2504
2505         /*
2506          * With CMD every object can have two locks in different namespaces:
2507          * lookup lock in space of mds storing direntry and update/open lock in
2508          * space of mds storing inode. Thus we check all targets, not only that
2509          * one fid was created in.
2510          */
2511         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2512                 if (!lmv->tgts[i] || !lmv->tgts[i]->ltd_exp ||
2513                     lmv->tgts[i]->ltd_active == 0)
2514                         continue;
2515
2516                 rc = md_lock_match(lmv->tgts[i]->ltd_exp, flags, fid,
2517                                    type, policy, mode, lockh);
2518                 if (rc)
2519                         return rc;
2520         }
2521
2522         return 0;
2523 }
2524
2525 static int lmv_get_lustre_md(struct obd_export *exp,
2526                              struct ptlrpc_request *req,
2527                              struct obd_export *dt_exp,
2528                              struct obd_export *md_exp,
2529                              struct lustre_md *md)
2530 {
2531         struct lmv_obd    *lmv = &exp->exp_obd->u.lmv;
2532
2533         return md_get_lustre_md(lmv->tgts[0]->ltd_exp, req, dt_exp, md_exp, md);
2534 }
2535
2536 static int lmv_free_lustre_md(struct obd_export *exp, struct lustre_md *md)
2537 {
2538         struct obd_device       *obd = exp->exp_obd;
2539         struct lmv_obd    *lmv = &obd->u.lmv;
2540
2541         if (md->mea)
2542                 obd_free_memmd(exp, (void *)&md->mea);
2543         return md_free_lustre_md(lmv->tgts[0]->ltd_exp, md);
2544 }
2545
2546 static int lmv_set_open_replay_data(struct obd_export *exp,
2547                                     struct obd_client_handle *och,
2548                                     struct lookup_intent *it)
2549 {
2550         struct obd_device       *obd = exp->exp_obd;
2551         struct lmv_obd    *lmv = &obd->u.lmv;
2552         struct lmv_tgt_desc     *tgt;
2553
2554         tgt = lmv_find_target(lmv, &och->och_fid);
2555         if (IS_ERR(tgt))
2556                 return PTR_ERR(tgt);
2557
2558         return md_set_open_replay_data(tgt->ltd_exp, och, it);
2559 }
2560
2561 static int lmv_clear_open_replay_data(struct obd_export *exp,
2562                                       struct obd_client_handle *och)
2563 {
2564         struct obd_device       *obd = exp->exp_obd;
2565         struct lmv_obd    *lmv = &obd->u.lmv;
2566         struct lmv_tgt_desc     *tgt;
2567
2568         tgt = lmv_find_target(lmv, &och->och_fid);
2569         if (IS_ERR(tgt))
2570                 return PTR_ERR(tgt);
2571
2572         return md_clear_open_replay_data(tgt->ltd_exp, och);
2573 }
2574
2575 static int lmv_get_remote_perm(struct obd_export *exp,
2576                                const struct lu_fid *fid,
2577                                __u32 suppgid, struct ptlrpc_request **request)
2578 {
2579         struct obd_device       *obd = exp->exp_obd;
2580         struct lmv_obd    *lmv = &obd->u.lmv;
2581         struct lmv_tgt_desc     *tgt;
2582         int                   rc;
2583
2584         rc = lmv_check_connect(obd);
2585         if (rc)
2586                 return rc;
2587
2588         tgt = lmv_find_target(lmv, fid);
2589         if (IS_ERR(tgt))
2590                 return PTR_ERR(tgt);
2591
2592         rc = md_get_remote_perm(tgt->ltd_exp, fid, suppgid, request);
2593         return rc;
2594 }
2595
2596 static int lmv_intent_getattr_async(struct obd_export *exp,
2597                                     struct md_enqueue_info *minfo,
2598                                     struct ldlm_enqueue_info *einfo)
2599 {
2600         struct md_op_data       *op_data = &minfo->mi_data;
2601         struct obd_device       *obd = exp->exp_obd;
2602         struct lmv_obd    *lmv = &obd->u.lmv;
2603         struct lmv_tgt_desc     *tgt = NULL;
2604         int                   rc;
2605
2606         rc = lmv_check_connect(obd);
2607         if (rc)
2608                 return rc;
2609
2610         tgt = lmv_find_target(lmv, &op_data->op_fid1);
2611         if (IS_ERR(tgt))
2612                 return PTR_ERR(tgt);
2613
2614         rc = md_intent_getattr_async(tgt->ltd_exp, minfo, einfo);
2615         return rc;
2616 }
2617
2618 static int lmv_revalidate_lock(struct obd_export *exp, struct lookup_intent *it,
2619                                struct lu_fid *fid, __u64 *bits)
2620 {
2621         struct obd_device       *obd = exp->exp_obd;
2622         struct lmv_obd    *lmv = &obd->u.lmv;
2623         struct lmv_tgt_desc     *tgt;
2624         int                   rc;
2625
2626         rc = lmv_check_connect(obd);
2627         if (rc)
2628                 return rc;
2629
2630         tgt = lmv_find_target(lmv, fid);
2631         if (IS_ERR(tgt))
2632                 return PTR_ERR(tgt);
2633
2634         rc = md_revalidate_lock(tgt->ltd_exp, it, fid, bits);
2635         return rc;
2636 }
2637
2638 /**
2639  * For lmv, only need to send request to master MDT, and the master MDT will
2640  * process with other slave MDTs. The only exception is Q_GETOQUOTA for which
2641  * we directly fetch data from the slave MDTs.
2642  */
2643 static int lmv_quotactl(struct obd_device *unused, struct obd_export *exp,
2644                         struct obd_quotactl *oqctl)
2645 {
2646         struct obd_device   *obd = class_exp2obd(exp);
2647         struct lmv_obd      *lmv = &obd->u.lmv;
2648         struct lmv_tgt_desc *tgt = lmv->tgts[0];
2649         int               rc = 0, i;
2650         __u64           curspace, curinodes;
2651
2652         if (!lmv->desc.ld_tgt_count || !tgt->ltd_active) {
2653                 CERROR("master lmv inactive\n");
2654                 return -EIO;
2655         }
2656
2657         if (oqctl->qc_cmd != Q_GETOQUOTA) {
2658                 rc = obd_quotactl(tgt->ltd_exp, oqctl);
2659                 return rc;
2660         }
2661
2662         curspace = curinodes = 0;
2663         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2664                 int err;
2665
2666                 tgt = lmv->tgts[i];
2667
2668                 if (!tgt || !tgt->ltd_exp || tgt->ltd_active == 0)
2669                         continue;
2670                 if (!tgt->ltd_active) {
2671                         CDEBUG(D_HA, "mdt %d is inactive.\n", i);
2672                         continue;
2673                 }
2674
2675                 err = obd_quotactl(tgt->ltd_exp, oqctl);
2676                 if (err) {
2677                         CERROR("getquota on mdt %d failed. %d\n", i, err);
2678                         if (!rc)
2679                                 rc = err;
2680                 } else {
2681                         curspace += oqctl->qc_dqblk.dqb_curspace;
2682                         curinodes += oqctl->qc_dqblk.dqb_curinodes;
2683                 }
2684         }
2685         oqctl->qc_dqblk.dqb_curspace = curspace;
2686         oqctl->qc_dqblk.dqb_curinodes = curinodes;
2687
2688         return rc;
2689 }
2690
2691 static int lmv_quotacheck(struct obd_device *unused, struct obd_export *exp,
2692                           struct obd_quotactl *oqctl)
2693 {
2694         struct obd_device   *obd = class_exp2obd(exp);
2695         struct lmv_obd      *lmv = &obd->u.lmv;
2696         struct lmv_tgt_desc *tgt;
2697         int               i, rc = 0;
2698
2699         for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2700                 int err;
2701
2702                 tgt = lmv->tgts[i];
2703                 if (!tgt || !tgt->ltd_exp || !tgt->ltd_active) {
2704                         CERROR("lmv idx %d inactive\n", i);
2705                         return -EIO;
2706                 }
2707
2708                 err = obd_quotacheck(tgt->ltd_exp, oqctl);
2709                 if (err && !rc)
2710                         rc = err;
2711         }
2712
2713         return rc;
2714 }
2715
2716 static struct obd_ops lmv_obd_ops = {
2717         .owner          = THIS_MODULE,
2718         .setup          = lmv_setup,
2719         .cleanup        = lmv_cleanup,
2720         .precleanup     = lmv_precleanup,
2721         .process_config = lmv_process_config,
2722         .connect        = lmv_connect,
2723         .disconnect     = lmv_disconnect,
2724         .statfs         = lmv_statfs,
2725         .get_info       = lmv_get_info,
2726         .set_info_async = lmv_set_info_async,
2727         .packmd         = lmv_packmd,
2728         .unpackmd       = lmv_unpackmd,
2729         .notify         = lmv_notify,
2730         .get_uuid       = lmv_get_uuid,
2731         .iocontrol      = lmv_iocontrol,
2732         .quotacheck     = lmv_quotacheck,
2733         .quotactl       = lmv_quotactl
2734 };
2735
2736 static struct md_ops lmv_md_ops = {
2737         .getstatus              = lmv_getstatus,
2738         .null_inode             = lmv_null_inode,
2739         .find_cbdata            = lmv_find_cbdata,
2740         .close                  = lmv_close,
2741         .create                 = lmv_create,
2742         .done_writing           = lmv_done_writing,
2743         .enqueue                = lmv_enqueue,
2744         .getattr                = lmv_getattr,
2745         .getxattr               = lmv_getxattr,
2746         .getattr_name           = lmv_getattr_name,
2747         .intent_lock            = lmv_intent_lock,
2748         .link                   = lmv_link,
2749         .rename                 = lmv_rename,
2750         .setattr                = lmv_setattr,
2751         .setxattr               = lmv_setxattr,
2752         .sync                   = lmv_sync,
2753         .readpage               = lmv_readpage,
2754         .unlink                 = lmv_unlink,
2755         .init_ea_size           = lmv_init_ea_size,
2756         .cancel_unused          = lmv_cancel_unused,
2757         .set_lock_data          = lmv_set_lock_data,
2758         .lock_match             = lmv_lock_match,
2759         .get_lustre_md          = lmv_get_lustre_md,
2760         .free_lustre_md         = lmv_free_lustre_md,
2761         .set_open_replay_data   = lmv_set_open_replay_data,
2762         .clear_open_replay_data = lmv_clear_open_replay_data,
2763         .get_remote_perm        = lmv_get_remote_perm,
2764         .intent_getattr_async   = lmv_intent_getattr_async,
2765         .revalidate_lock        = lmv_revalidate_lock
2766 };
2767
2768 static int __init lmv_init(void)
2769 {
2770         struct lprocfs_static_vars lvars;
2771         int                     rc;
2772
2773         lprocfs_lmv_init_vars(&lvars);
2774
2775         rc = class_register_type(&lmv_obd_ops, &lmv_md_ops,
2776                                  LUSTRE_LMV_NAME, NULL);
2777         return rc;
2778 }
2779
2780 static void lmv_exit(void)
2781 {
2782         class_unregister_type(LUSTRE_LMV_NAME);
2783 }
2784
2785 MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
2786 MODULE_DESCRIPTION("Lustre Logical Metadata Volume");
2787 MODULE_VERSION(LUSTRE_VERSION_STRING);
2788 MODULE_LICENSE("GPL");
2789
2790 module_init(lmv_init);
2791 module_exit(lmv_exit);