lustre/mds/mds_fs.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  *  mds/mds_fs.c
   5  *  Lustre Metadata Server (MDS) filesystem interface code
   6  *
   7  *  Copyright (C) 2002, 2003 Cluster File Systems, Inc.
   8  *   Author: Andreas Dilger <adilger@clusterfs.com>
   9  *
  10  *   This file is part of Lustre, http://www.lustre.org.
  11  *
  12  *   Lustre is free software; you can redistribute it and/or
  13  *   modify it under the terms of version 2 of the GNU General Public
  14  *   License as published by the Free Software Foundation.
  15  *
  16  *   Lustre is distributed in the hope that it will be useful,
  17  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  19  *   GNU General Public License for more details.
  20  *
  21  *   You should have received a copy of the GNU General Public License
  22  *   along with Lustre; if not, write to the Free Software
  23  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  24  */
  25
  26 #ifndef EXPORT_SYMTAB
  27 # define EXPORT_SYMTAB
  28 #endif
  29 #define DEBUG_SUBSYSTEM S_MDS
  30
  31 #include <linux/module.h>
  32 #include <linux/kmod.h>
  33 #include <linux/version.h>
  34 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
  35 #include <linux/mount.h>
  36 #endif
  37 #include <linux/lustre_mds.h>
  38 #include <linux/obd_class.h>
  39 #include <linux/obd_support.h>
  40 #include <linux/lustre_lib.h>
  41 #include <linux/lustre_fsfilt.h>
  42 #include <portals/list.h>
  43
  44 #include "mds_internal.h"
  45
  46 /* This limit is arbitrary, but for now we fit it in 1 page (32k clients) */
  47 #define MDS_MAX_CLIENTS (PAGE_SIZE * 8)
  48 #define MDS_MAX_CLIENT_WORDS (MDS_MAX_CLIENTS / sizeof(unsigned long))
  49
  50 #define LAST_RCVD "last_rcvd"
  51 #define LOV_OBJID "lov_objid"
  52
  53 /* Add client data to the MDS.  We use a bitmap to locate a free space
  54  * in the last_rcvd file if cl_off is -1 (i.e. a new client).
  55  * Otherwise, we have just read the data from the last_rcvd file and
  56  * we know its offset.
  57  */
  58 int mds_client_add(struct obd_device *obd, struct mds_obd *mds,
  59                    struct mds_export_data *med, int cl_idx)
  60 {
  61         unsigned long *bitmap = mds->mds_client_bitmap;
  62         int new_client = (cl_idx == -1);
  63         ENTRY;
  64
  65         LASSERT(bitmap != NULL);
  66
  67         /* XXX if mcd_uuid were a real obd_uuid, I could use obd_uuid_equals */
  68         if (!strcmp(med->med_mcd->mcd_uuid, obd->obd_uuid.uuid))
  69                 RETURN(0);
  70
  71         /* the bitmap operations can handle cl_idx > sizeof(long) * 8, so
  72          * there's no need for extra complication here
  73          */
  74         if (new_client) {
  75                 cl_idx = find_first_zero_bit(bitmap, MDS_MAX_CLIENTS);
  76         repeat:
  77                 if (cl_idx >= MDS_MAX_CLIENTS) {
  78                         CERROR("no room for clients - fix MDS_MAX_CLIENTS\n");
  79                         return -ENOMEM;
  80                 }
  81                 if (test_and_set_bit(cl_idx, bitmap)) {
  82                         cl_idx = find_next_zero_bit(bitmap, MDS_MAX_CLIENTS,
  83                                                     cl_idx);
  84                         goto repeat;
  85                 }
  86         } else {
  87                 if (test_and_set_bit(cl_idx, bitmap)) {
  88                         CERROR("MDS client %d: bit already set in bitmap!!\n",
  89                                cl_idx);
  90                         LBUG();
  91                 }
  92         }
  93
  94         CDEBUG(D_INFO, "client at index %d with UUID '%s' added\n",
  95                cl_idx, med->med_mcd->mcd_uuid);
  96
  97         med->med_idx = cl_idx;
  98         med->med_off = MDS_LR_CLIENT_START + (cl_idx * MDS_LR_CLIENT_SIZE);
  99
 100         if (new_client) {
 101                 struct obd_run_ctxt saved;
 102                 loff_t off = med->med_off;
 103                 struct file *file = mds->mds_rcvd_filp;
 104                 void *handle;
 105                 int rc, err;
 106
 107                 push_ctxt(&saved, &obd->obd_ctxt, NULL);
 108                 /* We need to start a transaction here first, to avoid a
 109                  * possible ordering deadlock on last_rcvd->i_sem and the
 110                  * journal lock. In most places we start the journal handle
 111                  * first (because we do compound transactions), and then
 112                  * later do the write into last_rcvd, which gets i_sem.
 113                  *
 114                  * Without this transaction, clients connecting at the same
 115                  * time other MDS operations are ongoing get last_rcvd->i_sem
 116                  * first (in generic_file_write()) and start the journal
 117                  * transaction afterwards, and can deadlock with other ops.
 118                  *
 119                  * We use FSFILT_OP_SETATTR because it is smallest, but all
 120                  * ops include enough space for the last_rcvd update so we
 121                  * could use any of them, or maybe an FSFILT_OP_NONE is best?
 122                  */
 123                 handle = fsfilt_start(obd, file->f_dentry->d_inode,
 124                                       FSFILT_OP_SETATTR, NULL);
 125                 if (IS_ERR(handle)) {
 126                         rc = PTR_ERR(handle);
 127                         CERROR("unable to start transaction: rc %d\n", rc);
 128                 } else {
 129                         rc = fsfilt_write_record(obd, file, med->med_mcd,
 130                                                  sizeof(*med->med_mcd),
 131                                                  &off, 1);
 132                         err = fsfilt_commit(obd, file->f_dentry->d_inode,
 133                                             handle, 1);
 134                         if (rc == 0)
 135                                 rc = err;
 136                 }
 137                 pop_ctxt(&saved, &obd->obd_ctxt, NULL);
 138
 139                 if (rc)
 140                         return rc;
 141                 CDEBUG(D_INFO, "wrote client mcd at idx %u off %llu (len %u)\n",
 142                        med->med_idx, med->med_off,
 143                        (unsigned int)sizeof(*med->med_mcd));
 144         }
 145         return 0;
 146 }
 147
 148 int mds_client_free(struct obd_export *exp, int clear_client)
 149 {
 150         struct mds_export_data *med = &exp->exp_mds_data;
 151         struct mds_obd *mds = &exp->exp_obd->u.mds;
 152         struct obd_device *obd = exp->exp_obd;
 153         struct mds_client_data zero_mcd;
 154         struct obd_run_ctxt saved;
 155         int rc;
 156         unsigned long *bitmap = mds->mds_client_bitmap;
 157
 158         LASSERT(bitmap);
 159         if (!med->med_mcd)
 160                 RETURN(0);
 161
 162         /* XXX if mcd_uuid were a real obd_uuid, I could use obd_uuid_equals */
 163         if (!strcmp(med->med_mcd->mcd_uuid, obd->obd_uuid.uuid))
 164                 GOTO(free_and_out, 0);
 165
 166         CDEBUG(D_INFO, "freeing client at index %u (%lld)with UUID '%s'\n",
 167                med->med_idx, med->med_off, med->med_mcd->mcd_uuid);
 168
 169         if (!test_and_clear_bit(med->med_idx, bitmap)) {
 170                 CERROR("MDS client %u: bit already clear in bitmap!!\n",
 171                        med->med_idx);
 172                 LBUG();
 173         }
 174
 175         if (clear_client) {
 176                 memset(&zero_mcd, 0, sizeof zero_mcd);
 177                 push_ctxt(&saved, &obd->obd_ctxt, NULL);
 178                 rc = fsfilt_write_record(obd, mds->mds_rcvd_filp,
 179                                               &zero_mcd, sizeof(zero_mcd),
 180                                               &med->med_off, 1);
 181                 pop_ctxt(&saved, &obd->obd_ctxt, NULL);
 182
 183                 CDEBUG(rc == 0 ? D_INFO : D_ERROR,
 184                        "zeroing out client %s off %u in %s rc %d\n",
 185                        med->med_mcd->mcd_uuid, med->med_idx, LAST_RCVD, rc);
 186         }
 187
 188  free_and_out:
 189         OBD_FREE(med->med_mcd, sizeof(*med->med_mcd));
 190
 191         return 0;
 192 }
 193
 194 static int mds_server_free_data(struct mds_obd *mds)
 195 {
 196         OBD_FREE(mds->mds_client_bitmap,
 197                  MDS_MAX_CLIENT_WORDS * sizeof(unsigned long));
 198         OBD_FREE(mds->mds_server_data, sizeof(*mds->mds_server_data));
 199         mds->mds_server_data = NULL;
 200
 201         return 0;
 202 }
 203
 204 static int mds_read_last_rcvd(struct obd_device *obd, struct file *file)
 205 {
 206         struct mds_obd *mds = &obd->u.mds;
 207         struct mds_server_data *msd;
 208         struct mds_client_data *mcd = NULL;
 209         loff_t off = 0;
 210         unsigned long last_rcvd_size = file->f_dentry->d_inode->i_size;
 211         __u64 last_transno = 0;
 212         __u64 mount_count;
 213         int cl_idx, rc = 0;
 214         ENTRY;
 215
 216         LASSERT(sizeof(struct mds_client_data) == MDS_LR_CLIENT_SIZE);
 217         LASSERT(sizeof(struct mds_server_data) <= MDS_LR_SERVER_SIZE);
 218
 219         OBD_ALLOC_WAIT(msd, sizeof(*msd));
 220         if (!msd)
 221                 RETURN(-ENOMEM);
 222
 223         OBD_ALLOC_WAIT(mds->mds_client_bitmap,
 224                   MDS_MAX_CLIENT_WORDS * sizeof(unsigned long));
 225         if (!mds->mds_client_bitmap) {
 226                 OBD_FREE(msd, sizeof(*msd));
 227                 RETURN(-ENOMEM);
 228         }
 229
 230         mds->mds_server_data = msd;
 231
 232         if (last_rcvd_size == 0) {
 233                 CWARN("%s: initializing new %s\n", obd->obd_name, LAST_RCVD);
 234                 memcpy(msd->msd_uuid, obd->obd_uuid.uuid,sizeof(msd->msd_uuid));
 235                 msd->msd_server_size = cpu_to_le32(MDS_LR_SERVER_SIZE);
 236                 msd->msd_client_start = cpu_to_le32(MDS_LR_CLIENT_START);
 237                 msd->msd_client_size = cpu_to_le16(MDS_LR_CLIENT_SIZE);
 238                 msd->msd_feature_rocompat = cpu_to_le32(MDS_ROCOMPAT_LOVOBJID);
 239                 rc = fsfilt_write_record(obd, file, msd, sizeof(*msd), &off, 1);
 240
 241                 if (rc == 0)
 242                         RETURN(0);
 243
 244                 CERROR("%s: error writing new MSD: %d\n", obd->obd_name, rc);
 245                 GOTO(err_msd, rc);
 246         }
 247
 248         rc = fsfilt_read_record(obd, file, msd, sizeof(*msd), &off);
 249         if (rc) {
 250                 CERROR("error reading MDS %s: rc = %d\n", LAST_RCVD, rc);
 251                 GOTO(err_msd, rc);
 252         }
 253         if (!msd->msd_server_size)
 254                 msd->msd_server_size = cpu_to_le32(MDS_LR_SERVER_SIZE);
 255         if (!msd->msd_client_start)
 256                 msd->msd_client_start = cpu_to_le32(MDS_LR_CLIENT_START);
 257         if (!msd->msd_client_size)
 258                 msd->msd_client_size = cpu_to_le16(MDS_LR_CLIENT_SIZE);
 259
 260         if (msd->msd_feature_incompat & ~cpu_to_le32(MDS_INCOMPAT_SUPP)) {
 261                 CERROR("unsupported incompat feature %x\n",
 262                        le32_to_cpu(msd->msd_feature_incompat) &
 263                        ~MDS_INCOMPAT_SUPP);
 264                 GOTO(err_msd, rc = -EINVAL);
 265         }
 266         /* XXX updating existing b_devel fs only, can be removed in future */
 267         msd->msd_feature_rocompat = cpu_to_le32(MDS_ROCOMPAT_LOVOBJID);
 268         if (msd->msd_feature_rocompat & ~cpu_to_le32(MDS_ROCOMPAT_SUPP)) {
 269                 CERROR("unsupported read-only feature %x\n",
 270                        le32_to_cpu(msd->msd_feature_rocompat) &
 271                        ~MDS_ROCOMPAT_SUPP);
 272                 /* Do something like remount filesystem read-only */
 273                 GOTO(err_msd, rc = -EINVAL);
 274         }
 275
 276         last_transno = le64_to_cpu(msd->msd_last_transno);
 277         mds->mds_last_transno = last_transno;
 278
 279         mount_count = le64_to_cpu(msd->msd_mount_count);
 280         mds->mds_mount_count = mount_count;
 281
 282         CDEBUG(D_INODE, "%s: server last_transno: "LPU64"\n",
 283                obd->obd_name, last_transno);
 284         CDEBUG(D_INODE, "%s: server mount_count: "LPU64"\n",
 285                obd->obd_name, mount_count);
 286         CDEBUG(D_INODE, "%s: server data size: %u\n",
 287                obd->obd_name, le32_to_cpu(msd->msd_server_size));
 288         CDEBUG(D_INODE, "%s: per-client data start: %u\n",
 289                obd->obd_name, le32_to_cpu(msd->msd_client_start));
 290         CDEBUG(D_INODE, "%s: per-client data size: %u\n",
 291                obd->obd_name, le32_to_cpu(msd->msd_client_size));
 292         CDEBUG(D_INODE, "%s: last_rcvd size: %lu\n",
 293                obd->obd_name, last_rcvd_size);
 294         CDEBUG(D_INODE, "%s: last_rcvd clients: %lu\n", obd->obd_name,
 295                last_rcvd_size <= MDS_LR_CLIENT_START ? 0 :
 296                (last_rcvd_size - MDS_LR_CLIENT_START) / MDS_LR_CLIENT_SIZE);
 297
 298         /* When we do a clean MDS shutdown, we save the last_transno into
 299          * the header.  If we find clients with higher last_transno values
 300          * then those clients may need recovery done. */
 301         for (cl_idx = 0, off = le32_to_cpu(msd->msd_client_start);
 302              off < last_rcvd_size; cl_idx++) {
 303                 __u64 last_transno;
 304                 int mount_age;
 305
 306                 if (!mcd) {
 307                         OBD_ALLOC_WAIT(mcd, sizeof(*mcd));
 308                         if (!mcd)
 309                                 GOTO(err_client, rc = -ENOMEM);
 310                 }
 311
 312                 /* Don't assume off is incremented properly by
 313                  * fsfilt_read_record(), in case sizeof(*mcd)
 314                  * isn't the same as msd->msd_client_size.  */
 315                 off = le32_to_cpu(msd->msd_client_start) +
 316                         cl_idx * le16_to_cpu(msd->msd_client_size);
 317                 rc = fsfilt_read_record(obd, file, mcd, sizeof(*mcd), &off);
 318                 if (rc) {
 319                         CERROR("error reading MDS %s idx %d, off %llu: rc %d\n",
 320                                LAST_RCVD, cl_idx, off, rc);
 321                         break; /* read error shouldn't cause startup to fail */
 322                 }
 323
 324                 if (mcd->mcd_uuid[0] == '\0') {
 325                         CDEBUG(D_INFO, "skipping zeroed client at offset %d\n",
 326                                cl_idx);
 327                         continue;
 328                 }
 329
 330                 last_transno = le64_to_cpu(mcd->mcd_last_transno);
 331
 332                 /* These exports are cleaned up by mds_disconnect(), so they
 333                  * need to be set up like real exports as mds_connect() does.
 334                  */
 335                 mount_age = mount_count - le64_to_cpu(mcd->mcd_mount_count);
 336                 if (mount_age < MDS_MOUNT_RECOV) {
 337                         struct obd_export *exp = class_new_export(obd);
 338                         struct mds_export_data *med;
 339                         CDEBUG(D_HA, "RCVRNG CLIENT uuid: %s idx: %d lr: "LPU64
 340                                " srv lr: "LPU64" mnt: "LPU64" last mount: "
 341                                LPU64"\n", mcd->mcd_uuid, cl_idx,
 342                                last_transno, le64_to_cpu(msd->msd_last_transno),
 343                                le64_to_cpu(mcd->mcd_mount_count), mount_count);
 344                         if (exp == NULL)
 345                                 GOTO(err_client, rc = -ENOMEM);
 346
 347                         memcpy(&exp->exp_client_uuid.uuid, mcd->mcd_uuid,
 348                                sizeof exp->exp_client_uuid.uuid);
 349                         med = &exp->exp_mds_data;
 350                         med->med_mcd = mcd;
 351                         mds_client_add(obd, mds, med, cl_idx);
 352                         /* create helper if export init gets more complex */
 353                         INIT_LIST_HEAD(&med->med_open_head);
 354                         spin_lock_init(&med->med_open_lock);
 355
 356                         mcd = NULL;
 357                         obd->obd_recoverable_clients++;
 358                         obd->obd_max_recoverable_clients++;
 359                         class_export_put(exp);
 360                 } else {
 361                         CDEBUG(D_INFO, "discarded client %d, UUID '%s', count "
 362                                LPU64"\n", cl_idx, mcd->mcd_uuid,
 363                                le64_to_cpu(mcd->mcd_mount_count));
 364                 }
 365
 366                 CDEBUG(D_OTHER, "client at idx %d has last_transno = "LPU64"\n",
 367                        cl_idx, last_transno);
 368
 369                 if (last_transno > mds->mds_last_transno)
 370                         mds->mds_last_transno = last_transno;
 371         }
 372
 373         obd->obd_last_committed = mds->mds_last_transno;
 374         if (obd->obd_recoverable_clients) {
 375                 CWARN("RECOVERY: %d recoverable clients, last_transno "
 376                        LPU64"\n", obd->obd_recoverable_clients,
 377                        mds->mds_last_transno);
 378                 obd->obd_next_recovery_transno = obd->obd_last_committed + 1;
 379                 obd->obd_recovering = 1;
 380         }
 381
 382         if (mcd)
 383                 OBD_FREE(mcd, sizeof(*mcd));
 384
 385         return 0;
 386
 387 err_client:
 388         class_disconnect_exports(obd, 0);
 389 err_msd:
 390         mds_server_free_data(mds);
 391         return rc;
 392 }
 393
 394 int mds_fs_setup(struct obd_device *obd, struct vfsmount *mnt)
 395 {
 396         struct mds_obd *mds = &obd->u.mds;
 397         struct obd_run_ctxt saved;
 398         struct dentry *dentry;
 399         struct file *file;
 400         int rc;
 401         ENTRY;
 402
 403
 404         /* Get rid of unneeded supplementary groups */
 405         current->ngroups = 0;
 406         memset(current->groups, 0, sizeof(current->groups));
 407
 408         mds->mds_vfsmnt = mnt;
 409         mds->mds_sb = mnt->mnt_root->d_inode->i_sb;
 410
 411         fsfilt_setup(obd, mds->mds_sb);
 412
 413         OBD_SET_CTXT_MAGIC(&obd->obd_ctxt);
 414         obd->obd_ctxt.pwdmnt = mnt;
 415         obd->obd_ctxt.pwd = mnt->mnt_root;
 416         obd->obd_ctxt.fs = get_ds();
 417         obd->obd_ctxt.cb_ops = mds_lvfs_ops;
 418
 419         /* setup the directory tree */
 420         push_ctxt(&saved, &obd->obd_ctxt, NULL);
 421         dentry = simple_mkdir(current->fs->pwd, "ROOT", 0755);
 422         if (IS_ERR(dentry)) {
 423                 rc = PTR_ERR(dentry);
 424                 CERROR("cannot create ROOT directory: rc = %d\n", rc);
 425                 GOTO(err_pop, rc);
 426         }
 427
 428         mds->mds_rootfid.id = dentry->d_inode->i_ino;
 429         mds->mds_rootfid.generation = dentry->d_inode->i_generation;
 430         mds->mds_rootfid.f_type = S_IFDIR;
 431
 432         dput(dentry);
 433
 434         dentry = lookup_one_len("__iopen__", current->fs->pwd,
 435                                 strlen("__iopen__"));
 436         if (IS_ERR(dentry) || !dentry->d_inode) {
 437                 rc = (IS_ERR(dentry)) ? PTR_ERR(dentry): -ENOENT;
 438                 CERROR("cannot open iopen FH directory: rc = %d\n", rc);
 439                 GOTO(err_pop, rc);
 440         }
 441         mds->mds_fid_de = dentry;
 442
 443         dentry = simple_mkdir(current->fs->pwd, "PENDING", 0777);
 444         if (IS_ERR(dentry)) {
 445                 rc = PTR_ERR(dentry);
 446                 CERROR("cannot create PENDING directory: rc = %d\n", rc);
 447                 GOTO(err_fid, rc);
 448         }
 449         mds->mds_pending_dir = dentry;
 450
 451         dentry = simple_mkdir(current->fs->pwd, "LOGS", 0777);
 452         if (IS_ERR(dentry)) {
 453                 rc = PTR_ERR(dentry);
 454                 CERROR("cannot create LOGS directory: rc = %d\n", rc);
 455                 GOTO(err_pending, rc);
 456         }
 457         mds->mds_logs_dir = dentry;
 458
 459         dentry = simple_mkdir(current->fs->pwd, "OBJECTS", 0777);
 460         if (IS_ERR(dentry)) {
 461                 rc = PTR_ERR(dentry);
 462                 CERROR("cannot create OBJECTS directory: rc = %d\n", rc);
 463                 GOTO(err_logs, rc);
 464         }
 465         mds->mds_objects_dir = dentry;
 466
 467         /* open and test the last rcvd file */
 468         file = filp_open(LAST_RCVD, O_RDWR | O_CREAT, 0644);
 469         if (IS_ERR(file)) {
 470                 rc = PTR_ERR(file);
 471                 CERROR("cannot open/create %s file: rc = %d\n", LAST_RCVD, rc);
 472                 GOTO(err_objects, rc = PTR_ERR(file));
 473         }
 474         mds->mds_rcvd_filp = file;
 475         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 476                 CERROR("%s is not a regular file!: mode = %o\n", LAST_RCVD,
 477                        file->f_dentry->d_inode->i_mode);
 478                 GOTO(err_last_rcvd, rc = -ENOENT);
 479         }
 480
 481         rc = mds_read_last_rcvd(obd, file);
 482         if (rc) {
 483                 CERROR("cannot read %s: rc = %d\n", LAST_RCVD, rc);
 484                 GOTO(err_last_rcvd, rc);
 485         }
 486
 487         /* open and test the lov objd file */
 488         file = filp_open(LOV_OBJID, O_RDWR | O_CREAT, 0644);
 489         if (IS_ERR(file)) {
 490                 rc = PTR_ERR(file);
 491                 CERROR("cannot open/create %s file: rc = %d\n", LOV_OBJID, rc);
 492                 GOTO(err_client, rc = PTR_ERR(file));
 493         }
 494         mds->mds_lov_objid_filp = file;
 495         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 496                 CERROR("%s is not a regular file!: mode = %o\n", LOV_OBJID,
 497                        file->f_dentry->d_inode->i_mode);
 498                 GOTO(err_lov_objid, rc = -ENOENT);
 499         }
 500 err_pop:
 501         pop_ctxt(&saved, &obd->obd_ctxt, NULL);
 502
 503         return rc;
 504
 505 err_lov_objid:
 506         if (mds->mds_lov_objid_filp && filp_close(mds->mds_lov_objid_filp, 0))
 507                 CERROR("can't close %s after error\n", LOV_OBJID);
 508 err_client:
 509         class_disconnect_exports(obd, 0);
 510 err_last_rcvd:
 511         if (mds->mds_rcvd_filp && filp_close(mds->mds_rcvd_filp, 0))
 512                 CERROR("can't close %s after error\n", LAST_RCVD);
 513 err_objects:
 514         dput(mds->mds_objects_dir);
 515 err_logs:
 516         dput(mds->mds_logs_dir);
 517 err_pending:
 518         dput(mds->mds_pending_dir);
 519 err_fid:
 520         dput(mds->mds_fid_de);
 521         goto err_pop;
 522 }
 523
 524
 525 int mds_fs_cleanup(struct obd_device *obd, int flags)
 526 {
 527         struct mds_obd *mds = &obd->u.mds;
 528         struct obd_run_ctxt saved;
 529         int rc = 0;
 530
 531         if (flags & OBD_OPT_FAILOVER)
 532                 CERROR("%s: shutting down for failover; client state will"
 533                        " be preserved.\n", obd->obd_name);
 534
 535         class_disconnect_exports(obd, flags); /* cleans up client info too */
 536         mds_server_free_data(mds);
 537
 538         push_ctxt(&saved, &obd->obd_ctxt, NULL);
 539         if (mds->mds_rcvd_filp) {
 540                 rc = filp_close(mds->mds_rcvd_filp, 0);
 541                 mds->mds_rcvd_filp = NULL;
 542                 if (rc)
 543                         CERROR("%s file won't close, rc=%d\n", LAST_RCVD, rc);
 544         }
 545         if (mds->mds_lov_objid_filp) {
 546                 rc = filp_close(mds->mds_lov_objid_filp, 0);
 547                 mds->mds_lov_objid_filp = NULL;
 548                 if (rc)
 549                         CERROR("%s file won't close, rc=%d\n", LOV_OBJID, rc);
 550         }
 551         if (mds->mds_objects_dir != NULL) {
 552                 l_dput(mds->mds_objects_dir);
 553                 mds->mds_objects_dir = NULL;
 554         }
 555         if (mds->mds_logs_dir) {
 556                 l_dput(mds->mds_logs_dir);
 557                 mds->mds_logs_dir = NULL;
 558         }
 559         if (mds->mds_pending_dir) {
 560                 l_dput(mds->mds_pending_dir);
 561                 mds->mds_pending_dir = NULL;
 562         }
 563         pop_ctxt(&saved, &obd->obd_ctxt, NULL);
 564         shrink_dcache_parent(mds->mds_fid_de);
 565         dput(mds->mds_fid_de);
 566
 567         return rc;
 568 }
 569
 570 /* Creates an object with the same name as its fid.  Because this is not at all
 571  * performance sensitive, it is accomplished by creating a file, checking the
 572  * fid, and renaming it. */
 573 int mds_obd_create(struct obd_export *exp, struct obdo *oa,
 574                       struct lov_stripe_md **ea, struct obd_trans_info *oti)
 575 {
 576         struct mds_obd *mds = &exp->exp_obd->u.mds;
 577         struct inode *parent_inode = mds->mds_objects_dir->d_inode;
 578         unsigned int tmpname = ll_insecure_random_int();
 579         struct file *filp;
 580         struct dentry *new_child;
 581         struct obd_run_ctxt saved;
 582         char fidname[LL_FID_NAMELEN];
 583         void *handle;
 584         int rc = 0, err, namelen;
 585         ENTRY;
 586
 587         push_ctxt(&saved, &exp->exp_obd->obd_ctxt, NULL);
 588
 589         sprintf(fidname, "OBJECTS/%u", tmpname);
 590         filp = filp_open(fidname, O_CREAT | O_EXCL, 0644);
 591         if (IS_ERR(filp)) {
 592                 rc = PTR_ERR(filp);
 593                 if (rc == -EEXIST) {
 594                         CERROR("impossible object name collision %u\n",
 595                                tmpname);
 596                         LBUG();
 597                 }
 598                 CERROR("error creating tmp object %u: rc %d\n", tmpname, rc);
 599                 GOTO(out_pop, rc);
 600         }
 601
 602         LASSERT(mds->mds_objects_dir == filp->f_dentry->d_parent);
 603
 604         oa->o_id = filp->f_dentry->d_inode->i_ino;
 605         oa->o_generation = filp->f_dentry->d_inode->i_generation;
 606         namelen = ll_fid2str(fidname, oa->o_id, oa->o_generation);
 607
 608         down(&parent_inode->i_sem);
 609         new_child = lookup_one_len(fidname, mds->mds_objects_dir, namelen);
 610
 611         if (IS_ERR(new_child)) {
 612                 CERROR("getting neg dentry for obj rename: %d\n", rc);
 613                 GOTO(out_close, rc = PTR_ERR(new_child));
 614         }
 615         if (new_child->d_inode != NULL) {
 616                 CERROR("impossible non-negative obj dentry " LPU64":%u!\n",
 617                        oa->o_id, oa->o_generation);
 618                 LBUG();
 619         }
 620
 621         handle = fsfilt_start(exp->exp_obd, mds->mds_objects_dir->d_inode,
 622                               FSFILT_OP_RENAME, NULL);
 623         if (IS_ERR(handle))
 624                 GOTO(out_dput, rc = PTR_ERR(handle));
 625
 626         lock_kernel();
 627         rc = vfs_rename(mds->mds_objects_dir->d_inode, filp->f_dentry,
 628                         mds->mds_objects_dir->d_inode, new_child);
 629         unlock_kernel();
 630         if (rc)
 631                 CERROR("error renaming new object "LPU64":%u: rc %d\n",
 632                        oa->o_id, oa->o_generation, rc);
 633
 634         err = fsfilt_commit(exp->exp_obd, mds->mds_objects_dir->d_inode,
 635                             handle, 0);
 636         if (!err)
 637                 oa->o_valid |= OBD_MD_FLID | OBD_MD_FLGENER;
 638         else if (!rc)
 639                 rc = err;
 640 out_dput:
 641         dput(new_child);
 642 out_close:
 643         up(&parent_inode->i_sem);
 644         err = filp_close(filp, 0);
 645         if (err) {
 646                 CERROR("closing tmpfile %u: rc %d\n", tmpname, rc);
 647                 if (!rc)
 648                         rc = err;
 649         }
 650 out_pop:
 651         pop_ctxt(&saved, &exp->exp_obd->obd_ctxt, NULL);
 652         RETURN(rc);
 653 }
 654
 655 int mds_obd_destroy(struct obd_export *exp, struct obdo *oa,
 656                     struct lov_stripe_md *ea, struct obd_trans_info *oti)
 657 {
 658         struct mds_obd *mds = &exp->exp_obd->u.mds;
 659         struct inode *parent_inode = mds->mds_objects_dir->d_inode;
 660         struct obd_device *obd = exp->exp_obd;
 661         struct obd_run_ctxt saved;
 662         char fidname[LL_FID_NAMELEN];
 663         struct dentry *de;
 664         void *handle;
 665         int err, namelen, rc = 0;
 666         ENTRY;
 667
 668         push_ctxt(&saved, &obd->obd_ctxt, NULL);
 669
 670         namelen = ll_fid2str(fidname, oa->o_id, oa->o_generation);
 671
 672         down(&parent_inode->i_sem);
 673         de = lookup_one_len(fidname, mds->mds_objects_dir, namelen);
 674         if (de == NULL || de->d_inode == NULL) {
 675                 CERROR("destroying non-existent object "LPU64"\n", oa->o_id);
 676                 GOTO(out, rc = IS_ERR(de) ? PTR_ERR(de) : -ENOENT);
 677         }
 678
 679         handle = fsfilt_start(obd, mds->mds_objects_dir->d_inode,
 680                               FSFILT_OP_UNLINK_LOG, oti);
 681         if (IS_ERR(handle)) {
 682                 GOTO(out_dput, rc = PTR_ERR(handle));
 683         }
 684
 685         rc = vfs_unlink(mds->mds_objects_dir->d_inode, de);
 686         if (rc)
 687                 CERROR("error destroying object "LPU64":%u: rc %d\n",
 688                        oa->o_id, oa->o_generation, rc);
 689
 690         err = fsfilt_commit(obd, mds->mds_objects_dir->d_inode, handle, 0);
 691         if (err && !rc)
 692                 rc = err;
 693 out_dput:
 694         l_dput(de);
 695 out:
 696         up(&parent_inode->i_sem);
 697         pop_ctxt(&saved, &obd->obd_ctxt, NULL);
 698         RETURN(rc);
 699 }