kernel/drivers/staging/lustre/lustre/mdc/mdc_request.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) 2001, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2011, 2012, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  */
  36
  37 #define DEBUG_SUBSYSTEM S_MDC
  38
  39 # include <linux/module.h>
  40 # include <linux/pagemap.h>
  41 # include <linux/miscdevice.h>
  42 # include <linux/init.h>
  43 # include <linux/utsname.h>
  44
  45 #include "../include/lustre_acl.h"
  46 #include "../include/obd_class.h"
  47 #include "../include/lustre_fid.h"
  48 #include "../include/lprocfs_status.h"
  49 #include "../include/lustre_param.h"
  50 #include "../include/lustre_log.h"
  51
  52 #include "mdc_internal.h"
  53
  54 #define REQUEST_MINOR 244
  55
  56 struct mdc_renew_capa_args {
  57         struct obd_capa *ra_oc;
  58         renew_capa_cb_t  ra_cb;
  59 };
  60
  61 static int mdc_cleanup(struct obd_device *obd);
  62
  63 static int mdc_unpack_capa(struct obd_export *exp, struct ptlrpc_request *req,
  64                     const struct req_msg_field *field, struct obd_capa **oc)
  65 {
  66         struct lustre_capa *capa;
  67         struct obd_capa *c;
  68
  69         /* swabbed already in mdc_enqueue */
  70         capa = req_capsule_server_get(&req->rq_pill, field);
  71         if (capa == NULL)
  72                 return -EPROTO;
  73
  74         c = alloc_capa(CAPA_SITE_CLIENT);
  75         if (IS_ERR(c)) {
  76                 CDEBUG(D_INFO, "alloc capa failed!\n");
  77                 return PTR_ERR(c);
  78         } else {
  79                 c->c_capa = *capa;
  80                 *oc = c;
  81                 return 0;
  82         }
  83 }
  84
  85 static inline int mdc_queue_wait(struct ptlrpc_request *req)
  86 {
  87         struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
  88         int rc;
  89
  90         /* mdc_enter_request() ensures that this client has no more
  91          * than cl_max_rpcs_in_flight RPCs simultaneously inf light
  92          * against an MDT. */
  93         rc = mdc_enter_request(cli);
  94         if (rc != 0)
  95                 return rc;
  96
  97         rc = ptlrpc_queue_wait(req);
  98         mdc_exit_request(cli);
  99
 100         return rc;
 101 }
 102
 103 /* Helper that implements most of mdc_getstatus and signal_completed_replay. */
 104 /* XXX this should become mdc_get_info("key"), sending MDS_GET_INFO RPC */
 105 static int send_getstatus(struct obd_import *imp, struct lu_fid *rootfid,
 106                           struct obd_capa **pc, int level, int msg_flags)
 107 {
 108         struct ptlrpc_request *req;
 109         struct mdt_body       *body;
 110         int                 rc;
 111
 112         req = ptlrpc_request_alloc_pack(imp, &RQF_MDS_GETSTATUS,
 113                                         LUSTRE_MDS_VERSION, MDS_GETSTATUS);
 114         if (req == NULL)
 115                 return -ENOMEM;
 116
 117         mdc_pack_body(req, NULL, NULL, 0, 0, -1, 0);
 118         lustre_msg_add_flags(req->rq_reqmsg, msg_flags);
 119         req->rq_send_state = level;
 120
 121         ptlrpc_request_set_replen(req);
 122
 123         rc = ptlrpc_queue_wait(req);
 124         if (rc)
 125                 goto out;
 126
 127         body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
 128         if (body == NULL) {
 129                 rc = -EPROTO;
 130                 goto out;
 131         }
 132
 133         if (body->valid & OBD_MD_FLMDSCAPA) {
 134                 rc = mdc_unpack_capa(NULL, req, &RMF_CAPA1, pc);
 135                 if (rc)
 136                         goto out;
 137         }
 138
 139         *rootfid = body->fid1;
 140         CDEBUG(D_NET,
 141                "root fid="DFID", last_committed=%llu\n",
 142                PFID(rootfid),
 143                lustre_msg_get_last_committed(req->rq_repmsg));
 144 out:
 145         ptlrpc_req_finished(req);
 146         return rc;
 147 }
 148
 149 /* This should be mdc_get_info("rootfid") */
 150 static int mdc_getstatus(struct obd_export *exp, struct lu_fid *rootfid,
 151                   struct obd_capa **pc)
 152 {
 153         return send_getstatus(class_exp2cliimp(exp), rootfid, pc,
 154                               LUSTRE_IMP_FULL, 0);
 155 }
 156
 157 /*
 158  * This function now is known to always saying that it will receive 4 buffers
 159  * from server. Even for cases when acl_size and md_size is zero, RPC header
 160  * will contain 4 fields and RPC itself will contain zero size fields. This is
 161  * because mdt_getattr*() _always_ returns 4 fields, but if acl is not needed
 162  * and thus zero, it shrinks it, making zero size. The same story about
 163  * md_size. And this is course of problem when client waits for smaller number
 164  * of fields. This issue will be fixed later when client gets aware of RPC
 165  * layouts.  --umka
 166  */
 167 static int mdc_getattr_common(struct obd_export *exp,
 168                               struct ptlrpc_request *req)
 169 {
 170         struct req_capsule *pill = &req->rq_pill;
 171         struct mdt_body    *body;
 172         void           *eadata;
 173         int              rc;
 174
 175         /* Request message already built. */
 176         rc = ptlrpc_queue_wait(req);
 177         if (rc != 0)
 178                 return rc;
 179
 180         /* sanity check for the reply */
 181         body = req_capsule_server_get(pill, &RMF_MDT_BODY);
 182         if (body == NULL)
 183                 return -EPROTO;
 184
 185         CDEBUG(D_NET, "mode: %o\n", body->mode);
 186
 187         if (body->eadatasize != 0) {
 188                 mdc_update_max_ea_from_body(exp, body);
 189
 190                 eadata = req_capsule_server_sized_get(pill, &RMF_MDT_MD,
 191                                                       body->eadatasize);
 192                 if (eadata == NULL)
 193                         return -EPROTO;
 194         }
 195
 196         if (body->valid & OBD_MD_FLRMTPERM) {
 197                 struct mdt_remote_perm *perm;
 198
 199                 LASSERT(client_is_remote(exp));
 200                 perm = req_capsule_server_swab_get(pill, &RMF_ACL,
 201                                                 lustre_swab_mdt_remote_perm);
 202                 if (perm == NULL)
 203                         return -EPROTO;
 204         }
 205
 206         if (body->valid & OBD_MD_FLMDSCAPA) {
 207                 struct lustre_capa *capa;
 208
 209                 capa = req_capsule_server_get(pill, &RMF_CAPA1);
 210                 if (capa == NULL)
 211                         return -EPROTO;
 212         }
 213
 214         return 0;
 215 }
 216
 217 static int mdc_getattr(struct obd_export *exp, struct md_op_data *op_data,
 218                 struct ptlrpc_request **request)
 219 {
 220         struct ptlrpc_request *req;
 221         int                 rc;
 222
 223         /* Single MDS without an LMV case */
 224         if (op_data->op_flags & MF_GET_MDT_IDX) {
 225                 op_data->op_mds = 0;
 226                 return 0;
 227         }
 228         *request = NULL;
 229         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_MDS_GETATTR);
 230         if (req == NULL)
 231                 return -ENOMEM;
 232
 233         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
 234
 235         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_GETATTR);
 236         if (rc) {
 237                 ptlrpc_request_free(req);
 238                 return rc;
 239         }
 240
 241         mdc_pack_body(req, &op_data->op_fid1, op_data->op_capa1,
 242                       op_data->op_valid, op_data->op_mode, -1, 0);
 243
 244         req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
 245                              op_data->op_mode);
 246         if (op_data->op_valid & OBD_MD_FLRMTPERM) {
 247                 LASSERT(client_is_remote(exp));
 248                 req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
 249                                      sizeof(struct mdt_remote_perm));
 250         }
 251         ptlrpc_request_set_replen(req);
 252
 253         rc = mdc_getattr_common(exp, req);
 254         if (rc)
 255                 ptlrpc_req_finished(req);
 256         else
 257                 *request = req;
 258         return rc;
 259 }
 260
 261 static int mdc_getattr_name(struct obd_export *exp, struct md_op_data *op_data,
 262                      struct ptlrpc_request **request)
 263 {
 264         struct ptlrpc_request *req;
 265         int                 rc;
 266
 267         *request = NULL;
 268         req = ptlrpc_request_alloc(class_exp2cliimp(exp),
 269                                    &RQF_MDS_GETATTR_NAME);
 270         if (req == NULL)
 271                 return -ENOMEM;
 272
 273         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
 274         req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
 275                              op_data->op_namelen + 1);
 276
 277         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_GETATTR_NAME);
 278         if (rc) {
 279                 ptlrpc_request_free(req);
 280                 return rc;
 281         }
 282
 283         mdc_pack_body(req, &op_data->op_fid1, op_data->op_capa1,
 284                       op_data->op_valid, op_data->op_mode,
 285                       op_data->op_suppgids[0], 0);
 286
 287         if (op_data->op_name) {
 288                 char *name = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
 289
 290                 LASSERT(strnlen(op_data->op_name, op_data->op_namelen) ==
 291                                 op_data->op_namelen);
 292                 memcpy(name, op_data->op_name, op_data->op_namelen);
 293         }
 294
 295         req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
 296                              op_data->op_mode);
 297         ptlrpc_request_set_replen(req);
 298
 299         rc = mdc_getattr_common(exp, req);
 300         if (rc)
 301                 ptlrpc_req_finished(req);
 302         else
 303                 *request = req;
 304         return rc;
 305 }
 306
 307 static int mdc_is_subdir(struct obd_export *exp,
 308                          const struct lu_fid *pfid,
 309                          const struct lu_fid *cfid,
 310                          struct ptlrpc_request **request)
 311 {
 312         struct ptlrpc_request  *req;
 313         int                  rc;
 314
 315         *request = NULL;
 316         req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
 317                                         &RQF_MDS_IS_SUBDIR, LUSTRE_MDS_VERSION,
 318                                         MDS_IS_SUBDIR);
 319         if (req == NULL)
 320                 return -ENOMEM;
 321
 322         mdc_is_subdir_pack(req, pfid, cfid, 0);
 323         ptlrpc_request_set_replen(req);
 324
 325         rc = ptlrpc_queue_wait(req);
 326         if (rc && rc != -EREMOTE)
 327                 ptlrpc_req_finished(req);
 328         else
 329                 *request = req;
 330         return rc;
 331 }
 332
 333 static int mdc_xattr_common(struct obd_export *exp,
 334                             const struct req_format *fmt,
 335                             const struct lu_fid *fid,
 336                             struct obd_capa *oc, int opcode, u64 valid,
 337                             const char *xattr_name, const char *input,
 338                             int input_size, int output_size, int flags,
 339                             __u32 suppgid, struct ptlrpc_request **request)
 340 {
 341         struct ptlrpc_request *req;
 342         int   xattr_namelen = 0;
 343         char *tmp;
 344         int   rc;
 345
 346         *request = NULL;
 347         req = ptlrpc_request_alloc(class_exp2cliimp(exp), fmt);
 348         if (req == NULL)
 349                 return -ENOMEM;
 350
 351         mdc_set_capa_size(req, &RMF_CAPA1, oc);
 352         if (xattr_name) {
 353                 xattr_namelen = strlen(xattr_name) + 1;
 354                 req_capsule_set_size(&req->rq_pill, &RMF_NAME, RCL_CLIENT,
 355                                      xattr_namelen);
 356         }
 357         if (input_size) {
 358                 LASSERT(input);
 359                 req_capsule_set_size(&req->rq_pill, &RMF_EADATA, RCL_CLIENT,
 360                                      input_size);
 361         }
 362
 363         /* Flush local XATTR locks to get rid of a possible cancel RPC */
 364         if (opcode == MDS_REINT && fid_is_sane(fid) &&
 365             exp->exp_connect_data.ocd_ibits_known & MDS_INODELOCK_XATTR) {
 366                 LIST_HEAD(cancels);
 367                 int count;
 368
 369                 /* Without that packing would fail */
 370                 if (input_size == 0)
 371                         req_capsule_set_size(&req->rq_pill, &RMF_EADATA,
 372                                              RCL_CLIENT, 0);
 373
 374                 count = mdc_resource_get_unused(exp, fid,
 375                                                 &cancels, LCK_EX,
 376                                                 MDS_INODELOCK_XATTR);
 377
 378                 rc = mdc_prep_elc_req(exp, req, MDS_REINT, &cancels, count);
 379                 if (rc) {
 380                         ptlrpc_request_free(req);
 381                         return rc;
 382                 }
 383         } else {
 384                 rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, opcode);
 385                 if (rc) {
 386                         ptlrpc_request_free(req);
 387                         return rc;
 388                 }
 389         }
 390
 391         if (opcode == MDS_REINT) {
 392                 struct mdt_rec_setxattr *rec;
 393
 394                 CLASSERT(sizeof(struct mdt_rec_setxattr) ==
 395                          sizeof(struct mdt_rec_reint));
 396                 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
 397                 rec->sx_opcode = REINT_SETXATTR;
 398                 rec->sx_fsuid  = from_kuid(&init_user_ns, current_fsuid());
 399                 rec->sx_fsgid  = from_kgid(&init_user_ns, current_fsgid());
 400                 rec->sx_cap    = cfs_curproc_cap_pack();
 401                 rec->sx_suppgid1 = suppgid;
 402                 rec->sx_suppgid2 = -1;
 403                 rec->sx_fid    = *fid;
 404                 rec->sx_valid  = valid | OBD_MD_FLCTIME;
 405                 rec->sx_time   = get_seconds();
 406                 rec->sx_size   = output_size;
 407                 rec->sx_flags  = flags;
 408
 409                 mdc_pack_capa(req, &RMF_CAPA1, oc);
 410         } else {
 411                 mdc_pack_body(req, fid, oc, valid, output_size, suppgid, flags);
 412         }
 413
 414         if (xattr_name) {
 415                 tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
 416                 memcpy(tmp, xattr_name, xattr_namelen);
 417         }
 418         if (input_size) {
 419                 tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
 420                 memcpy(tmp, input, input_size);
 421         }
 422
 423         if (req_capsule_has_field(&req->rq_pill, &RMF_EADATA, RCL_SERVER))
 424                 req_capsule_set_size(&req->rq_pill, &RMF_EADATA,
 425                                      RCL_SERVER, output_size);
 426         ptlrpc_request_set_replen(req);
 427
 428         /* make rpc */
 429         if (opcode == MDS_REINT)
 430                 mdc_get_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
 431
 432         rc = ptlrpc_queue_wait(req);
 433
 434         if (opcode == MDS_REINT)
 435                 mdc_put_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
 436
 437         if (rc)
 438                 ptlrpc_req_finished(req);
 439         else
 440                 *request = req;
 441         return rc;
 442 }
 443
 444 static int mdc_setxattr(struct obd_export *exp, const struct lu_fid *fid,
 445                  struct obd_capa *oc, u64 valid, const char *xattr_name,
 446                  const char *input, int input_size, int output_size,
 447                  int flags, __u32 suppgid, struct ptlrpc_request **request)
 448 {
 449         return mdc_xattr_common(exp, &RQF_MDS_REINT_SETXATTR,
 450                                 fid, oc, MDS_REINT, valid, xattr_name,
 451                                 input, input_size, output_size, flags,
 452                                 suppgid, request);
 453 }
 454
 455 static int mdc_getxattr(struct obd_export *exp, const struct lu_fid *fid,
 456                  struct obd_capa *oc, u64 valid, const char *xattr_name,
 457                  const char *input, int input_size, int output_size,
 458                  int flags, struct ptlrpc_request **request)
 459 {
 460         return mdc_xattr_common(exp, &RQF_MDS_GETXATTR,
 461                                 fid, oc, MDS_GETXATTR, valid, xattr_name,
 462                                 input, input_size, output_size, flags,
 463                                 -1, request);
 464 }
 465
 466 #ifdef CONFIG_FS_POSIX_ACL
 467 static int mdc_unpack_acl(struct ptlrpc_request *req, struct lustre_md *md)
 468 {
 469         struct req_capsule     *pill = &req->rq_pill;
 470         struct mdt_body *body = md->body;
 471         struct posix_acl       *acl;
 472         void               *buf;
 473         int                  rc;
 474
 475         if (!body->aclsize)
 476                 return 0;
 477
 478         buf = req_capsule_server_sized_get(pill, &RMF_ACL, body->aclsize);
 479
 480         if (!buf)
 481                 return -EPROTO;
 482
 483         acl = posix_acl_from_xattr(&init_user_ns, buf, body->aclsize);
 484         if (acl == NULL)
 485                 return 0;
 486
 487         if (IS_ERR(acl)) {
 488                 rc = PTR_ERR(acl);
 489                 CERROR("convert xattr to acl: %d\n", rc);
 490                 return rc;
 491         }
 492
 493         rc = posix_acl_valid(acl);
 494         if (rc) {
 495                 CERROR("validate acl: %d\n", rc);
 496                 posix_acl_release(acl);
 497                 return rc;
 498         }
 499
 500         md->posix_acl = acl;
 501         return 0;
 502 }
 503 #else
 504 #define mdc_unpack_acl(req, md) 0
 505 #endif
 506
 507 int mdc_get_lustre_md(struct obd_export *exp, struct ptlrpc_request *req,
 508                       struct obd_export *dt_exp, struct obd_export *md_exp,
 509                       struct lustre_md *md)
 510 {
 511         struct req_capsule *pill = &req->rq_pill;
 512         int rc;
 513
 514         LASSERT(md);
 515         memset(md, 0, sizeof(*md));
 516
 517         md->body = req_capsule_server_get(pill, &RMF_MDT_BODY);
 518         LASSERT(md->body != NULL);
 519
 520         if (md->body->valid & OBD_MD_FLEASIZE) {
 521                 int lmmsize;
 522                 struct lov_mds_md *lmm;
 523
 524                 if (!S_ISREG(md->body->mode)) {
 525                         CDEBUG(D_INFO,
 526                                "OBD_MD_FLEASIZE set, should be a regular file, but is not\n");
 527                         rc = -EPROTO;
 528                         goto out;
 529                 }
 530
 531                 if (md->body->eadatasize == 0) {
 532                         CDEBUG(D_INFO,
 533                                "OBD_MD_FLEASIZE set, but eadatasize 0\n");
 534                         rc = -EPROTO;
 535                         goto out;
 536                 }
 537                 lmmsize = md->body->eadatasize;
 538                 lmm = req_capsule_server_sized_get(pill, &RMF_MDT_MD, lmmsize);
 539                 if (!lmm) {
 540                         rc = -EPROTO;
 541                         goto out;
 542                 }
 543
 544                 rc = obd_unpackmd(dt_exp, &md->lsm, lmm, lmmsize);
 545                 if (rc < 0)
 546                         goto out;
 547
 548                 if (rc < sizeof(*md->lsm)) {
 549                         CDEBUG(D_INFO,
 550                                "lsm size too small: rc < sizeof (*md->lsm) (%d < %d)\n",
 551                                rc, (int)sizeof(*md->lsm));
 552                         rc = -EPROTO;
 553                         goto out;
 554                 }
 555
 556         } else if (md->body->valid & OBD_MD_FLDIREA) {
 557                 int lmvsize;
 558                 struct lov_mds_md *lmv;
 559
 560                 if (!S_ISDIR(md->body->mode)) {
 561                         CDEBUG(D_INFO,
 562                                "OBD_MD_FLDIREA set, should be a directory, but is not\n");
 563                         rc = -EPROTO;
 564                         goto out;
 565                 }
 566
 567                 if (md->body->eadatasize == 0) {
 568                         CDEBUG(D_INFO,
 569                                "OBD_MD_FLDIREA is set, but eadatasize 0\n");
 570                         return -EPROTO;
 571                 }
 572                 if (md->body->valid & OBD_MD_MEA) {
 573                         lmvsize = md->body->eadatasize;
 574                         lmv = req_capsule_server_sized_get(pill, &RMF_MDT_MD,
 575                                                            lmvsize);
 576                         if (!lmv) {
 577                                 rc = -EPROTO;
 578                                 goto out;
 579                         }
 580
 581                         rc = obd_unpackmd(md_exp, (void *)&md->mea, lmv,
 582                                           lmvsize);
 583                         if (rc < 0)
 584                                 goto out;
 585
 586                         if (rc < sizeof(*md->mea)) {
 587                                 CDEBUG(D_INFO,
 588                                        "size too small: rc < sizeof(*md->mea) (%d < %d)\n",
 589                                         rc, (int)sizeof(*md->mea));
 590                                 rc = -EPROTO;
 591                                 goto out;
 592                         }
 593                 }
 594         }
 595         rc = 0;
 596
 597         if (md->body->valid & OBD_MD_FLRMTPERM) {
 598                 /* remote permission */
 599                 LASSERT(client_is_remote(exp));
 600                 md->remote_perm = req_capsule_server_swab_get(pill, &RMF_ACL,
 601                                                 lustre_swab_mdt_remote_perm);
 602                 if (!md->remote_perm) {
 603                         rc = -EPROTO;
 604                         goto out;
 605                 }
 606         } else if (md->body->valid & OBD_MD_FLACL) {
 607                 /* for ACL, it's possible that FLACL is set but aclsize is zero.
 608                  * only when aclsize != 0 there's an actual segment for ACL
 609                  * in reply buffer.
 610                  */
 611                 if (md->body->aclsize) {
 612                         rc = mdc_unpack_acl(req, md);
 613                         if (rc)
 614                                 goto out;
 615 #ifdef CONFIG_FS_POSIX_ACL
 616                 } else {
 617                         md->posix_acl = NULL;
 618 #endif
 619                 }
 620         }
 621         if (md->body->valid & OBD_MD_FLMDSCAPA) {
 622                 struct obd_capa *oc = NULL;
 623
 624                 rc = mdc_unpack_capa(NULL, req, &RMF_CAPA1, &oc);
 625                 if (rc)
 626                         goto out;
 627                 md->mds_capa = oc;
 628         }
 629
 630         if (md->body->valid & OBD_MD_FLOSSCAPA) {
 631                 struct obd_capa *oc = NULL;
 632
 633                 rc = mdc_unpack_capa(NULL, req, &RMF_CAPA2, &oc);
 634                 if (rc)
 635                         goto out;
 636                 md->oss_capa = oc;
 637         }
 638
 639 out:
 640         if (rc) {
 641                 if (md->oss_capa) {
 642                         capa_put(md->oss_capa);
 643                         md->oss_capa = NULL;
 644                 }
 645                 if (md->mds_capa) {
 646                         capa_put(md->mds_capa);
 647                         md->mds_capa = NULL;
 648                 }
 649 #ifdef CONFIG_FS_POSIX_ACL
 650                 posix_acl_release(md->posix_acl);
 651 #endif
 652                 if (md->lsm)
 653                         obd_free_memmd(dt_exp, &md->lsm);
 654         }
 655         return rc;
 656 }
 657
 658 int mdc_free_lustre_md(struct obd_export *exp, struct lustre_md *md)
 659 {
 660         return 0;
 661 }
 662
 663 /**
 664  * Handles both OPEN and SETATTR RPCs for OPEN-CLOSE and SETATTR-DONE_WRITING
 665  * RPC chains.
 666  */
 667 void mdc_replay_open(struct ptlrpc_request *req)
 668 {
 669         struct md_open_data *mod = req->rq_cb_data;
 670         struct ptlrpc_request *close_req;
 671         struct obd_client_handle *och;
 672         struct lustre_handle old;
 673         struct mdt_body *body;
 674
 675         if (mod == NULL) {
 676                 DEBUG_REQ(D_ERROR, req,
 677                           "Can't properly replay without open data.");
 678                 return;
 679         }
 680
 681         body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
 682         LASSERT(body != NULL);
 683
 684         och = mod->mod_och;
 685         if (och != NULL) {
 686                 struct lustre_handle *file_fh;
 687
 688                 LASSERT(och->och_magic == OBD_CLIENT_HANDLE_MAGIC);
 689
 690                 file_fh = &och->och_fh;
 691                 CDEBUG(D_HA, "updating handle from %#llx to %#llx\n",
 692                        file_fh->cookie, body->handle.cookie);
 693                 old = *file_fh;
 694                 *file_fh = body->handle;
 695         }
 696         close_req = mod->mod_close_req;
 697         if (close_req != NULL) {
 698                 __u32 opc = lustre_msg_get_opc(close_req->rq_reqmsg);
 699                 struct mdt_ioepoch *epoch;
 700
 701                 LASSERT(opc == MDS_CLOSE || opc == MDS_DONE_WRITING);
 702                 epoch = req_capsule_client_get(&close_req->rq_pill,
 703                                                &RMF_MDT_EPOCH);
 704                 LASSERT(epoch);
 705
 706                 if (och != NULL)
 707                         LASSERT(!memcmp(&old, &epoch->handle, sizeof(old)));
 708                 DEBUG_REQ(D_HA, close_req, "updating close body with new fh");
 709                 epoch->handle = body->handle;
 710         }
 711 }
 712
 713 void mdc_commit_open(struct ptlrpc_request *req)
 714 {
 715         struct md_open_data *mod = req->rq_cb_data;
 716
 717         if (mod == NULL)
 718                 return;
 719
 720         /**
 721          * No need to touch md_open_data::mod_och, it holds a reference on
 722          * \var mod and will zero references to each other, \var mod will be
 723          * freed after that when md_open_data::mod_och will put the reference.
 724          */
 725
 726         /**
 727          * Do not let open request to disappear as it still may be needed
 728          * for close rpc to happen (it may happen on evict only, otherwise
 729          * ptlrpc_request::rq_replay does not let mdc_commit_open() to be
 730          * called), just mark this rpc as committed to distinguish these 2
 731          * cases, see mdc_close() for details. The open request reference will
 732          * be put along with freeing \var mod.
 733          */
 734         ptlrpc_request_addref(req);
 735         spin_lock(&req->rq_lock);
 736         req->rq_committed = 1;
 737         spin_unlock(&req->rq_lock);
 738         req->rq_cb_data = NULL;
 739         obd_mod_put(mod);
 740 }
 741
 742 int mdc_set_open_replay_data(struct obd_export *exp,
 743                              struct obd_client_handle *och,
 744                              struct lookup_intent *it)
 745 {
 746         struct md_open_data   *mod;
 747         struct mdt_rec_create *rec;
 748         struct mdt_body       *body;
 749         struct ptlrpc_request *open_req = it->d.lustre.it_data;
 750         struct obd_import     *imp = open_req->rq_import;
 751
 752         if (!open_req->rq_replay)
 753                 return 0;
 754
 755         rec = req_capsule_client_get(&open_req->rq_pill, &RMF_REC_REINT);
 756         body = req_capsule_server_get(&open_req->rq_pill, &RMF_MDT_BODY);
 757         LASSERT(rec != NULL);
 758         /* Incoming message in my byte order (it's been swabbed). */
 759         /* Outgoing messages always in my byte order. */
 760         LASSERT(body != NULL);
 761
 762         /* Only if the import is replayable, we set replay_open data */
 763         if (och && imp->imp_replayable) {
 764                 mod = obd_mod_alloc();
 765                 if (mod == NULL) {
 766                         DEBUG_REQ(D_ERROR, open_req,
 767                                   "Can't allocate md_open_data");
 768                         return 0;
 769                 }
 770
 771                 /**
 772                  * Take a reference on \var mod, to be freed on mdc_close().
 773                  * It protects \var mod from being freed on eviction (commit
 774                  * callback is called despite rq_replay flag).
 775                  * Another reference for \var och.
 776                  */
 777                 obd_mod_get(mod);
 778                 obd_mod_get(mod);
 779
 780                 spin_lock(&open_req->rq_lock);
 781                 och->och_mod = mod;
 782                 mod->mod_och = och;
 783                 mod->mod_is_create = it_disposition(it, DISP_OPEN_CREATE) ||
 784                                      it_disposition(it, DISP_OPEN_STRIPE);
 785                 mod->mod_open_req = open_req;
 786                 open_req->rq_cb_data = mod;
 787                 open_req->rq_commit_cb = mdc_commit_open;
 788                 spin_unlock(&open_req->rq_lock);
 789         }
 790
 791         rec->cr_fid2 = body->fid1;
 792         rec->cr_ioepoch = body->ioepoch;
 793         rec->cr_old_handle.cookie = body->handle.cookie;
 794         open_req->rq_replay_cb = mdc_replay_open;
 795         if (!fid_is_sane(&body->fid1)) {
 796                 DEBUG_REQ(D_ERROR, open_req,
 797                           "Saving replay request with insane fid");
 798                 LBUG();
 799         }
 800
 801         DEBUG_REQ(D_RPCTRACE, open_req, "Set up open replay data");
 802         return 0;
 803 }
 804
 805 static void mdc_free_open(struct md_open_data *mod)
 806 {
 807         int committed = 0;
 808
 809         if (mod->mod_is_create == 0 &&
 810             imp_connect_disp_stripe(mod->mod_open_req->rq_import))
 811                 committed = 1;
 812
 813         LASSERT(mod->mod_open_req->rq_replay == 0);
 814
 815         DEBUG_REQ(D_RPCTRACE, mod->mod_open_req, "free open request\n");
 816
 817         ptlrpc_request_committed(mod->mod_open_req, committed);
 818         if (mod->mod_close_req)
 819                 ptlrpc_request_committed(mod->mod_close_req, committed);
 820 }
 821
 822 int mdc_clear_open_replay_data(struct obd_export *exp,
 823                                struct obd_client_handle *och)
 824 {
 825         struct md_open_data *mod = och->och_mod;
 826
 827         /**
 828          * It is possible to not have \var mod in a case of eviction between
 829          * lookup and ll_file_open().
 830          **/
 831         if (mod == NULL)
 832                 return 0;
 833
 834         LASSERT(mod != LP_POISON);
 835         LASSERT(mod->mod_open_req != NULL);
 836         mdc_free_open(mod);
 837
 838         mod->mod_och = NULL;
 839         och->och_mod = NULL;
 840         obd_mod_put(mod);
 841
 842         return 0;
 843 }
 844
 845 /* Prepares the request for the replay by the given reply */
 846 static void mdc_close_handle_reply(struct ptlrpc_request *req,
 847                                    struct md_op_data *op_data, int rc) {
 848         struct mdt_body  *repbody;
 849         struct mdt_ioepoch *epoch;
 850
 851         if (req && rc == -EAGAIN) {
 852                 repbody = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
 853                 epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
 854
 855                 epoch->flags |= MF_SOM_AU;
 856                 if (repbody->valid & OBD_MD_FLGETATTRLOCK)
 857                         op_data->op_flags |= MF_GETATTR_LOCK;
 858         }
 859 }
 860
 861 static int mdc_close(struct obd_export *exp, struct md_op_data *op_data,
 862                      struct md_open_data *mod, struct ptlrpc_request **request)
 863 {
 864         struct obd_device     *obd = class_exp2obd(exp);
 865         struct ptlrpc_request *req;
 866         struct req_format     *req_fmt;
 867         int                    rc;
 868         int                    saved_rc = 0;
 869
 870
 871         req_fmt = &RQF_MDS_CLOSE;
 872         if (op_data->op_bias & MDS_HSM_RELEASE) {
 873                 req_fmt = &RQF_MDS_RELEASE_CLOSE;
 874
 875                 /* allocate a FID for volatile file */
 876                 rc = mdc_fid_alloc(exp, &op_data->op_fid2, op_data);
 877                 if (rc < 0) {
 878                         CERROR("%s: "DFID" failed to allocate FID: %d\n",
 879                                obd->obd_name, PFID(&op_data->op_fid1), rc);
 880                         /* save the errcode and proceed to close */
 881                         saved_rc = rc;
 882                 }
 883         }
 884
 885         *request = NULL;
 886         req = ptlrpc_request_alloc(class_exp2cliimp(exp), req_fmt);
 887         if (req == NULL)
 888                 return -ENOMEM;
 889
 890         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
 891
 892         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_CLOSE);
 893         if (rc) {
 894                 ptlrpc_request_free(req);
 895                 return rc;
 896         }
 897
 898         /* To avoid a livelock (bug 7034), we need to send CLOSE RPCs to a
 899          * portal whose threads are not taking any DLM locks and are therefore
 900          * always progressing */
 901         req->rq_request_portal = MDS_READPAGE_PORTAL;
 902         ptlrpc_at_set_req_timeout(req);
 903
 904         /* Ensure that this close's handle is fixed up during replay. */
 905         if (likely(mod != NULL)) {
 906                 LASSERTF(mod->mod_open_req != NULL &&
 907                          mod->mod_open_req->rq_type != LI_POISON,
 908                          "POISONED open %p!\n", mod->mod_open_req);
 909
 910                 mod->mod_close_req = req;
 911
 912                 DEBUG_REQ(D_HA, mod->mod_open_req, "matched open");
 913                 /* We no longer want to preserve this open for replay even
 914                  * though the open was committed. b=3632, b=3633 */
 915                 spin_lock(&mod->mod_open_req->rq_lock);
 916                 mod->mod_open_req->rq_replay = 0;
 917                 spin_unlock(&mod->mod_open_req->rq_lock);
 918         } else {
 919                  CDEBUG(D_HA,
 920                         "couldn't find open req; expecting close error\n");
 921         }
 922
 923         mdc_close_pack(req, op_data);
 924
 925         req_capsule_set_size(&req->rq_pill, &RMF_MDT_MD, RCL_SERVER,
 926                              obd->u.cli.cl_default_mds_easize);
 927         req_capsule_set_size(&req->rq_pill, &RMF_LOGCOOKIES, RCL_SERVER,
 928                              obd->u.cli.cl_default_mds_cookiesize);
 929
 930         ptlrpc_request_set_replen(req);
 931
 932         mdc_get_rpc_lock(obd->u.cli.cl_close_lock, NULL);
 933         rc = ptlrpc_queue_wait(req);
 934         mdc_put_rpc_lock(obd->u.cli.cl_close_lock, NULL);
 935
 936         if (req->rq_repmsg == NULL) {
 937                 CDEBUG(D_RPCTRACE, "request failed to send: %p, %d\n", req,
 938                        req->rq_status);
 939                 if (rc == 0)
 940                         rc = req->rq_status ?: -EIO;
 941         } else if (rc == 0 || rc == -EAGAIN) {
 942                 struct mdt_body *body;
 943
 944                 rc = lustre_msg_get_status(req->rq_repmsg);
 945                 if (lustre_msg_get_type(req->rq_repmsg) == PTL_RPC_MSG_ERR) {
 946                         DEBUG_REQ(D_ERROR, req,
 947                                   "type == PTL_RPC_MSG_ERR, err = %d", rc);
 948                         if (rc > 0)
 949                                 rc = -rc;
 950                 }
 951                 body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
 952                 if (body == NULL)
 953                         rc = -EPROTO;
 954         } else if (rc == -ESTALE) {
 955                 /**
 956                  * it can be allowed error after 3633 if open was committed and
 957                  * server failed before close was sent. Let's check if mod
 958                  * exists and return no error in that case
 959                  */
 960                 if (mod) {
 961                         DEBUG_REQ(D_HA, req, "Reset ESTALE = %d", rc);
 962                         LASSERT(mod->mod_open_req != NULL);
 963                         if (mod->mod_open_req->rq_committed)
 964                                 rc = 0;
 965                 }
 966         }
 967
 968         if (mod) {
 969                 if (rc != 0)
 970                         mod->mod_close_req = NULL;
 971                 /* Since now, mod is accessed through open_req only,
 972                  * thus close req does not keep a reference on mod anymore. */
 973                 obd_mod_put(mod);
 974         }
 975         *request = req;
 976         mdc_close_handle_reply(req, op_data, rc);
 977         return rc < 0 ? rc : saved_rc;
 978 }
 979
 980 static int mdc_done_writing(struct obd_export *exp, struct md_op_data *op_data,
 981                             struct md_open_data *mod)
 982 {
 983         struct obd_device     *obd = class_exp2obd(exp);
 984         struct ptlrpc_request *req;
 985         int                 rc;
 986
 987         req = ptlrpc_request_alloc(class_exp2cliimp(exp),
 988                                    &RQF_MDS_DONE_WRITING);
 989         if (req == NULL)
 990                 return -ENOMEM;
 991
 992         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
 993         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_DONE_WRITING);
 994         if (rc) {
 995                 ptlrpc_request_free(req);
 996                 return rc;
 997         }
 998
 999         if (mod != NULL) {
1000                 LASSERTF(mod->mod_open_req != NULL &&
1001                          mod->mod_open_req->rq_type != LI_POISON,
1002                          "POISONED setattr %p!\n", mod->mod_open_req);
1003
1004                 mod->mod_close_req = req;
1005                 DEBUG_REQ(D_HA, mod->mod_open_req, "matched setattr");
1006                 /* We no longer want to preserve this setattr for replay even
1007                  * though the open was committed. b=3632, b=3633 */
1008                 spin_lock(&mod->mod_open_req->rq_lock);
1009                 mod->mod_open_req->rq_replay = 0;
1010                 spin_unlock(&mod->mod_open_req->rq_lock);
1011         }
1012
1013         mdc_close_pack(req, op_data);
1014         ptlrpc_request_set_replen(req);
1015
1016         mdc_get_rpc_lock(obd->u.cli.cl_close_lock, NULL);
1017         rc = ptlrpc_queue_wait(req);
1018         mdc_put_rpc_lock(obd->u.cli.cl_close_lock, NULL);
1019
1020         if (rc == -ESTALE) {
1021                 /**
1022                  * it can be allowed error after 3633 if open or setattr were
1023                  * committed and server failed before close was sent.
1024                  * Let's check if mod exists and return no error in that case
1025                  */
1026                 if (mod) {
1027                         LASSERT(mod->mod_open_req != NULL);
1028                         if (mod->mod_open_req->rq_committed)
1029                                 rc = 0;
1030                 }
1031         }
1032
1033         if (mod) {
1034                 if (rc != 0)
1035                         mod->mod_close_req = NULL;
1036                 LASSERT(mod->mod_open_req != NULL);
1037                 mdc_free_open(mod);
1038
1039                 /* Since now, mod is accessed through setattr req only,
1040                  * thus DW req does not keep a reference on mod anymore. */
1041                 obd_mod_put(mod);
1042         }
1043
1044         mdc_close_handle_reply(req, op_data, rc);
1045         ptlrpc_req_finished(req);
1046         return rc;
1047 }
1048
1049
1050 static int mdc_readpage(struct obd_export *exp, struct md_op_data *op_data,
1051                         struct page **pages, struct ptlrpc_request **request)
1052 {
1053         struct ptlrpc_request   *req;
1054         struct ptlrpc_bulk_desc *desc;
1055         int                   i;
1056         wait_queue_head_t             waitq;
1057         int                   resends = 0;
1058         struct l_wait_info       lwi;
1059         int                   rc;
1060
1061         *request = NULL;
1062         init_waitqueue_head(&waitq);
1063
1064 restart_bulk:
1065         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_MDS_READPAGE);
1066         if (req == NULL)
1067                 return -ENOMEM;
1068
1069         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
1070
1071         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_READPAGE);
1072         if (rc) {
1073                 ptlrpc_request_free(req);
1074                 return rc;
1075         }
1076
1077         req->rq_request_portal = MDS_READPAGE_PORTAL;
1078         ptlrpc_at_set_req_timeout(req);
1079
1080         desc = ptlrpc_prep_bulk_imp(req, op_data->op_npages, 1, BULK_PUT_SINK,
1081                                     MDS_BULK_PORTAL);
1082         if (desc == NULL) {
1083                 ptlrpc_request_free(req);
1084                 return -ENOMEM;
1085         }
1086
1087         /* NB req now owns desc and will free it when it gets freed */
1088         for (i = 0; i < op_data->op_npages; i++)
1089                 ptlrpc_prep_bulk_page_pin(desc, pages[i], 0, PAGE_CACHE_SIZE);
1090
1091         mdc_readdir_pack(req, op_data->op_offset,
1092                          PAGE_CACHE_SIZE * op_data->op_npages,
1093                          &op_data->op_fid1, op_data->op_capa1);
1094
1095         ptlrpc_request_set_replen(req);
1096         rc = ptlrpc_queue_wait(req);
1097         if (rc) {
1098                 ptlrpc_req_finished(req);
1099                 if (rc != -ETIMEDOUT)
1100                         return rc;
1101
1102                 resends++;
1103                 if (!client_should_resend(resends, &exp->exp_obd->u.cli)) {
1104                         CERROR("too many resend retries, returning error\n");
1105                         return -EIO;
1106                 }
1107                 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends),
1108                                        NULL, NULL, NULL);
1109                 l_wait_event(waitq, 0, &lwi);
1110
1111                 goto restart_bulk;
1112         }
1113
1114         rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk,
1115                                           req->rq_bulk->bd_nob_transferred);
1116         if (rc < 0) {
1117                 ptlrpc_req_finished(req);
1118                 return rc;
1119         }
1120
1121         if (req->rq_bulk->bd_nob_transferred & ~LU_PAGE_MASK) {
1122                 CERROR("Unexpected # bytes transferred: %d (%ld expected)\n",
1123                         req->rq_bulk->bd_nob_transferred,
1124                         PAGE_CACHE_SIZE * op_data->op_npages);
1125                 ptlrpc_req_finished(req);
1126                 return -EPROTO;
1127         }
1128
1129         *request = req;
1130         return 0;
1131 }
1132
1133 static int mdc_statfs(const struct lu_env *env,
1134                       struct obd_export *exp, struct obd_statfs *osfs,
1135                       __u64 max_age, __u32 flags)
1136 {
1137         struct obd_device     *obd = class_exp2obd(exp);
1138         struct ptlrpc_request *req;
1139         struct obd_statfs     *msfs;
1140         struct obd_import     *imp = NULL;
1141         int                 rc;
1142
1143         /*
1144          * Since the request might also come from lprocfs, so we need
1145          * sync this with client_disconnect_export Bug15684
1146          */
1147         down_read(&obd->u.cli.cl_sem);
1148         if (obd->u.cli.cl_import)
1149                 imp = class_import_get(obd->u.cli.cl_import);
1150         up_read(&obd->u.cli.cl_sem);
1151         if (!imp)
1152                 return -ENODEV;
1153
1154         req = ptlrpc_request_alloc_pack(imp, &RQF_MDS_STATFS,
1155                                         LUSTRE_MDS_VERSION, MDS_STATFS);
1156         if (req == NULL) {
1157                 rc = -ENOMEM;
1158                 goto output;
1159         }
1160
1161         ptlrpc_request_set_replen(req);
1162
1163         if (flags & OBD_STATFS_NODELAY) {
1164                 /* procfs requests not want stay in wait for avoid deadlock */
1165                 req->rq_no_resend = 1;
1166                 req->rq_no_delay = 1;
1167         }
1168
1169         rc = ptlrpc_queue_wait(req);
1170         if (rc) {
1171                 /* check connection error first */
1172                 if (imp->imp_connect_error)
1173                         rc = imp->imp_connect_error;
1174                 goto out;
1175         }
1176
1177         msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
1178         if (msfs == NULL) {
1179                 rc = -EPROTO;
1180                 goto out;
1181         }
1182
1183         *osfs = *msfs;
1184 out:
1185         ptlrpc_req_finished(req);
1186 output:
1187         class_import_put(imp);
1188         return rc;
1189 }
1190
1191 static int mdc_ioc_fid2path(struct obd_export *exp, struct getinfo_fid2path *gf)
1192 {
1193         __u32 keylen, vallen;
1194         void *key;
1195         int rc;
1196
1197         if (gf->gf_pathlen > PATH_MAX)
1198                 return -ENAMETOOLONG;
1199         if (gf->gf_pathlen < 2)
1200                 return -EOVERFLOW;
1201
1202         /* Key is KEY_FID2PATH + getinfo_fid2path description */
1203         keylen = cfs_size_round(sizeof(KEY_FID2PATH)) + sizeof(*gf);
1204         OBD_ALLOC(key, keylen);
1205         if (key == NULL)
1206                 return -ENOMEM;
1207         memcpy(key, KEY_FID2PATH, sizeof(KEY_FID2PATH));
1208         memcpy(key + cfs_size_round(sizeof(KEY_FID2PATH)), gf, sizeof(*gf));
1209
1210         CDEBUG(D_IOCTL, "path get "DFID" from %llu #%d\n",
1211                PFID(&gf->gf_fid), gf->gf_recno, gf->gf_linkno);
1212
1213         if (!fid_is_sane(&gf->gf_fid)) {
1214                 rc = -EINVAL;
1215                 goto out;
1216         }
1217
1218         /* Val is struct getinfo_fid2path result plus path */
1219         vallen = sizeof(*gf) + gf->gf_pathlen;
1220
1221         rc = obd_get_info(NULL, exp, keylen, key, &vallen, gf, NULL);
1222         if (rc != 0 && rc != -EREMOTE)
1223                 goto out;
1224
1225         if (vallen <= sizeof(*gf)) {
1226                 rc = -EPROTO;
1227                 goto out;
1228         } else if (vallen > sizeof(*gf) + gf->gf_pathlen) {
1229                 rc = -EOVERFLOW;
1230                 goto out;
1231         }
1232
1233         CDEBUG(D_IOCTL, "path get "DFID" from %llu #%d\n%s\n",
1234                PFID(&gf->gf_fid), gf->gf_recno, gf->gf_linkno, gf->gf_path);
1235
1236 out:
1237         OBD_FREE(key, keylen);
1238         return rc;
1239 }
1240
1241 static int mdc_ioc_hsm_progress(struct obd_export *exp,
1242                                 struct hsm_progress_kernel *hpk)
1243 {
1244         struct obd_import               *imp = class_exp2cliimp(exp);
1245         struct hsm_progress_kernel      *req_hpk;
1246         struct ptlrpc_request           *req;
1247         int                              rc;
1248
1249         req = ptlrpc_request_alloc_pack(imp, &RQF_MDS_HSM_PROGRESS,
1250                                         LUSTRE_MDS_VERSION, MDS_HSM_PROGRESS);
1251         if (req == NULL) {
1252                 rc = -ENOMEM;
1253                 goto out;
1254         }
1255
1256         mdc_pack_body(req, NULL, NULL, OBD_MD_FLRMTPERM, 0, 0, 0);
1257
1258         /* Copy hsm_progress struct */
1259         req_hpk = req_capsule_client_get(&req->rq_pill, &RMF_MDS_HSM_PROGRESS);
1260         if (req_hpk == NULL) {
1261                 rc = -EPROTO;
1262                 goto out;
1263         }
1264
1265         *req_hpk = *hpk;
1266         req_hpk->hpk_errval = lustre_errno_hton(hpk->hpk_errval);
1267
1268         ptlrpc_request_set_replen(req);
1269
1270         rc = mdc_queue_wait(req);
1271         goto out;
1272 out:
1273         ptlrpc_req_finished(req);
1274         return rc;
1275 }
1276
1277 static int mdc_ioc_hsm_ct_register(struct obd_import *imp, __u32 archives)
1278 {
1279         __u32                   *archive_mask;
1280         struct ptlrpc_request   *req;
1281         int                      rc;
1282
1283         req = ptlrpc_request_alloc_pack(imp, &RQF_MDS_HSM_CT_REGISTER,
1284                                         LUSTRE_MDS_VERSION,
1285                                         MDS_HSM_CT_REGISTER);
1286         if (req == NULL) {
1287                 rc = -ENOMEM;
1288                 goto out;
1289         }
1290
1291         mdc_pack_body(req, NULL, NULL, OBD_MD_FLRMTPERM, 0, 0, 0);
1292
1293         /* Copy hsm_progress struct */
1294         archive_mask = req_capsule_client_get(&req->rq_pill,
1295                                               &RMF_MDS_HSM_ARCHIVE);
1296         if (archive_mask == NULL) {
1297                 rc = -EPROTO;
1298                 goto out;
1299         }
1300
1301         *archive_mask = archives;
1302
1303         ptlrpc_request_set_replen(req);
1304
1305         rc = mdc_queue_wait(req);
1306         goto out;
1307 out:
1308         ptlrpc_req_finished(req);
1309         return rc;
1310 }
1311
1312 static int mdc_ioc_hsm_current_action(struct obd_export *exp,
1313                                       struct md_op_data *op_data)
1314 {
1315         struct hsm_current_action       *hca = op_data->op_data;
1316         struct hsm_current_action       *req_hca;
1317         struct ptlrpc_request           *req;
1318         int                              rc;
1319
1320         req = ptlrpc_request_alloc(class_exp2cliimp(exp),
1321                                    &RQF_MDS_HSM_ACTION);
1322         if (req == NULL)
1323                 return -ENOMEM;
1324
1325         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
1326
1327         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_HSM_ACTION);
1328         if (rc) {
1329                 ptlrpc_request_free(req);
1330                 return rc;
1331         }
1332
1333         mdc_pack_body(req, &op_data->op_fid1, op_data->op_capa1,
1334                       OBD_MD_FLRMTPERM, 0, op_data->op_suppgids[0], 0);
1335
1336         ptlrpc_request_set_replen(req);
1337
1338         rc = mdc_queue_wait(req);
1339         if (rc)
1340                 goto out;
1341
1342         req_hca = req_capsule_server_get(&req->rq_pill,
1343                                          &RMF_MDS_HSM_CURRENT_ACTION);
1344         if (req_hca == NULL) {
1345                 rc = -EPROTO;
1346                 goto out;
1347         }
1348
1349         *hca = *req_hca;
1350
1351 out:
1352         ptlrpc_req_finished(req);
1353         return rc;
1354 }
1355
1356 static int mdc_ioc_hsm_ct_unregister(struct obd_import *imp)
1357 {
1358         struct ptlrpc_request   *req;
1359         int                      rc;
1360
1361         req = ptlrpc_request_alloc_pack(imp, &RQF_MDS_HSM_CT_UNREGISTER,
1362                                         LUSTRE_MDS_VERSION,
1363                                         MDS_HSM_CT_UNREGISTER);
1364         if (req == NULL) {
1365                 rc = -ENOMEM;
1366                 goto out;
1367         }
1368
1369         mdc_pack_body(req, NULL, NULL, OBD_MD_FLRMTPERM, 0, 0, 0);
1370
1371         ptlrpc_request_set_replen(req);
1372
1373         rc = mdc_queue_wait(req);
1374         goto out;
1375 out:
1376         ptlrpc_req_finished(req);
1377         return rc;
1378 }
1379
1380 static int mdc_ioc_hsm_state_get(struct obd_export *exp,
1381                                  struct md_op_data *op_data)
1382 {
1383         struct hsm_user_state   *hus = op_data->op_data;
1384         struct hsm_user_state   *req_hus;
1385         struct ptlrpc_request   *req;
1386         int                      rc;
1387
1388         req = ptlrpc_request_alloc(class_exp2cliimp(exp),
1389                                    &RQF_MDS_HSM_STATE_GET);
1390         if (req == NULL)
1391                 return -ENOMEM;
1392
1393         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
1394
1395         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_HSM_STATE_GET);
1396         if (rc != 0) {
1397                 ptlrpc_request_free(req);
1398                 return rc;
1399         }
1400
1401         mdc_pack_body(req, &op_data->op_fid1, op_data->op_capa1,
1402                       OBD_MD_FLRMTPERM, 0, op_data->op_suppgids[0], 0);
1403
1404         ptlrpc_request_set_replen(req);
1405
1406         rc = mdc_queue_wait(req);
1407         if (rc)
1408                 goto out;
1409
1410         req_hus = req_capsule_server_get(&req->rq_pill, &RMF_HSM_USER_STATE);
1411         if (req_hus == NULL) {
1412                 rc = -EPROTO;
1413                 goto out;
1414         }
1415
1416         *hus = *req_hus;
1417
1418 out:
1419         ptlrpc_req_finished(req);
1420         return rc;
1421 }
1422
1423 static int mdc_ioc_hsm_state_set(struct obd_export *exp,
1424                                  struct md_op_data *op_data)
1425 {
1426         struct hsm_state_set    *hss = op_data->op_data;
1427         struct hsm_state_set    *req_hss;
1428         struct ptlrpc_request   *req;
1429         int                      rc;
1430
1431         req = ptlrpc_request_alloc(class_exp2cliimp(exp),
1432                                    &RQF_MDS_HSM_STATE_SET);
1433         if (req == NULL)
1434                 return -ENOMEM;
1435
1436         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
1437
1438         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_HSM_STATE_SET);
1439         if (rc) {
1440                 ptlrpc_request_free(req);
1441                 return rc;
1442         }
1443
1444         mdc_pack_body(req, &op_data->op_fid1, op_data->op_capa1,
1445                       OBD_MD_FLRMTPERM, 0, op_data->op_suppgids[0], 0);
1446
1447         /* Copy states */
1448         req_hss = req_capsule_client_get(&req->rq_pill, &RMF_HSM_STATE_SET);
1449         if (req_hss == NULL) {
1450                 rc = -EPROTO;
1451                 goto out;
1452         }
1453         *req_hss = *hss;
1454
1455         ptlrpc_request_set_replen(req);
1456
1457         rc = mdc_queue_wait(req);
1458         goto out;
1459
1460 out:
1461         ptlrpc_req_finished(req);
1462         return rc;
1463 }
1464
1465 static int mdc_ioc_hsm_request(struct obd_export *exp,
1466                                struct hsm_user_request *hur)
1467 {
1468         struct obd_import       *imp = class_exp2cliimp(exp);
1469         struct ptlrpc_request   *req;
1470         struct hsm_request      *req_hr;
1471         struct hsm_user_item    *req_hui;
1472         char                    *req_opaque;
1473         int                      rc;
1474
1475         req = ptlrpc_request_alloc(imp, &RQF_MDS_HSM_REQUEST);
1476         if (req == NULL) {
1477                 rc = -ENOMEM;
1478                 goto out;
1479         }
1480
1481         req_capsule_set_size(&req->rq_pill, &RMF_MDS_HSM_USER_ITEM, RCL_CLIENT,
1482                              hur->hur_request.hr_itemcount
1483                              * sizeof(struct hsm_user_item));
1484         req_capsule_set_size(&req->rq_pill, &RMF_GENERIC_DATA, RCL_CLIENT,
1485                              hur->hur_request.hr_data_len);
1486
1487         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_HSM_REQUEST);
1488         if (rc) {
1489                 ptlrpc_request_free(req);
1490                 return rc;
1491         }
1492
1493         mdc_pack_body(req, NULL, NULL, OBD_MD_FLRMTPERM, 0, 0, 0);
1494
1495         /* Copy hsm_request struct */
1496         req_hr = req_capsule_client_get(&req->rq_pill, &RMF_MDS_HSM_REQUEST);
1497         if (req_hr == NULL) {
1498                 rc = -EPROTO;
1499                 goto out;
1500         }
1501         *req_hr = hur->hur_request;
1502
1503         /* Copy hsm_user_item structs */
1504         req_hui = req_capsule_client_get(&req->rq_pill, &RMF_MDS_HSM_USER_ITEM);
1505         if (req_hui == NULL) {
1506                 rc = -EPROTO;
1507                 goto out;
1508         }
1509         memcpy(req_hui, hur->hur_user_item,
1510                hur->hur_request.hr_itemcount * sizeof(struct hsm_user_item));
1511
1512         /* Copy opaque field */
1513         req_opaque = req_capsule_client_get(&req->rq_pill, &RMF_GENERIC_DATA);
1514         if (req_opaque == NULL) {
1515                 rc = -EPROTO;
1516                 goto out;
1517         }
1518         memcpy(req_opaque, hur_data(hur), hur->hur_request.hr_data_len);
1519
1520         ptlrpc_request_set_replen(req);
1521
1522         rc = mdc_queue_wait(req);
1523         goto out;
1524
1525 out:
1526         ptlrpc_req_finished(req);
1527         return rc;
1528 }
1529
1530 static struct kuc_hdr *changelog_kuc_hdr(char *buf, int len, int flags)
1531 {
1532         struct kuc_hdr *lh = (struct kuc_hdr *)buf;
1533
1534         LASSERT(len <= KUC_CHANGELOG_MSG_MAXSIZE);
1535
1536         lh->kuc_magic = KUC_MAGIC;
1537         lh->kuc_transport = KUC_TRANSPORT_CHANGELOG;
1538         lh->kuc_flags = flags;
1539         lh->kuc_msgtype = CL_RECORD;
1540         lh->kuc_msglen = len;
1541         return lh;
1542 }
1543
1544 #define D_CHANGELOG 0
1545
1546 struct changelog_show {
1547         __u64           cs_startrec;
1548         __u32           cs_flags;
1549         struct file     *cs_fp;
1550         char            *cs_buf;
1551         struct obd_device *cs_obd;
1552 };
1553
1554 static int changelog_kkuc_cb(const struct lu_env *env, struct llog_handle *llh,
1555                              struct llog_rec_hdr *hdr, void *data)
1556 {
1557         struct changelog_show *cs = data;
1558         struct llog_changelog_rec *rec = (struct llog_changelog_rec *)hdr;
1559         struct kuc_hdr *lh;
1560         int len, rc;
1561
1562         if (rec->cr_hdr.lrh_type != CHANGELOG_REC) {
1563                 rc = -EINVAL;
1564                 CERROR("%s: not a changelog rec %x/%d: rc = %d\n",
1565                        cs->cs_obd->obd_name, rec->cr_hdr.lrh_type,
1566                        rec->cr.cr_type, rc);
1567                 return rc;
1568         }
1569
1570         if (rec->cr.cr_index < cs->cs_startrec) {
1571                 /* Skip entries earlier than what we are interested in */
1572                 CDEBUG(D_CHANGELOG, "rec=%llu start=%llu\n",
1573                        rec->cr.cr_index, cs->cs_startrec);
1574                 return 0;
1575         }
1576
1577         CDEBUG(D_CHANGELOG, "%llu %02d%-5s %llu 0x%x t="DFID" p="DFID
1578                 " %.*s\n", rec->cr.cr_index, rec->cr.cr_type,
1579                 changelog_type2str(rec->cr.cr_type), rec->cr.cr_time,
1580                 rec->cr.cr_flags & CLF_FLAGMASK,
1581                 PFID(&rec->cr.cr_tfid), PFID(&rec->cr.cr_pfid),
1582                 rec->cr.cr_namelen, changelog_rec_name(&rec->cr));
1583
1584         len = sizeof(*lh) + changelog_rec_size(&rec->cr) + rec->cr.cr_namelen;
1585
1586         /* Set up the message */
1587         lh = changelog_kuc_hdr(cs->cs_buf, len, cs->cs_flags);
1588         memcpy(lh + 1, &rec->cr, len - sizeof(*lh));
1589
1590         rc = libcfs_kkuc_msg_put(cs->cs_fp, lh);
1591         CDEBUG(D_CHANGELOG, "kucmsg fp %p len %d rc %d\n", cs->cs_fp, len, rc);
1592
1593         return rc;
1594 }
1595
1596 static int mdc_changelog_send_thread(void *csdata)
1597 {
1598         struct changelog_show *cs = csdata;
1599         struct llog_ctxt *ctxt = NULL;
1600         struct llog_handle *llh = NULL;
1601         struct kuc_hdr *kuch;
1602         int rc;
1603
1604         CDEBUG(D_CHANGELOG, "changelog to fp=%p start %llu\n",
1605                cs->cs_fp, cs->cs_startrec);
1606
1607         OBD_ALLOC(cs->cs_buf, KUC_CHANGELOG_MSG_MAXSIZE);
1608         if (cs->cs_buf == NULL) {
1609                 rc = -ENOMEM;
1610                 goto out;
1611         }
1612
1613         /* Set up the remote catalog handle */
1614         ctxt = llog_get_context(cs->cs_obd, LLOG_CHANGELOG_REPL_CTXT);
1615         if (ctxt == NULL) {
1616                 rc = -ENOENT;
1617                 goto out;
1618         }
1619         rc = llog_open(NULL, ctxt, &llh, NULL, CHANGELOG_CATALOG,
1620                        LLOG_OPEN_EXISTS);
1621         if (rc) {
1622                 CERROR("%s: fail to open changelog catalog: rc = %d\n",
1623                        cs->cs_obd->obd_name, rc);
1624                 goto out;
1625         }
1626         rc = llog_init_handle(NULL, llh, LLOG_F_IS_CAT, NULL);
1627         if (rc) {
1628                 CERROR("llog_init_handle failed %d\n", rc);
1629                 goto out;
1630         }
1631
1632         rc = llog_cat_process(NULL, llh, changelog_kkuc_cb, cs, 0, 0);
1633
1634         /* Send EOF no matter what our result */
1635         kuch = changelog_kuc_hdr(cs->cs_buf, sizeof(*kuch), cs->cs_flags);
1636         if (kuch) {
1637                 kuch->kuc_msgtype = CL_EOF;
1638                 libcfs_kkuc_msg_put(cs->cs_fp, kuch);
1639         }
1640
1641 out:
1642         fput(cs->cs_fp);
1643         if (llh)
1644                 llog_cat_close(NULL, llh);
1645         if (ctxt)
1646                 llog_ctxt_put(ctxt);
1647         if (cs->cs_buf)
1648                 OBD_FREE(cs->cs_buf, KUC_CHANGELOG_MSG_MAXSIZE);
1649         OBD_FREE_PTR(cs);
1650         return rc;
1651 }
1652
1653 static int mdc_ioc_changelog_send(struct obd_device *obd,
1654                                   struct ioc_changelog *icc)
1655 {
1656         struct changelog_show *cs;
1657         int rc;
1658
1659         /* Freed in mdc_changelog_send_thread */
1660         OBD_ALLOC_PTR(cs);
1661         if (!cs)
1662                 return -ENOMEM;
1663
1664         cs->cs_obd = obd;
1665         cs->cs_startrec = icc->icc_recno;
1666         /* matching fput in mdc_changelog_send_thread */
1667         cs->cs_fp = fget(icc->icc_id);
1668         cs->cs_flags = icc->icc_flags;
1669
1670         /*
1671          * New thread because we should return to user app before
1672          * writing into our pipe
1673          */
1674         rc = PTR_ERR(kthread_run(mdc_changelog_send_thread, cs,
1675                                  "mdc_clg_send_thread"));
1676         if (!IS_ERR_VALUE(rc)) {
1677                 CDEBUG(D_CHANGELOG, "start changelog thread\n");
1678                 return 0;
1679         }
1680
1681         CERROR("Failed to start changelog thread: %d\n", rc);
1682         OBD_FREE_PTR(cs);
1683         return rc;
1684 }
1685
1686 static int mdc_ioc_hsm_ct_start(struct obd_export *exp,
1687                                 struct lustre_kernelcomm *lk);
1688
1689 static int mdc_quotacheck(struct obd_device *unused, struct obd_export *exp,
1690                           struct obd_quotactl *oqctl)
1691 {
1692         struct client_obd       *cli = &exp->exp_obd->u.cli;
1693         struct ptlrpc_request   *req;
1694         struct obd_quotactl     *body;
1695         int                   rc;
1696
1697         req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
1698                                         &RQF_MDS_QUOTACHECK, LUSTRE_MDS_VERSION,
1699                                         MDS_QUOTACHECK);
1700         if (req == NULL)
1701                 return -ENOMEM;
1702
1703         body = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
1704         *body = *oqctl;
1705
1706         ptlrpc_request_set_replen(req);
1707
1708         /* the next poll will find -ENODATA, that means quotacheck is
1709          * going on */
1710         cli->cl_qchk_stat = -ENODATA;
1711         rc = ptlrpc_queue_wait(req);
1712         if (rc)
1713                 cli->cl_qchk_stat = rc;
1714         ptlrpc_req_finished(req);
1715         return rc;
1716 }
1717
1718 static int mdc_quota_poll_check(struct obd_export *exp,
1719                                 struct if_quotacheck *qchk)
1720 {
1721         struct client_obd *cli = &exp->exp_obd->u.cli;
1722         int rc;
1723
1724         qchk->obd_uuid = cli->cl_target_uuid;
1725         memcpy(qchk->obd_type, LUSTRE_MDS_NAME, strlen(LUSTRE_MDS_NAME));
1726
1727         rc = cli->cl_qchk_stat;
1728         /* the client is not the previous one */
1729         if (rc == CL_NOT_QUOTACHECKED)
1730                 rc = -EINTR;
1731         return rc;
1732 }
1733
1734 static int mdc_quotactl(struct obd_device *unused, struct obd_export *exp,
1735                         struct obd_quotactl *oqctl)
1736 {
1737         struct ptlrpc_request   *req;
1738         struct obd_quotactl     *oqc;
1739         int                   rc;
1740
1741         req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
1742                                         &RQF_MDS_QUOTACTL, LUSTRE_MDS_VERSION,
1743                                         MDS_QUOTACTL);
1744         if (req == NULL)
1745                 return -ENOMEM;
1746
1747         oqc = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
1748         *oqc = *oqctl;
1749
1750         ptlrpc_request_set_replen(req);
1751         ptlrpc_at_set_req_timeout(req);
1752         req->rq_no_resend = 1;
1753
1754         rc = ptlrpc_queue_wait(req);
1755         if (rc)
1756                 CERROR("ptlrpc_queue_wait failed, rc: %d\n", rc);
1757
1758         if (req->rq_repmsg) {
1759                 oqc = req_capsule_server_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
1760                 if (oqc) {
1761                         *oqctl = *oqc;
1762                 } else if (!rc) {
1763                         CERROR("Can't unpack obd_quotactl\n");
1764                         rc = -EPROTO;
1765                 }
1766         } else if (!rc) {
1767                 CERROR("Can't unpack obd_quotactl\n");
1768                 rc = -EPROTO;
1769         }
1770         ptlrpc_req_finished(req);
1771
1772         return rc;
1773 }
1774
1775 static int mdc_ioc_swap_layouts(struct obd_export *exp,
1776                                 struct md_op_data *op_data)
1777 {
1778         LIST_HEAD(cancels);
1779         struct ptlrpc_request   *req;
1780         int                      rc, count;
1781         struct mdc_swap_layouts *msl, *payload;
1782
1783         msl = op_data->op_data;
1784
1785         /* When the MDT will get the MDS_SWAP_LAYOUTS RPC the
1786          * first thing it will do is to cancel the 2 layout
1787          * locks hold by this client.
1788          * So the client must cancel its layout locks on the 2 fids
1789          * with the request RPC to avoid extra RPC round trips
1790          */
1791         count = mdc_resource_get_unused(exp, &op_data->op_fid1, &cancels,
1792                                         LCK_CR, MDS_INODELOCK_LAYOUT);
1793         count += mdc_resource_get_unused(exp, &op_data->op_fid2, &cancels,
1794                                          LCK_CR, MDS_INODELOCK_LAYOUT);
1795
1796         req = ptlrpc_request_alloc(class_exp2cliimp(exp),
1797                                    &RQF_MDS_SWAP_LAYOUTS);
1798         if (req == NULL) {
1799                 ldlm_lock_list_put(&cancels, l_bl_ast, count);
1800                 return -ENOMEM;
1801         }
1802
1803         mdc_set_capa_size(req, &RMF_CAPA1, op_data->op_capa1);
1804         mdc_set_capa_size(req, &RMF_CAPA2, op_data->op_capa2);
1805
1806         rc = mdc_prep_elc_req(exp, req, MDS_SWAP_LAYOUTS, &cancels, count);
1807         if (rc) {
1808                 ptlrpc_request_free(req);
1809                 return rc;
1810         }
1811
1812         mdc_swap_layouts_pack(req, op_data);
1813
1814         payload = req_capsule_client_get(&req->rq_pill, &RMF_SWAP_LAYOUTS);
1815         LASSERT(payload);
1816
1817         *payload = *msl;
1818
1819         ptlrpc_request_set_replen(req);
1820
1821         rc = ptlrpc_queue_wait(req);
1822         if (rc)
1823                 goto out;
1824
1825 out:
1826         ptlrpc_req_finished(req);
1827         return rc;
1828 }
1829
1830 static int mdc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
1831                          void *karg, void *uarg)
1832 {
1833         struct obd_device *obd = exp->exp_obd;
1834         struct obd_ioctl_data *data = karg;
1835         struct obd_import *imp = obd->u.cli.cl_import;
1836         int rc;
1837
1838         if (!try_module_get(THIS_MODULE)) {
1839                 CERROR("Can't get module. Is it alive?");
1840                 return -EINVAL;
1841         }
1842         switch (cmd) {
1843         case OBD_IOC_CHANGELOG_SEND:
1844                 rc = mdc_ioc_changelog_send(obd, karg);
1845                 goto out;
1846         case OBD_IOC_CHANGELOG_CLEAR: {
1847                 struct ioc_changelog *icc = karg;
1848                 struct changelog_setinfo cs = {
1849                         .cs_recno = icc->icc_recno,
1850                         .cs_id = icc->icc_id
1851                 };
1852
1853                 rc = obd_set_info_async(NULL, exp, strlen(KEY_CHANGELOG_CLEAR),
1854                                         KEY_CHANGELOG_CLEAR, sizeof(cs), &cs,
1855                                         NULL);
1856                 goto out;
1857         }
1858         case OBD_IOC_FID2PATH:
1859                 rc = mdc_ioc_fid2path(exp, karg);
1860                 goto out;
1861         case LL_IOC_HSM_CT_START:
1862                 rc = mdc_ioc_hsm_ct_start(exp, karg);
1863                 /* ignore if it was already registered on this MDS. */
1864                 if (rc == -EEXIST)
1865                         rc = 0;
1866                 goto out;
1867         case LL_IOC_HSM_PROGRESS:
1868                 rc = mdc_ioc_hsm_progress(exp, karg);
1869                 goto out;
1870         case LL_IOC_HSM_STATE_GET:
1871                 rc = mdc_ioc_hsm_state_get(exp, karg);
1872                 goto out;
1873         case LL_IOC_HSM_STATE_SET:
1874                 rc = mdc_ioc_hsm_state_set(exp, karg);
1875                 goto out;
1876         case LL_IOC_HSM_ACTION:
1877                 rc = mdc_ioc_hsm_current_action(exp, karg);
1878                 goto out;
1879         case LL_IOC_HSM_REQUEST:
1880                 rc = mdc_ioc_hsm_request(exp, karg);
1881                 goto out;
1882         case OBD_IOC_CLIENT_RECOVER:
1883                 rc = ptlrpc_recover_import(imp, data->ioc_inlbuf1, 0);
1884                 if (rc < 0)
1885                         goto out;
1886                 rc = 0;
1887                 goto out;
1888         case IOC_OSC_SET_ACTIVE:
1889                 rc = ptlrpc_set_import_active(imp, data->ioc_offset);
1890                 goto out;
1891         case OBD_IOC_POLL_QUOTACHECK:
1892                 rc = mdc_quota_poll_check(exp, (struct if_quotacheck *)karg);
1893                 goto out;
1894         case OBD_IOC_PING_TARGET:
1895                 rc = ptlrpc_obd_ping(obd);
1896                 goto out;
1897         /*
1898          * Normally IOC_OBD_STATFS, OBD_IOC_QUOTACTL iocontrol are handled by
1899          * LMV instead of MDC. But when the cluster is upgraded from 1.8,
1900          * there'd be no LMV layer thus we might be called here. Eventually
1901          * this code should be removed.
1902          * bz20731, LU-592.
1903          */
1904         case IOC_OBD_STATFS: {
1905                 struct obd_statfs stat_buf = {0};
1906
1907                 if (*((__u32 *) data->ioc_inlbuf2) != 0) {
1908                         rc = -ENODEV;
1909                         goto out;
1910                 }
1911
1912                 /* copy UUID */
1913                 if (copy_to_user(data->ioc_pbuf2, obd2cli_tgt(obd),
1914                                  min_t(size_t, data->ioc_plen2,
1915                                                sizeof(struct obd_uuid)))) {
1916                         rc = -EFAULT;
1917                         goto out;
1918                 }
1919
1920                 rc = mdc_statfs(NULL, obd->obd_self_export, &stat_buf,
1921                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
1922                                 0);
1923                 if (rc != 0)
1924                         goto out;
1925
1926                 if (copy_to_user(data->ioc_pbuf1, &stat_buf,
1927                                  min_t(size_t, data->ioc_plen1,
1928                                                sizeof(stat_buf)))) {
1929                         rc = -EFAULT;
1930                         goto out;
1931                 }
1932
1933                 rc = 0;
1934                 goto out;
1935         }
1936         case OBD_IOC_QUOTACTL: {
1937                 struct if_quotactl *qctl = karg;
1938                 struct obd_quotactl *oqctl;
1939
1940                 OBD_ALLOC_PTR(oqctl);
1941                 if (oqctl == NULL) {
1942                         rc = -ENOMEM;
1943                         goto out;
1944                 }
1945
1946                 QCTL_COPY(oqctl, qctl);
1947                 rc = obd_quotactl(exp, oqctl);
1948                 if (rc == 0) {
1949                         QCTL_COPY(qctl, oqctl);
1950                         qctl->qc_valid = QC_MDTIDX;
1951                         qctl->obd_uuid = obd->u.cli.cl_target_uuid;
1952                 }
1953
1954                 OBD_FREE_PTR(oqctl);
1955                 goto out;
1956         }
1957         case LL_IOC_GET_CONNECT_FLAGS:
1958                 if (copy_to_user(uarg, exp_connect_flags_ptr(exp),
1959                                  sizeof(*exp_connect_flags_ptr(exp)))) {
1960                         rc = -EFAULT;
1961                         goto out;
1962                 }
1963
1964                 rc = 0;
1965                 goto out;
1966         case LL_IOC_LOV_SWAP_LAYOUTS:
1967                 rc = mdc_ioc_swap_layouts(exp, karg);
1968                 goto out;
1969         default:
1970                 CERROR("unrecognised ioctl: cmd = %#x\n", cmd);
1971                 rc = -ENOTTY;
1972                 goto out;
1973         }
1974 out:
1975         module_put(THIS_MODULE);
1976
1977         return rc;
1978 }
1979
1980 static int mdc_get_info_rpc(struct obd_export *exp,
1981                             u32 keylen, void *key,
1982                             int vallen, void *val)
1983 {
1984         struct obd_import      *imp = class_exp2cliimp(exp);
1985         struct ptlrpc_request  *req;
1986         char               *tmp;
1987         int                  rc = -EINVAL;
1988
1989         req = ptlrpc_request_alloc(imp, &RQF_MDS_GET_INFO);
1990         if (req == NULL)
1991                 return -ENOMEM;
1992
1993         req_capsule_set_size(&req->rq_pill, &RMF_GETINFO_KEY,
1994                              RCL_CLIENT, keylen);
1995         req_capsule_set_size(&req->rq_pill, &RMF_GETINFO_VALLEN,
1996                              RCL_CLIENT, sizeof(__u32));
1997
1998         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_GET_INFO);
1999         if (rc) {
2000                 ptlrpc_request_free(req);
2001                 return rc;
2002         }
2003
2004         tmp = req_capsule_client_get(&req->rq_pill, &RMF_GETINFO_KEY);
2005         memcpy(tmp, key, keylen);
2006         tmp = req_capsule_client_get(&req->rq_pill, &RMF_GETINFO_VALLEN);
2007         memcpy(tmp, &vallen, sizeof(__u32));
2008
2009         req_capsule_set_size(&req->rq_pill, &RMF_GETINFO_VAL,
2010                              RCL_SERVER, vallen);
2011         ptlrpc_request_set_replen(req);
2012
2013         rc = ptlrpc_queue_wait(req);
2014         /* -EREMOTE means the get_info result is partial, and it needs to
2015          * continue on another MDT, see fid2path part in lmv_iocontrol */
2016         if (rc == 0 || rc == -EREMOTE) {
2017                 tmp = req_capsule_server_get(&req->rq_pill, &RMF_GETINFO_VAL);
2018                 memcpy(val, tmp, vallen);
2019                 if (ptlrpc_rep_need_swab(req)) {
2020                         if (KEY_IS(KEY_FID2PATH))
2021                                 lustre_swab_fid2path(val);
2022                 }
2023         }
2024         ptlrpc_req_finished(req);
2025
2026         return rc;
2027 }
2028
2029 static void lustre_swab_hai(struct hsm_action_item *h)
2030 {
2031         __swab32s(&h->hai_len);
2032         __swab32s(&h->hai_action);
2033         lustre_swab_lu_fid(&h->hai_fid);
2034         lustre_swab_lu_fid(&h->hai_dfid);
2035         __swab64s(&h->hai_cookie);
2036         __swab64s(&h->hai_extent.offset);
2037         __swab64s(&h->hai_extent.length);
2038         __swab64s(&h->hai_gid);
2039 }
2040
2041 static void lustre_swab_hal(struct hsm_action_list *h)
2042 {
2043         struct hsm_action_item  *hai;
2044         int                      i;
2045
2046         __swab32s(&h->hal_version);
2047         __swab32s(&h->hal_count);
2048         __swab32s(&h->hal_archive_id);
2049         __swab64s(&h->hal_flags);
2050         hai = hai_zero(h);
2051         for (i = 0; i < h->hal_count; i++, hai = hai_next(hai))
2052                 lustre_swab_hai(hai);
2053 }
2054
2055 static void lustre_swab_kuch(struct kuc_hdr *l)
2056 {
2057         __swab16s(&l->kuc_magic);
2058         /* __u8 l->kuc_transport */
2059         __swab16s(&l->kuc_msgtype);
2060         __swab16s(&l->kuc_msglen);
2061 }
2062
2063 static int mdc_ioc_hsm_ct_start(struct obd_export *exp,
2064                                 struct lustre_kernelcomm *lk)
2065 {
2066         struct obd_import  *imp = class_exp2cliimp(exp);
2067         __u32               archive = lk->lk_data;
2068         int                 rc = 0;
2069
2070         if (lk->lk_group != KUC_GRP_HSM) {
2071                 CERROR("Bad copytool group %d\n", lk->lk_group);
2072                 return -EINVAL;
2073         }
2074
2075         CDEBUG(D_HSM, "CT start r%d w%d u%d g%d f%#x\n", lk->lk_rfd, lk->lk_wfd,
2076                lk->lk_uid, lk->lk_group, lk->lk_flags);
2077
2078         if (lk->lk_flags & LK_FLG_STOP) {
2079                 /* Unregister with the coordinator */
2080                 rc = mdc_ioc_hsm_ct_unregister(imp);
2081         } else {
2082                 rc = mdc_ioc_hsm_ct_register(imp, archive);
2083         }
2084
2085         return rc;
2086 }
2087
2088 /**
2089  * Send a message to any listening copytools
2090  * @param val KUC message (kuc_hdr + hsm_action_list)
2091  * @param len total length of message
2092  */
2093 static int mdc_hsm_copytool_send(int len, void *val)
2094 {
2095         struct kuc_hdr          *lh = (struct kuc_hdr *)val;
2096         struct hsm_action_list  *hal = (struct hsm_action_list *)(lh + 1);
2097         int                      rc;
2098
2099         if (len < sizeof(*lh) + sizeof(*hal)) {
2100                 CERROR("Short HSM message %d < %d\n", len,
2101                        (int) (sizeof(*lh) + sizeof(*hal)));
2102                 return -EPROTO;
2103         }
2104         if (lh->kuc_magic == __swab16(KUC_MAGIC)) {
2105                 lustre_swab_kuch(lh);
2106                 lustre_swab_hal(hal);
2107         } else if (lh->kuc_magic != KUC_MAGIC) {
2108                 CERROR("Bad magic %x!=%x\n", lh->kuc_magic, KUC_MAGIC);
2109                 return -EPROTO;
2110         }
2111
2112         CDEBUG(D_HSM,
2113                "Received message mg=%x t=%d m=%d l=%d actions=%d on %s\n",
2114                lh->kuc_magic, lh->kuc_transport, lh->kuc_msgtype,
2115                lh->kuc_msglen, hal->hal_count, hal->hal_fsname);
2116
2117         /* Broadcast to HSM listeners */
2118         rc = libcfs_kkuc_group_put(KUC_GRP_HSM, lh);
2119
2120         return rc;
2121 }
2122
2123 /**
2124  * callback function passed to kuc for re-registering each HSM copytool
2125  * running on MDC, after MDT shutdown/recovery.
2126  * @param data archive id served by the copytool
2127  * @param cb_arg callback argument (obd_import)
2128  */
2129 static int mdc_hsm_ct_reregister(__u32 data, void *cb_arg)
2130 {
2131         struct obd_import       *imp = (struct obd_import *)cb_arg;
2132         __u32                    archive = data;
2133         int                      rc;
2134
2135         CDEBUG(D_HA, "recover copytool registration to MDT (archive=%#x)\n",
2136                archive);
2137         rc = mdc_ioc_hsm_ct_register(imp, archive);
2138
2139         /* ignore error if the copytool is already registered */
2140         return ((rc != 0) && (rc != -EEXIST)) ? rc : 0;
2141 }
2142
2143 /**
2144  * Re-establish all kuc contexts with MDT
2145  * after MDT shutdown/recovery.
2146  */
2147 static int mdc_kuc_reregister(struct obd_import *imp)
2148 {
2149         /* re-register HSM agents */
2150         return libcfs_kkuc_group_foreach(KUC_GRP_HSM, mdc_hsm_ct_reregister,
2151                                          (void *)imp);
2152 }
2153
2154 static int mdc_set_info_async(const struct lu_env *env,
2155                               struct obd_export *exp,
2156                               u32 keylen, void *key,
2157                               u32 vallen, void *val,
2158                               struct ptlrpc_request_set *set)
2159 {
2160         struct obd_import       *imp = class_exp2cliimp(exp);
2161         int                      rc;
2162
2163         if (KEY_IS(KEY_READ_ONLY)) {
2164                 if (vallen != sizeof(int))
2165                         return -EINVAL;
2166
2167                 spin_lock(&imp->imp_lock);
2168                 if (*((int *)val)) {
2169                         imp->imp_connect_flags_orig |= OBD_CONNECT_RDONLY;
2170                         imp->imp_connect_data.ocd_connect_flags |=
2171                                                         OBD_CONNECT_RDONLY;
2172                 } else {
2173                         imp->imp_connect_flags_orig &= ~OBD_CONNECT_RDONLY;
2174                         imp->imp_connect_data.ocd_connect_flags &=
2175                                                         ~OBD_CONNECT_RDONLY;
2176                 }
2177                 spin_unlock(&imp->imp_lock);
2178
2179                 rc = do_set_info_async(imp, MDS_SET_INFO, LUSTRE_MDS_VERSION,
2180                                        keylen, key, vallen, val, set);
2181                 return rc;
2182         }
2183         if (KEY_IS(KEY_SPTLRPC_CONF)) {
2184                 sptlrpc_conf_client_adapt(exp->exp_obd);
2185                 return 0;
2186         }
2187         if (KEY_IS(KEY_FLUSH_CTX)) {
2188                 sptlrpc_import_flush_my_ctx(imp);
2189                 return 0;
2190         }
2191         if (KEY_IS(KEY_CHANGELOG_CLEAR)) {
2192                 rc = do_set_info_async(imp, MDS_SET_INFO, LUSTRE_MDS_VERSION,
2193                                        keylen, key, vallen, val, set);
2194                 return rc;
2195         }
2196         if (KEY_IS(KEY_HSM_COPYTOOL_SEND)) {
2197                 rc = mdc_hsm_copytool_send(vallen, val);
2198                 return rc;
2199         }
2200
2201         CERROR("Unknown key %s\n", (char *)key);
2202         return -EINVAL;
2203 }
2204
2205 static int mdc_get_info(const struct lu_env *env, struct obd_export *exp,
2206                         __u32 keylen, void *key, __u32 *vallen, void *val,
2207                         struct lov_stripe_md *lsm)
2208 {
2209         int rc = -EINVAL;
2210
2211         if (KEY_IS(KEY_MAX_EASIZE)) {
2212                 int mdsize, *max_easize;
2213
2214                 if (*vallen != sizeof(int))
2215                         return -EINVAL;
2216                 mdsize = *(int *)val;
2217                 if (mdsize > exp->exp_obd->u.cli.cl_max_mds_easize)
2218                         exp->exp_obd->u.cli.cl_max_mds_easize = mdsize;
2219                 max_easize = val;
2220                 *max_easize = exp->exp_obd->u.cli.cl_max_mds_easize;
2221                 return 0;
2222         } else if (KEY_IS(KEY_DEFAULT_EASIZE)) {
2223                 int *default_easize;
2224
2225                 if (*vallen != sizeof(int))
2226                         return -EINVAL;
2227                 default_easize = val;
2228                 *default_easize = exp->exp_obd->u.cli.cl_default_mds_easize;
2229                 return 0;
2230         } else if (KEY_IS(KEY_MAX_COOKIESIZE)) {
2231                 int mdsize, *max_cookiesize;
2232
2233                 if (*vallen != sizeof(int))
2234                         return -EINVAL;
2235                 mdsize = *(int *)val;
2236                 if (mdsize > exp->exp_obd->u.cli.cl_max_mds_cookiesize)
2237                         exp->exp_obd->u.cli.cl_max_mds_cookiesize = mdsize;
2238                 max_cookiesize = val;
2239                 *max_cookiesize = exp->exp_obd->u.cli.cl_max_mds_cookiesize;
2240                 return 0;
2241         } else if (KEY_IS(KEY_DEFAULT_COOKIESIZE)) {
2242                 int *default_cookiesize;
2243
2244                 if (*vallen != sizeof(int))
2245                         return -EINVAL;
2246                 default_cookiesize = val;
2247                 *default_cookiesize =
2248                         exp->exp_obd->u.cli.cl_default_mds_cookiesize;
2249                 return 0;
2250         } else if (KEY_IS(KEY_CONN_DATA)) {
2251                 struct obd_import *imp = class_exp2cliimp(exp);
2252                 struct obd_connect_data *data = val;
2253
2254                 if (*vallen != sizeof(*data))
2255                         return -EINVAL;
2256
2257                 *data = imp->imp_connect_data;
2258                 return 0;
2259         } else if (KEY_IS(KEY_TGT_COUNT)) {
2260                 *((int *)val) = 1;
2261                 return 0;
2262         }
2263
2264         rc = mdc_get_info_rpc(exp, keylen, key, *vallen, val);
2265
2266         return rc;
2267 }
2268
2269 static int mdc_sync(struct obd_export *exp, const struct lu_fid *fid,
2270                     struct obd_capa *oc, struct ptlrpc_request **request)
2271 {
2272         struct ptlrpc_request *req;
2273         int                 rc;
2274
2275         *request = NULL;
2276         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_MDS_SYNC);
2277         if (req == NULL)
2278                 return -ENOMEM;
2279
2280         mdc_set_capa_size(req, &RMF_CAPA1, oc);
2281
2282         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_SYNC);
2283         if (rc) {
2284                 ptlrpc_request_free(req);
2285                 return rc;
2286         }
2287
2288         mdc_pack_body(req, fid, oc, 0, 0, -1, 0);
2289
2290         ptlrpc_request_set_replen(req);
2291
2292         rc = ptlrpc_queue_wait(req);
2293         if (rc)
2294                 ptlrpc_req_finished(req);
2295         else
2296                 *request = req;
2297         return rc;
2298 }
2299
2300 static int mdc_import_event(struct obd_device *obd, struct obd_import *imp,
2301                             enum obd_import_event event)
2302 {
2303         int rc = 0;
2304
2305         LASSERT(imp->imp_obd == obd);
2306
2307         switch (event) {
2308         case IMP_EVENT_DISCON: {
2309 #if 0
2310                 /* XXX Pass event up to OBDs stack. used only for FLD now */
2311                 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DISCON, NULL);
2312 #endif
2313                 break;
2314         }
2315         case IMP_EVENT_INACTIVE: {
2316                 struct client_obd *cli = &obd->u.cli;
2317                 /*
2318                  * Flush current sequence to make client obtain new one
2319                  * from server in case of disconnect/reconnect.
2320                  */
2321                 if (cli->cl_seq != NULL)
2322                         seq_client_flush(cli->cl_seq);
2323
2324                 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
2325                 break;
2326         }
2327         case IMP_EVENT_INVALIDATE: {
2328                 struct ldlm_namespace *ns = obd->obd_namespace;
2329
2330                 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
2331
2332                 break;
2333         }
2334         case IMP_EVENT_ACTIVE:
2335                 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
2336                 /* redo the kuc registration after reconnecting */
2337                 if (rc == 0)
2338                         rc = mdc_kuc_reregister(imp);
2339                 break;
2340         case IMP_EVENT_OCD:
2341                 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
2342                 break;
2343         case IMP_EVENT_DEACTIVATE:
2344         case IMP_EVENT_ACTIVATE:
2345                 break;
2346         default:
2347                 CERROR("Unknown import event %x\n", event);
2348                 LBUG();
2349         }
2350         return rc;
2351 }
2352
2353 int mdc_fid_alloc(struct obd_export *exp, struct lu_fid *fid,
2354                   struct md_op_data *op_data)
2355 {
2356         struct client_obd *cli = &exp->exp_obd->u.cli;
2357         struct lu_client_seq *seq = cli->cl_seq;
2358
2359         return seq_client_alloc_fid(NULL, seq, fid);
2360 }
2361
2362 static struct obd_uuid *mdc_get_uuid(struct obd_export *exp)
2363 {
2364         struct client_obd *cli = &exp->exp_obd->u.cli;
2365
2366         return &cli->cl_target_uuid;
2367 }
2368
2369 /**
2370  * Determine whether the lock can be canceled before replaying it during
2371  * recovery, non zero value will be return if the lock can be canceled,
2372  * or zero returned for not
2373  */
2374 static int mdc_cancel_for_recovery(struct ldlm_lock *lock)
2375 {
2376         if (lock->l_resource->lr_type != LDLM_IBITS)
2377                 return 0;
2378
2379         /* FIXME: if we ever get into a situation where there are too many
2380          * opened files with open locks on a single node, then we really
2381          * should replay these open locks to reget it */
2382         if (lock->l_policy_data.l_inodebits.bits & MDS_INODELOCK_OPEN)
2383                 return 0;
2384
2385         return 1;
2386 }
2387
2388 static int mdc_resource_inode_free(struct ldlm_resource *res)
2389 {
2390         if (res->lr_lvb_inode)
2391                 res->lr_lvb_inode = NULL;
2392
2393         return 0;
2394 }
2395
2396 static struct ldlm_valblock_ops inode_lvbo = {
2397         .lvbo_free = mdc_resource_inode_free,
2398 };
2399
2400 static int mdc_llog_init(struct obd_device *obd)
2401 {
2402         struct obd_llog_group   *olg = &obd->obd_olg;
2403         struct llog_ctxt        *ctxt;
2404         int                      rc;
2405
2406         rc = llog_setup(NULL, obd, olg, LLOG_CHANGELOG_REPL_CTXT, obd,
2407                         &llog_client_ops);
2408         if (rc)
2409                 return rc;
2410
2411         ctxt = llog_group_get_ctxt(olg, LLOG_CHANGELOG_REPL_CTXT);
2412         llog_initiator_connect(ctxt);
2413         llog_ctxt_put(ctxt);
2414
2415         return 0;
2416 }
2417
2418 static void mdc_llog_finish(struct obd_device *obd)
2419 {
2420         struct llog_ctxt *ctxt;
2421
2422         ctxt = llog_get_context(obd, LLOG_CHANGELOG_REPL_CTXT);
2423         if (ctxt)
2424                 llog_cleanup(NULL, ctxt);
2425 }
2426
2427 static int mdc_setup(struct obd_device *obd, struct lustre_cfg *cfg)
2428 {
2429         struct client_obd *cli = &obd->u.cli;
2430         struct lprocfs_static_vars lvars = { NULL };
2431         int rc;
2432
2433         OBD_ALLOC(cli->cl_rpc_lock, sizeof(*cli->cl_rpc_lock));
2434         if (!cli->cl_rpc_lock)
2435                 return -ENOMEM;
2436         mdc_init_rpc_lock(cli->cl_rpc_lock);
2437
2438         ptlrpcd_addref();
2439
2440         OBD_ALLOC(cli->cl_close_lock, sizeof(*cli->cl_close_lock));
2441         if (!cli->cl_close_lock) {
2442                 rc = -ENOMEM;
2443                 goto err_rpc_lock;
2444         }
2445         mdc_init_rpc_lock(cli->cl_close_lock);
2446
2447         rc = client_obd_setup(obd, cfg);
2448         if (rc)
2449                 goto err_close_lock;
2450         lprocfs_mdc_init_vars(&lvars);
2451         lprocfs_obd_setup(obd, lvars.obd_vars);
2452         sptlrpc_lprocfs_cliobd_attach(obd);
2453         ptlrpc_lprocfs_register_obd(obd);
2454
2455         ns_register_cancel(obd->obd_namespace, mdc_cancel_for_recovery);
2456
2457         obd->obd_namespace->ns_lvbo = &inode_lvbo;
2458
2459         rc = mdc_llog_init(obd);
2460         if (rc) {
2461                 mdc_cleanup(obd);
2462                 CERROR("failed to setup llogging subsystems\n");
2463         }
2464
2465         return rc;
2466
2467 err_close_lock:
2468         OBD_FREE(cli->cl_close_lock, sizeof(*cli->cl_close_lock));
2469 err_rpc_lock:
2470         OBD_FREE(cli->cl_rpc_lock, sizeof(*cli->cl_rpc_lock));
2471         ptlrpcd_decref();
2472         return rc;
2473 }
2474
2475 /* Initialize the default and maximum LOV EA and cookie sizes.  This allows
2476  * us to make MDS RPCs with large enough reply buffers to hold a default
2477  * sized EA and cookie without having to calculate this (via a call into the
2478  * LOV + OSCs) each time we make an RPC.  The maximum size is also tracked
2479  * but not used to avoid wastefully vmalloc()'ing large reply buffers when
2480  * a large number of stripes is possible.  If a larger reply buffer is
2481  * required it will be reallocated in the ptlrpc layer due to overflow.
2482  */
2483 static int mdc_init_ea_size(struct obd_export *exp, int easize,
2484                             int def_easize, int cookiesize, int def_cookiesize)
2485 {
2486         struct obd_device *obd = exp->exp_obd;
2487         struct client_obd *cli = &obd->u.cli;
2488
2489         if (cli->cl_max_mds_easize < easize)
2490                 cli->cl_max_mds_easize = easize;
2491
2492         if (cli->cl_default_mds_easize < def_easize)
2493                 cli->cl_default_mds_easize = def_easize;
2494
2495         if (cli->cl_max_mds_cookiesize < cookiesize)
2496                 cli->cl_max_mds_cookiesize = cookiesize;
2497
2498         if (cli->cl_default_mds_cookiesize < def_cookiesize)
2499                 cli->cl_default_mds_cookiesize = def_cookiesize;
2500
2501         return 0;
2502 }
2503
2504 static int mdc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
2505 {
2506         switch (stage) {
2507         case OBD_CLEANUP_EARLY:
2508                 break;
2509         case OBD_CLEANUP_EXPORTS:
2510                 /* Failsafe, ok if racy */
2511                 if (obd->obd_type->typ_refcnt <= 1)
2512                         libcfs_kkuc_group_rem(0, KUC_GRP_HSM);
2513
2514                 obd_cleanup_client_import(obd);
2515                 ptlrpc_lprocfs_unregister_obd(obd);
2516                 lprocfs_obd_cleanup(obd);
2517
2518                 mdc_llog_finish(obd);
2519                 break;
2520         }
2521         return 0;
2522 }
2523
2524 static int mdc_cleanup(struct obd_device *obd)
2525 {
2526         struct client_obd *cli = &obd->u.cli;
2527
2528         OBD_FREE(cli->cl_rpc_lock, sizeof(*cli->cl_rpc_lock));
2529         OBD_FREE(cli->cl_close_lock, sizeof(*cli->cl_close_lock));
2530
2531         ptlrpcd_decref();
2532
2533         return client_obd_cleanup(obd);
2534 }
2535
2536 static int mdc_process_config(struct obd_device *obd, u32 len, void *buf)
2537 {
2538         struct lustre_cfg *lcfg = buf;
2539         struct lprocfs_static_vars lvars = { NULL };
2540         int rc = 0;
2541
2542         lprocfs_mdc_init_vars(&lvars);
2543         switch (lcfg->lcfg_command) {
2544         default:
2545                 rc = class_process_proc_param(PARAM_MDC, lvars.obd_vars,
2546                                               lcfg, obd);
2547                 if (rc > 0)
2548                         rc = 0;
2549                 break;
2550         }
2551         return rc;
2552 }
2553
2554
2555 /* get remote permission for current user on fid */
2556 static int mdc_get_remote_perm(struct obd_export *exp, const struct lu_fid *fid,
2557                                struct obd_capa *oc, __u32 suppgid,
2558                                struct ptlrpc_request **request)
2559 {
2560         struct ptlrpc_request  *req;
2561         int                 rc;
2562
2563         LASSERT(client_is_remote(exp));
2564
2565         *request = NULL;
2566         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_MDS_GETATTR);
2567         if (req == NULL)
2568                 return -ENOMEM;
2569
2570         mdc_set_capa_size(req, &RMF_CAPA1, oc);
2571
2572         rc = ptlrpc_request_pack(req, LUSTRE_MDS_VERSION, MDS_GETATTR);
2573         if (rc) {
2574                 ptlrpc_request_free(req);
2575                 return rc;
2576         }
2577
2578         mdc_pack_body(req, fid, oc, OBD_MD_FLRMTPERM, 0, suppgid, 0);
2579
2580         req_capsule_set_size(&req->rq_pill, &RMF_ACL, RCL_SERVER,
2581                              sizeof(struct mdt_remote_perm));
2582
2583         ptlrpc_request_set_replen(req);
2584
2585         rc = ptlrpc_queue_wait(req);
2586         if (rc)
2587                 ptlrpc_req_finished(req);
2588         else
2589                 *request = req;
2590         return rc;
2591 }
2592
2593 static int mdc_interpret_renew_capa(const struct lu_env *env,
2594                                     struct ptlrpc_request *req, void *args,
2595                                     int status)
2596 {
2597         struct mdc_renew_capa_args *ra = args;
2598         struct mdt_body *body = NULL;
2599         struct lustre_capa *capa;
2600
2601         if (status) {
2602                 capa = ERR_PTR(status);
2603                 goto out;
2604         }
2605
2606         body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
2607         if (body == NULL) {
2608                 capa = ERR_PTR(-EFAULT);
2609                 goto out;
2610         }
2611
2612         if ((body->valid & OBD_MD_FLOSSCAPA) == 0) {
2613                 capa = ERR_PTR(-ENOENT);
2614                 goto out;
2615         }
2616
2617         capa = req_capsule_server_get(&req->rq_pill, &RMF_CAPA2);
2618         if (!capa) {
2619                 capa = ERR_PTR(-EFAULT);
2620                 goto out;
2621         }
2622 out:
2623         ra->ra_cb(ra->ra_oc, capa);
2624         return 0;
2625 }
2626
2627 static int mdc_renew_capa(struct obd_export *exp, struct obd_capa *oc,
2628                           renew_capa_cb_t cb)
2629 {
2630         struct ptlrpc_request *req;
2631         struct mdc_renew_capa_args *ra;
2632
2633         req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), &RQF_MDS_GETATTR,
2634                                         LUSTRE_MDS_VERSION, MDS_GETATTR);
2635         if (req == NULL)
2636                 return -ENOMEM;
2637
2638         /* NB, OBD_MD_FLOSSCAPA is set here, but it doesn't necessarily mean the
2639          * capa to renew is oss capa.
2640          */
2641         mdc_pack_body(req, &oc->c_capa.lc_fid, oc, OBD_MD_FLOSSCAPA, 0, -1, 0);
2642         ptlrpc_request_set_replen(req);
2643
2644         CLASSERT(sizeof(*ra) <= sizeof(req->rq_async_args));
2645         ra = ptlrpc_req_async_args(req);
2646         ra->ra_oc = oc;
2647         ra->ra_cb = cb;
2648         req->rq_interpret_reply = mdc_interpret_renew_capa;
2649         ptlrpcd_add_req(req, PDL_POLICY_LOCAL, -1);
2650         return 0;
2651 }
2652
2653 static struct obd_ops mdc_obd_ops = {
2654         .o_owner            = THIS_MODULE,
2655         .o_setup            = mdc_setup,
2656         .o_precleanup       = mdc_precleanup,
2657         .o_cleanup        = mdc_cleanup,
2658         .o_add_conn      = client_import_add_conn,
2659         .o_del_conn      = client_import_del_conn,
2660         .o_connect          = client_connect_import,
2661         .o_disconnect       = client_disconnect_export,
2662         .o_iocontrol    = mdc_iocontrol,
2663         .o_set_info_async   = mdc_set_info_async,
2664         .o_statfs          = mdc_statfs,
2665         .o_fid_init         = client_fid_init,
2666         .o_fid_fini         = client_fid_fini,
2667         .o_fid_alloc    = mdc_fid_alloc,
2668         .o_import_event     = mdc_import_event,
2669         .o_get_info      = mdc_get_info,
2670         .o_process_config   = mdc_process_config,
2671         .o_get_uuid      = mdc_get_uuid,
2672         .o_quotactl      = mdc_quotactl,
2673         .o_quotacheck       = mdc_quotacheck
2674 };
2675
2676 static struct md_ops mdc_md_ops = {
2677         .m_getstatus    = mdc_getstatus,
2678         .m_null_inode       = mdc_null_inode,
2679         .m_find_cbdata      = mdc_find_cbdata,
2680         .m_close            = mdc_close,
2681         .m_create          = mdc_create,
2682         .m_done_writing     = mdc_done_writing,
2683         .m_enqueue        = mdc_enqueue,
2684         .m_getattr        = mdc_getattr,
2685         .m_getattr_name     = mdc_getattr_name,
2686         .m_intent_lock      = mdc_intent_lock,
2687         .m_link      = mdc_link,
2688         .m_is_subdir    = mdc_is_subdir,
2689         .m_rename          = mdc_rename,
2690         .m_setattr        = mdc_setattr,
2691         .m_setxattr      = mdc_setxattr,
2692         .m_getxattr      = mdc_getxattr,
2693         .m_sync      = mdc_sync,
2694         .m_readpage      = mdc_readpage,
2695         .m_unlink          = mdc_unlink,
2696         .m_cancel_unused    = mdc_cancel_unused,
2697         .m_init_ea_size     = mdc_init_ea_size,
2698         .m_set_lock_data    = mdc_set_lock_data,
2699         .m_lock_match       = mdc_lock_match,
2700         .m_get_lustre_md    = mdc_get_lustre_md,
2701         .m_free_lustre_md   = mdc_free_lustre_md,
2702         .m_set_open_replay_data = mdc_set_open_replay_data,
2703         .m_clear_open_replay_data = mdc_clear_open_replay_data,
2704         .m_renew_capa       = mdc_renew_capa,
2705         .m_unpack_capa      = mdc_unpack_capa,
2706         .m_get_remote_perm  = mdc_get_remote_perm,
2707         .m_intent_getattr_async = mdc_intent_getattr_async,
2708         .m_revalidate_lock      = mdc_revalidate_lock
2709 };
2710
2711 static int __init mdc_init(void)
2712 {
2713         struct lprocfs_static_vars lvars = { NULL };
2714
2715         lprocfs_mdc_init_vars(&lvars);
2716
2717         return class_register_type(&mdc_obd_ops, &mdc_md_ops, lvars.module_vars,
2718                                  LUSTRE_MDC_NAME, NULL);
2719 }
2720
2721 static void /*__exit*/ mdc_exit(void)
2722 {
2723         class_unregister_type(LUSTRE_MDC_NAME);
2724 }
2725
2726 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
2727 MODULE_DESCRIPTION("Lustre Metadata Client");
2728 MODULE_LICENSE("GPL");
2729
2730 module_init(mdc_init);
2731 module_exit(mdc_exit);