lustre/lvfs/llog_lvfs.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  *  Copyright (C) 2001-2003 Cluster File Systems, Inc.
   5  *   Author: Andreas Dilger <adilger@clusterfs.com>
   6  *
   7  *   This file is part of Lustre, http://www.lustre.org.
   8  *
   9  *   Lustre is free software; you can redistribute it and/or
  10  *   modify it under the terms of version 2 of the GNU General Public
  11  *   License as published by the Free Software Foundation.
  12  *
  13  *   Lustre is distributed in the hope that it will be useful,
  14  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16  *   GNU General Public License for more details.
  17  *
  18  *   You should have received a copy of the GNU General Public License
  19  *   along with Lustre; if not, write to the Free Software
  20  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  21  *
  22  * OST<->MDS recovery logging infrastructure.
  23  *
  24  * Invariants in implementation:
  25  * - we do not share logs among different OST<->MDS connections, so that
  26  *   if an OST or MDS fails it need only look at log(s) relevant to itself
  27  */
  28
  29 #define DEBUG_SUBSYSTEM S_LOG
  30
  31 #ifndef EXPORT_SYMTAB
  32 #define EXPORT_SYMTAB
  33 #endif
  34
  35 #ifdef __KERNEL__
  36 #include <linux/fs.h>
  37 #else
  38 #include <liblustre.h>
  39 #endif
  40
  41 #include <linux/lvfs.h>
  42 #include <linux/lustre_fsfilt.h>
  43 #include <linux/lustre_log.h>
  44
  45 #ifdef __KERNEL__
  46
  47 static int llog_lvfs_pad(struct llog_ctxt *ctxt, struct l_file *file,
  48                          int len, int index)
  49 {
  50         struct llog_rec_hdr rec;
  51         struct llog_rec_tail tail;
  52         int rc;
  53         ENTRY;
  54
  55         LASSERT(len >= LLOG_MIN_REC_SIZE && (len & 0x7) == 0);
  56
  57         tail.lrt_len = rec.lrh_len = cpu_to_le32(len);
  58         tail.lrt_index = rec.lrh_index = cpu_to_le32(index);
  59         rec.lrh_type = 0;
  60
  61         rc = llog_fsfilt_write_record(ctxt, file, &rec, sizeof(rec),
  62                                       &file->f_pos, 0);
  63         if (rc) {
  64                 CERROR("error writing padding record: rc %d\n", rc);
  65                 goto out;
  66         }
  67
  68         file->f_pos += len - sizeof(rec) - sizeof(tail);
  69         rc = llog_fsfilt_write_record(ctxt, file, &tail, sizeof(tail),
  70                                       &file->f_pos, 0);
  71         if (rc) {
  72                 CERROR("error writing padding record: rc %d\n", rc);
  73                 goto out;
  74         }
  75
  76  out:
  77         RETURN(rc);
  78 }
  79
  80 static int llog_lvfs_write_blob(struct llog_ctxt *ctxt, struct l_file *file,
  81                                 struct llog_rec_hdr *rec, void *buf, loff_t off)
  82 {
  83         int rc;
  84         struct llog_rec_tail end;
  85         loff_t saved_off = file->f_pos;
  86         int buflen = le32_to_cpu(rec->lrh_len);
  87
  88         ENTRY;
  89         file->f_pos = off;
  90
  91         if (!buf) {
  92                 rc = llog_fsfilt_write_record(ctxt, file, rec, buflen,
  93                                               &file->f_pos, 0);
  94                 if (rc) {
  95                         CERROR("error writing log record: rc %d\n", rc);
  96                         goto out;
  97                 }
  98                 GOTO(out, rc = 0);
  99         }
 100
 101         /* the buf case */
 102         rec->lrh_len = cpu_to_le32(sizeof(*rec) + buflen + sizeof(end));
 103         rc = llog_fsfilt_write_record(ctxt, file, rec, sizeof(*rec),
 104                                       &file->f_pos, 0);
 105         if (rc) {
 106                 CERROR("error writing log hdr: rc %d\n", rc);
 107                 goto out;
 108         }
 109
 110         rc = llog_fsfilt_write_record(ctxt, file, buf, buflen,
 111                                       &file->f_pos, 0);
 112         if (rc) {
 113                 CERROR("error writing log buffer: rc %d\n", rc);
 114                 goto out;
 115         }
 116
 117         end.lrt_len = rec->lrh_len;
 118         end.lrt_index = rec->lrh_index;
 119         rc = llog_fsfilt_write_record(ctxt, file, &end, sizeof(end),
 120                                       &file->f_pos, 0);
 121         if (rc) {
 122                 CERROR("error writing log tail: rc %d\n", rc);
 123                 goto out;
 124         }
 125
 126         rc = 0;
 127  out:
 128         if (saved_off > file->f_pos)
 129                 file->f_pos = saved_off;
 130         LASSERT(rc <= 0);
 131         RETURN(rc);
 132 }
 133
 134 static int llog_lvfs_read_blob(struct llog_ctxt *ctxt, struct l_file *file,
 135                                void *buf, int size, loff_t off)
 136 {
 137         loff_t offset = off;
 138         int rc;
 139         ENTRY;
 140
 141         rc = llog_fsfilt_read_record(ctxt, file, buf, size, &offset);
 142         if (rc) {
 143                 CERROR("error reading log record: rc %d\n", rc);
 144                 RETURN(rc);
 145         }
 146         RETURN(0);
 147 }
 148
 149 static int llog_lvfs_read_header(struct llog_handle *handle)
 150 {
 151         struct llog_ctxt *ctxt = handle->lgh_ctxt;
 152         int rc;
 153         ENTRY;
 154
 155         LASSERT(sizeof(*handle->lgh_hdr) == LLOG_CHUNK_SIZE);
 156         LASSERT(ctxt != NULL);
 157
 158         if (handle->lgh_file->f_dentry->d_inode->i_size == 0) {
 159                 CDEBUG(D_HA, "not reading header from 0-byte log\n");
 160                 RETURN(LLOG_EEMPTY);
 161         }
 162
 163         rc = llog_lvfs_read_blob(ctxt, handle->lgh_file, handle->lgh_hdr,
 164                                  LLOG_CHUNK_SIZE, 0);
 165         if (rc)
 166                 CERROR("error reading log header\n");
 167
 168         handle->lgh_last_idx = le32_to_cpu(handle->lgh_hdr->llh_tail.lrt_index);
 169         handle->lgh_file->f_pos = handle->lgh_file->f_dentry->d_inode->i_size;
 170
 171         RETURN(rc);
 172 }
 173
 174 /* returns negative in on error; 0 if success && reccookie == 0; 1 otherwise */
 175 /* appends if idx == -1, otherwise overwrites record idx. */
 176 static int llog_lvfs_write_rec(struct llog_handle *loghandle,
 177                                struct llog_rec_hdr *rec,
 178                                struct llog_cookie *reccookie,
 179                                int cookiecount,
 180                                void *buf, int idx)
 181 {
 182         struct llog_log_hdr *llh;
 183         int reclen = le32_to_cpu(rec->lrh_len), index, rc;
 184         struct llog_rec_tail *lrt;
 185         struct llog_ctxt *ctxt = loghandle->lgh_ctxt;
 186         struct file *file;
 187         loff_t offset;
 188         size_t left;
 189         ENTRY;
 190
 191         llh = loghandle->lgh_hdr;
 192         file = loghandle->lgh_file;
 193
 194         /* record length should not bigger than LLOG_CHUNK_SIZE */
 195         if (buf)
 196                 rc = (reclen > LLOG_CHUNK_SIZE - sizeof(struct llog_rec_hdr)
 197                       - sizeof(struct llog_rec_tail)) ? -E2BIG : 0;
 198         else
 199                 rc = (reclen > LLOG_CHUNK_SIZE) ? -E2BIG : 0;
 200         if (rc)
 201                 RETURN(rc);
 202
 203         if (idx != -1) {
 204                 loff_t saved_offset;
 205
 206                 /* no header: only allowed to insert record 1 */
 207                 if (idx > 1 && !file->f_dentry->d_inode->i_size) {
 208                         CERROR("idx != -1 in empty log\n");
 209                         LBUG();
 210                 }
 211
 212                 if (idx && llh->llh_size && llh->llh_size != reclen)
 213                         RETURN(-EINVAL);
 214
 215                 rc = llog_lvfs_write_blob(ctxt, file, &llh->llh_hdr, NULL, 0);
 216                 /* we are done if we only write the header or on error */
 217                 if (rc || idx == 0)
 218                         RETURN(rc);
 219
 220                 saved_offset = sizeof(*llh) + (idx-1)*le32_to_cpu(rec->lrh_len);
 221                 rc = llog_lvfs_write_blob(ctxt, file, rec, buf, saved_offset);
 222                 if (rc == 0 && reccookie) {
 223                         reccookie->lgc_lgl = loghandle->lgh_id;
 224                         reccookie->lgc_index = idx;
 225                         rc = 1;
 226                 }
 227                 RETURN(rc);
 228         }
 229
 230         /* Make sure that records don't cross a chunk boundary, so we can
 231          * process them page-at-a-time if needed.  If it will cross a chunk
 232          * boundary, write in a fake (but referenced) entry to pad the chunk.
 233          *
 234          * We know that llog_current_log() will return a loghandle that is
 235          * big enough to hold reclen, so all we care about is padding here.
 236          */
 237         left = LLOG_CHUNK_SIZE - (file->f_pos & (LLOG_CHUNK_SIZE - 1));
 238         if (buf)
 239                 reclen = sizeof(*rec) + le32_to_cpu(rec->lrh_len) +
 240                          sizeof(struct llog_rec_tail);
 241
 242         /* NOTE: padding is a record, but no bit is set */
 243         if (left != 0 && left != reclen &&
 244             left < (reclen + LLOG_MIN_REC_SIZE)) {
 245                 loghandle->lgh_last_idx++;
 246                 rc = llog_lvfs_pad(ctxt, file, left, loghandle->lgh_last_idx);
 247                 if (rc)
 248                         RETURN(rc);
 249         }
 250
 251         loghandle->lgh_last_idx++;
 252         index = loghandle->lgh_last_idx;
 253         rec->lrh_index = cpu_to_le32(index);
 254         if (buf == NULL) {
 255                 lrt = (void *)rec + le32_to_cpu(rec->lrh_len) - sizeof(*lrt);
 256                 lrt->lrt_len = rec->lrh_len;
 257                 lrt->lrt_index = rec->lrh_index;
 258         }
 259         if (ext2_set_bit(index, llh->llh_bitmap)) {
 260                 CERROR("argh, index %u already set in log bitmap?\n", index);
 261                 LBUG(); /* should never happen */
 262         }
 263         llh->llh_count = cpu_to_le32(le32_to_cpu(llh->llh_count) + 1);
 264         llh->llh_tail.lrt_index = cpu_to_le32(index);
 265
 266         offset = 0;
 267         rc = llog_lvfs_write_blob(ctxt, file, &llh->llh_hdr, NULL, 0);
 268         if (rc)
 269                 RETURN(rc);
 270
 271         CDEBUG(D_HA, "adding record "LPX64": idx: %u, %u bytes off: %lld\n",
 272                loghandle->lgh_id.lgl_oid, index, le32_to_cpu(rec->lrh_len),
 273                file->f_pos);
 274
 275         rc = llog_lvfs_write_blob(ctxt, file, rec, buf, file->f_pos);
 276         if (rc)
 277                 RETURN(rc);
 278
 279         if (rc == 0 && reccookie) {
 280                 if (llog_cookie_get_flags(reccookie) & LLOG_COOKIE_REPLAY) {
 281                         LASSERT(EQ_LOGID(reccookie->lgc_lgl, loghandle->lgh_id));
 282                         LASSERT(reccookie->lgc_index == index);
 283                 } else {
 284                         reccookie->lgc_lgl = loghandle->lgh_id;
 285                         reccookie->lgc_index = index;
 286                         llog_cookie_add_flags(reccookie, LLOG_COOKIE_REPLAY);
 287                 }
 288
 289                 if (le32_to_cpu(rec->lrh_type) == MDS_UNLINK_REC)
 290                         reccookie->lgc_subsys = LLOG_UNLINK_ORIG_CTXT;
 291                 else if (le32_to_cpu(rec->lrh_type) == OST_SZ_REC)
 292                         reccookie->lgc_subsys = LLOG_SIZE_ORIG_CTXT;
 293                 else if (le32_to_cpu(rec->lrh_type) == OST_RAID1_REC)
 294                         reccookie->lgc_subsys = LLOG_RD1_ORIG_CTXT;
 295                 else
 296                         reccookie->lgc_subsys = -1;
 297                 rc = 1;
 298         }
 299         if (rc == 0 && (le32_to_cpu(rec->lrh_type) == LLOG_GEN_REC ||
 300             le32_to_cpu(rec->lrh_type) == SMFS_UPDATE_REC))
 301                 rc = 1;
 302
 303         RETURN(rc);
 304 }
 305
 306 /* We can skip reading at least as many log blocks as the number of
 307 * minimum sized log records we are skipping.  If it turns out
 308 * that we are not far enough along the log (because the
 309 * actual records are larger than minimum size) we just skip
 310 * some more records. */
 311
 312 static void llog_skip_over(__u64 *off, int curr, int goal)
 313 {
 314         if (goal <= curr)
 315                 return;
 316         *off = (*off + (goal-curr-1) * LLOG_MIN_REC_SIZE) &
 317                 ~(LLOG_CHUNK_SIZE - 1);
 318 }
 319
 320 /* sets:
 321  *  - curr_offset to the furthest point read in the log file
 322  *  - curr_idx to the log index preceeding curr_offset
 323  * returns -EIO/-EINVAL on error
 324  */
 325 static int llog_lvfs_next_block(struct llog_handle *loghandle, int *curr_idx,
 326                                 int next_idx, __u64 *curr_offset, void *buf,
 327                                 int len)
 328 {
 329         struct llog_ctxt *ctxt = loghandle->lgh_ctxt;
 330         ENTRY;
 331
 332         if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
 333                 RETURN(-EINVAL);
 334
 335         CDEBUG(D_OTHER, "looking for log index %u (cur idx %u off "LPU64")\n",
 336                next_idx, *curr_idx, *curr_offset);
 337
 338         while (*curr_offset < loghandle->lgh_file->f_dentry->d_inode->i_size) {
 339                 struct llog_rec_hdr *rec;
 340                 struct llog_rec_tail *tail;
 341                 loff_t ppos;
 342                 int nbytes, rc;
 343
 344                 llog_skip_over(curr_offset, *curr_idx, next_idx);
 345
 346                 ppos = *curr_offset;
 347                 rc = llog_fsfilt_read_record(ctxt, loghandle->lgh_file,
 348                                              buf, len, &ppos);
 349
 350                 if (rc) {
 351                         CERROR("Cant read llog block at log id "LPU64
 352                                "/%u offset "LPU64"\n",
 353                                loghandle->lgh_id.lgl_oid,
 354                                loghandle->lgh_id.lgl_ogen,
 355                                *curr_offset);
 356                         RETURN(rc);
 357                 }
 358
 359                 nbytes = ppos - *curr_offset;
 360                 *curr_offset = ppos;
 361
 362                 if (nbytes == 0) /* end of file, nothing to do */
 363                         RETURN(0);
 364
 365                 if (nbytes < sizeof(*tail)) {
 366                         CERROR("Invalid llog block at log id "LPU64"/%u offset "
 367                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 368                                loghandle->lgh_id.lgl_ogen, *curr_offset);
 369                         RETURN(-EINVAL);
 370                 }
 371
 372                 tail = buf + nbytes - sizeof(struct llog_rec_tail);
 373                 *curr_idx = le32_to_cpu(tail->lrt_index);
 374
 375                 /* this shouldn't happen */
 376                 if (tail->lrt_index == 0) {
 377                         CERROR("Invalid llog tail at log id "LPU64"/%u offset "
 378                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 379                                loghandle->lgh_id.lgl_ogen, *curr_offset);
 380                         RETURN(-EINVAL);
 381                 }
 382                 if (le32_to_cpu(tail->lrt_index) < next_idx)
 383                         continue;
 384
 385                 /* sanity check that the start of the new buffer is no farther
 386                  * than the record that we wanted.  This shouldn't happen. */
 387                 rec = buf;
 388                 if (le32_to_cpu(rec->lrh_index) > next_idx) {
 389                         CERROR("missed desired record? %u > %u\n",
 390                                le32_to_cpu(rec->lrh_index), next_idx);
 391                         RETURN(-ENOENT);
 392                 }
 393                 RETURN(0);
 394         }
 395         RETURN(-EIO);
 396 }
 397
 398 static int llog_lvfs_prev_block(struct llog_handle *loghandle,
 399                                 int prev_idx, void *buf, int len)
 400 {
 401         struct llog_ctxt *ctxt = loghandle->lgh_ctxt;
 402         __u64 curr_offset;
 403         int rc;
 404         ENTRY;
 405
 406         if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
 407                 RETURN(-EINVAL);
 408
 409         CDEBUG(D_OTHER, "looking for log index %u n", prev_idx);
 410
 411         curr_offset = LLOG_CHUNK_SIZE;
 412         llog_skip_over(&curr_offset, 0, prev_idx);
 413
 414         while (curr_offset < loghandle->lgh_file->f_dentry->d_inode->i_size) {
 415                 struct llog_rec_hdr *rec;
 416                 struct llog_rec_tail *tail;
 417                 loff_t ppos;
 418
 419                 ppos = curr_offset;
 420                 rc = llog_fsfilt_read_record(ctxt, loghandle->lgh_file,
 421                                              buf, len, &ppos);
 422
 423                 if (rc) {
 424                         CERROR("Cant read llog block at log id "LPU64
 425                                "/%u offset "LPU64"\n",
 426                                loghandle->lgh_id.lgl_oid,
 427                                loghandle->lgh_id.lgl_ogen,
 428                                curr_offset);
 429                         RETURN(rc);
 430                 }
 431
 432                 /* put number of bytes read into rc to make code simpler */
 433                 rc = ppos - curr_offset;
 434                 curr_offset = ppos;
 435
 436                 if (rc == 0) /* end of file, nothing to do */
 437                         RETURN(0);
 438
 439                 if (rc < sizeof(*tail)) {
 440                         CERROR("Invalid llog block at log id "LPU64"/%u offset "
 441                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 442                                loghandle->lgh_id.lgl_ogen, curr_offset);
 443                         RETURN(-EINVAL);
 444                 }
 445
 446                 tail = buf + rc - sizeof(struct llog_rec_tail);
 447
 448                 /* this shouldn't happen */
 449                 if (tail->lrt_index == 0) {
 450                         CERROR("Invalid llog tail at log id "LPU64"/%u offset "
 451                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 452                                loghandle->lgh_id.lgl_ogen, curr_offset);
 453                         RETURN(-EINVAL);
 454                 }
 455                 if (le32_to_cpu(tail->lrt_index) < prev_idx)
 456                         continue;
 457
 458                 /* sanity check that the start of the new buffer is no farther
 459                  * than the record that we wanted.  This shouldn't happen. */
 460                 rec = buf;
 461                 if (le32_to_cpu(rec->lrh_index) > prev_idx) {
 462                         CERROR("missed desired record? %u > %u\n",
 463                                le32_to_cpu(rec->lrh_index), prev_idx);
 464                         RETURN(-ENOENT);
 465                 }
 466                 RETURN(0);
 467         }
 468         RETURN(-EIO);
 469 }
 470
 471 static struct file *llog_filp_open(char *name, int flags, int mode)
 472 {
 473         char *logname;
 474         struct file *filp;
 475         int len;
 476
 477         OBD_ALLOC(logname, PATH_MAX);
 478         if (logname == NULL)
 479                 return ERR_PTR(-ENOMEM);
 480
 481         len = snprintf(logname, PATH_MAX, "LOGS/%s", name);
 482         if (len >= PATH_MAX - 1) {
 483                 filp = ERR_PTR(-ENAMETOOLONG);
 484         } else {
 485                 filp = l_filp_open(logname, flags, mode);
 486                 if (IS_ERR(filp)) {
 487                         CERROR("logfile %s(%s): %ld\n",
 488                                flags & O_CREAT ? "create" : "open", logname,
 489                                PTR_ERR(filp));
 490                 }
 491         }
 492
 493         OBD_FREE(logname, PATH_MAX);
 494         return filp;
 495 }
 496
 497 /* creates object for the case when we have no obd (smfs). */
 498 static struct file *
 499 llog_object_create_alone(struct llog_ctxt *ctxt, struct llog_logid *lgh_id)
 500 {
 501         struct file *filp;
 502         int rc = 0;
 503         ENTRY;
 504
 505         LASSERT(lgh_id != NULL);
 506         if (lgh_id->lgl_oid) {
 507                 struct dentry *dchild;
 508                 char fidname[LL_FID_NAMELEN];
 509                 int fidlen = 0;
 510
 511                 down(&ctxt->loc_objects_dir->d_inode->i_sem);
 512                 fidlen = ll_fid2str(fidname, lgh_id->lgl_oid, lgh_id->lgl_ogen);
 513                 dchild = lookup_one_len(fidname, ctxt->loc_objects_dir, fidlen);
 514                 if (IS_ERR(dchild)) {
 515                         up(&ctxt->loc_objects_dir->d_inode->i_sem);
 516                         RETURN((struct file *)dchild);
 517                 }
 518                 if (dchild->d_inode == NULL) {
 519                         struct dentry_params dp;
 520                         struct inode *inode;
 521
 522                         dchild->d_fsdata = (void *) &dp;
 523                         dp.p_ptr = NULL;
 524                         dp.p_inum = lgh_id->lgl_oid;
 525                         rc = ll_vfs_create(ctxt->loc_objects_dir->d_inode,
 526                                            dchild, S_IFREG, NULL);
 527                         if (dchild->d_fsdata == (void *)(unsigned long)lgh_id->lgl_oid)
 528                                 dchild->d_fsdata = NULL;
 529                         if (rc) {
 530                                 CDEBUG(D_INODE, "err during create: %d\n", rc);
 531                                 dput(dchild);
 532                                 up(&ctxt->loc_objects_dir->d_inode->i_sem);
 533                                 RETURN(ERR_PTR(rc));
 534                         }
 535                         inode = dchild->d_inode;
 536                         LASSERT(inode->i_ino == lgh_id->lgl_oid);
 537                         inode->i_generation = lgh_id->lgl_ogen;
 538                         CDEBUG(D_HA, "recreated ino %lu with gen %u\n",
 539                                inode->i_ino, inode->i_generation);
 540                         mark_inode_dirty(inode);
 541                 }
 542
 543                 mntget(ctxt->loc_lvfs_ctxt->pwdmnt);
 544                 filp = dentry_open(dchild, ctxt->loc_lvfs_ctxt->pwdmnt,
 545                                     O_RDWR | O_LARGEFILE);
 546                 if (IS_ERR(filp)) {
 547                         dput(dchild);
 548                         up(&ctxt->loc_objects_dir->d_inode->i_sem);
 549                         RETURN(filp);
 550                 }
 551                 if (!S_ISREG(filp->f_dentry->d_inode->i_mode)) {
 552                         CERROR("%s is not a regular file!: mode = %o\n", fidname,
 553                                filp->f_dentry->d_inode->i_mode);
 554                         filp_close(filp, 0);
 555                         up(&ctxt->loc_objects_dir->d_inode->i_sem);
 556                         RETURN(ERR_PTR(-ENOENT));
 557                 }
 558
 559                 up(&ctxt->loc_objects_dir->d_inode->i_sem);
 560                 RETURN(filp);
 561
 562         } else {
 563                 unsigned int tmpname = ll_insecure_random_int();
 564                 char fidname[LL_FID_NAMELEN];
 565                 struct dentry *new_child, *parent;
 566                 void *handle;
 567                 int err, namelen;
 568
 569                 sprintf(fidname, "OBJECTS/%u", tmpname);
 570                 filp = filp_open(fidname, O_CREAT | O_EXCL, 0644);
 571                 if (IS_ERR(filp)) {
 572                         rc = PTR_ERR(filp);
 573                         if (rc == -EEXIST) {
 574                                 CERROR("impossible object name collision %u\n",
 575                                         tmpname);
 576                                 LBUG();
 577                         }
 578                         CERROR("error creating tmp object %u: rc %d\n", tmpname, rc);
 579                         RETURN(filp);
 580                 }
 581
 582                 namelen = ll_fid2str(fidname, filp->f_dentry->d_inode->i_ino,
 583                                      filp->f_dentry->d_inode->i_generation);
 584                 parent = filp->f_dentry->d_parent;
 585                 down(&parent->d_inode->i_sem);
 586                 new_child = lookup_one_len(fidname, parent, namelen);
 587                 if (IS_ERR(new_child)) {
 588                         CERROR("getting neg dentry for obj rename: %d\n", rc);
 589                         GOTO(out_close, rc = PTR_ERR(new_child));
 590                 }
 591                 if (new_child->d_inode != NULL) {
 592                         CERROR("impossible non-negative obj dentry %lu:%u!\n",
 593                                 filp->f_dentry->d_inode->i_ino,
 594                                 filp->f_dentry->d_inode->i_generation);
 595                         LBUG();
 596                 }
 597
 598                 handle = llog_fsfilt_start(ctxt, parent->d_inode, FSFILT_OP_RENAME, NULL);
 599                 if (IS_ERR(handle))
 600                         GOTO(out_dput, rc = PTR_ERR(handle));
 601
 602                 lock_kernel();
 603                 rc = vfs_rename(parent->d_inode, filp->f_dentry,
 604                                 parent->d_inode, new_child);
 605                 unlock_kernel();
 606                 if (rc)
 607                         CERROR("error renaming new object %lu:%u: rc %d\n",
 608                                 filp->f_dentry->d_inode->i_ino,
 609                                 filp->f_dentry->d_inode->i_generation, rc);
 610
 611                 err = llog_fsfilt_commit(ctxt, parent->d_inode, handle, 0);
 612                 if (!rc)
 613                         rc = err;
 614
 615         out_dput:
 616                 dput(new_child);
 617         out_close:
 618                 up(&parent->d_inode->i_sem);
 619                 if (rc) {
 620                         filp_close(filp, 0);
 621                         filp = ERR_PTR(rc);
 622                 } else {
 623                         /* FIXME: is this group 1 is correct? */
 624                         lgh_id->lgl_ogr = 1;
 625                         lgh_id->lgl_oid = filp->f_dentry->d_inode->i_ino;
 626                         lgh_id->lgl_ogen = filp->f_dentry->d_inode->i_generation;
 627                 }
 628                 RETURN(filp);
 629         }
 630 }
 631
 632 /* creates object for generic case (obd exists) */
 633 static struct file *
 634 llog_object_create_generic(struct llog_ctxt *ctxt, struct llog_logid *lgh_id)
 635 {
 636         struct file *filp;
 637         struct dentry *dchild;
 638         struct obd_device *obd;
 639         struct obdo *oa = NULL;
 640         int open_flags = O_RDWR | O_LARGEFILE;
 641         int rc = 0;
 642         ENTRY;
 643
 644         obd = ctxt->loc_exp->exp_obd;
 645         LASSERT(obd != NULL);
 646
 647         if (lgh_id->lgl_oid) {
 648                 dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, lgh_id->lgl_oid,
 649                                              lgh_id->lgl_ogen, lgh_id->lgl_ogr);
 650                 if (IS_ERR(dchild) == -ENOENT) {
 651                         OBD_ALLOC(oa, sizeof(*oa));
 652                         if (!oa)
 653                                 RETURN(ERR_PTR(-ENOMEM));
 654
 655                         oa->o_id = lgh_id->lgl_oid;
 656                         oa->o_generation = lgh_id->lgl_ogen;
 657                         oa->o_gr = lgh_id->lgl_ogr;
 658                         oa->o_valid = OBD_MD_FLGENER | OBD_MD_FLGROUP;
 659                         rc = obd_create(ctxt->loc_exp, oa, NULL, NULL);
 660                         if (rc) {
 661                                 CDEBUG(D_INODE, "err during create: %d\n", rc);
 662                                 GOTO(out_free_oa, rc);
 663                         }
 664                         CDEBUG(D_HA, "re-create log object "LPX64":0x%x:"LPX64"\n",
 665                                lgh_id->lgl_oid, lgh_id->lgl_ogen, lgh_id->lgl_ogr);
 666
 667                         dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, lgh_id->lgl_oid,
 668                                                      lgh_id->lgl_ogen, lgh_id->lgl_ogr);
 669                 } else if (IS_ERR(dchild)) {
 670                         CERROR("error looking up logfile "LPX64":0x%x: rc %d\n",
 671                                lgh_id->lgl_oid, lgh_id->lgl_ogen, rc);
 672                         RETURN((struct file *)dchild);
 673                 }
 674
 675                 filp = l_dentry_open(&obd->obd_lvfs_ctxt, dchild, open_flags);
 676                 if (IS_ERR(filp)) {
 677                         l_dput(dchild);
 678                         rc = PTR_ERR(filp);
 679                         CERROR("error opening logfile "LPX64"0x%x: rc %d\n",
 680                                lgh_id->lgl_oid, lgh_id->lgl_ogen, rc);
 681                 }
 682                 GOTO(out_free_oa, rc);
 683         } else {
 684                 /* this is important to work here over obd_create() as it manages
 685                   groups and we need it. Yet another reason is that mds_obd_create()
 686                  is fully the same as old version of this function and this helps
 687                  us to avoid code duplicating and layering violating. */
 688                 OBD_ALLOC(oa, sizeof(*oa));
 689                 if (!oa)
 690                         RETURN(ERR_PTR(-ENOMEM));
 691
 692                 oa->o_gr = FILTER_GROUP_LLOG;
 693                 oa->o_valid = OBD_MD_FLGENER | OBD_MD_FLGROUP;
 694                 rc = obd_create(ctxt->loc_exp, oa, NULL, NULL);
 695                 if (rc)
 696                         GOTO(out_free_oa, rc);
 697
 698                 dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, oa->o_id,
 699                                              oa->o_generation, oa->o_gr);
 700                 if (IS_ERR(dchild))
 701                         GOTO(out_free_oa, rc = PTR_ERR(dchild));
 702
 703                 filp = l_dentry_open(&obd->obd_lvfs_ctxt, dchild,
 704                                      open_flags);
 705                 if (IS_ERR(filp)) {
 706                         l_dput(dchild);
 707                         GOTO(out_free_oa, rc = PTR_ERR(filp));
 708                 }
 709
 710                 /* group 1 is not longer valid, we use the group which is set
 711                 by obd_create()->mds_obd_create(). */
 712                 lgh_id->lgl_ogr = oa->o_gr;
 713                 lgh_id->lgl_oid = oa->o_id;
 714                 lgh_id->lgl_ogen = oa->o_generation;
 715         }
 716
 717 out_free_oa:
 718         if (rc)
 719                 filp = ERR_PTR(rc);
 720         if (oa)
 721                 OBD_FREE(oa, sizeof(*oa));
 722         RETURN(filp);
 723 }
 724
 725 static struct file *
 726 llog_object_create(struct llog_ctxt *ctxt, struct llog_logid *lgh_id)
 727 {
 728         if (ctxt->loc_alone)
 729                 return llog_object_create_alone(ctxt, lgh_id);
 730         else
 731                 return llog_object_create_generic(ctxt, lgh_id);
 732 }
 733
 734 static int llog_add_link_object(struct llog_ctxt *ctxt, struct llog_logid logid,
 735                                 struct dentry *dentry)
 736 {
 737         struct dentry *new_child;
 738         char fidname[LL_FID_NAMELEN];
 739         void *handle;
 740         int namelen, rc = 0, err;
 741         ENTRY;
 742
 743         namelen = ll_fid2str(fidname, logid.lgl_oid, logid.lgl_ogen);
 744         down(&ctxt->loc_objects_dir->d_inode->i_sem);
 745         new_child = lookup_one_len(fidname, ctxt->loc_objects_dir, namelen);
 746         if (IS_ERR(new_child)) {
 747                 CERROR("getting neg dentry for obj rename: %d\n", rc);
 748                 GOTO(out, rc = PTR_ERR(new_child));
 749         }
 750         if (new_child->d_inode == dentry->d_inode)
 751                 GOTO(out_dput, rc);
 752         if (new_child->d_inode != NULL) {
 753                 CERROR("impossible non-negative obj dentry "LPX64":%u!\n",
 754                        logid.lgl_oid, logid.lgl_ogen);
 755                 LBUG();
 756         }
 757         handle = llog_fsfilt_start(ctxt, ctxt->loc_objects_dir->d_inode,
 758                                    FSFILT_OP_LINK, NULL);
 759         if (IS_ERR(handle))
 760                 GOTO(out_dput, rc = PTR_ERR(handle));
 761
 762         lock_kernel();
 763         rc = vfs_link(dentry, ctxt->loc_objects_dir->d_inode, new_child);
 764         unlock_kernel();
 765         if (rc)
 766                 CERROR("error link new object "LPX64":%u: rc %d\n",
 767                        logid.lgl_oid, logid.lgl_ogen, rc);
 768         err = llog_fsfilt_commit(ctxt, ctxt->loc_objects_dir->d_inode, handle, 0);
 769 out_dput:
 770         l_dput(new_child);
 771 out:
 772         up(&ctxt->loc_objects_dir->d_inode->i_sem);
 773         RETURN(rc);
 774 }
 775
 776 static int llog_lvfs_open(struct llog_ctxt *ctxt, struct llog_handle **res,
 777                           struct llog_logid *logid, char *name, int flags)
 778 {
 779         struct llog_handle *handle;
 780         struct lvfs_run_ctxt saved;
 781         int rc = 0;
 782         int open_flags = O_RDWR | O_LARGEFILE;
 783         ENTRY;
 784
 785         if (flags & OBD_LLOG_FL_CREATE)
 786                 open_flags |= O_CREAT;
 787
 788         handle = llog_alloc_handle();
 789         if (handle == NULL)
 790                 RETURN(-ENOMEM);
 791         *res = handle;
 792
 793         LASSERT(ctxt);
 794         if (ctxt->loc_lvfs_ctxt)
 795                 push_ctxt(&saved, ctxt->loc_lvfs_ctxt, NULL);
 796
 797         if (logid != NULL) {
 798                 handle->lgh_file = llog_object_create(ctxt, logid);
 799                 if (IS_ERR(handle->lgh_file)) {
 800                         CERROR("cannot create/open llog object "LPX64":%x "
 801                                "error = %ld", logid->lgl_oid, logid->lgl_ogen,
 802                                PTR_ERR(handle->lgh_file));
 803                         GOTO(cleanup, rc = PTR_ERR(handle->lgh_file));
 804                 }
 805                 handle->lgh_id = *logid;
 806
 807         } else if (name) {
 808                 handle->lgh_file = llog_filp_open(name, open_flags, 0644);
 809                 if (IS_ERR(handle->lgh_file)) {
 810                         CERROR("cannot open %s file, error = %ld\n",
 811                                name, PTR_ERR(handle->lgh_file));
 812                         GOTO(cleanup, rc = PTR_ERR(handle->lgh_file));
 813                 }
 814                 LASSERT(handle->lgh_file->f_dentry->d_parent == ctxt->loc_logs_dir);
 815
 816                 handle->lgh_id.lgl_ogr = 1;
 817                 handle->lgh_id.lgl_oid = handle->lgh_file->f_dentry->d_inode->i_ino;
 818                 handle->lgh_id.lgl_ogen = handle->lgh_file->f_dentry->d_inode->i_generation;
 819                 rc = llog_add_link_object(ctxt, handle->lgh_id, handle->lgh_file->f_dentry);
 820                 if (rc)
 821                         GOTO(cleanup, rc);
 822
 823         } else {
 824                 handle->lgh_file = llog_object_create(ctxt, &handle->lgh_id);
 825                 if (IS_ERR(handle->lgh_file)) {
 826                         CERROR("cannot create llog object, error = %ld\n",
 827                                PTR_ERR(handle->lgh_file));
 828                         GOTO(cleanup, rc = PTR_ERR(handle->lgh_file));
 829                 }
 830         }
 831
 832         handle->lgh_ctxt = ctxt;
 833 finish:
 834         if (ctxt->loc_lvfs_ctxt)
 835                 pop_ctxt(&saved, ctxt->loc_lvfs_ctxt, NULL);
 836         RETURN(rc);
 837 cleanup:
 838         llog_free_handle(handle);
 839         goto finish;
 840 }
 841
 842 static int llog_lvfs_close(struct llog_handle *handle)
 843 {
 844         int rc;
 845         ENTRY;
 846
 847         rc = filp_close(handle->lgh_file, 0);
 848         if (rc)
 849                 CERROR("error closing log: rc %d\n", rc);
 850         RETURN(rc);
 851 }
 852
 853 static int llog_lvfs_destroy(struct llog_handle *loghandle)
 854 {
 855         struct llog_ctxt *ctxt = loghandle->lgh_ctxt;
 856         struct lvfs_run_ctxt saved;
 857         struct dentry *fdentry;
 858         struct inode *parent_inode;
 859         char fidname[LL_FID_NAMELEN];
 860         void *handle;
 861         int rc = -EINVAL, err, namelen;
 862         ENTRY;
 863
 864         if (ctxt->loc_lvfs_ctxt)
 865                 push_ctxt(&saved, ctxt->loc_lvfs_ctxt, NULL);
 866
 867         fdentry = loghandle->lgh_file->f_dentry;
 868         parent_inode = fdentry->d_parent->d_inode;
 869
 870         if (!strcmp(fdentry->d_parent->d_name.name, "LOGS")) {
 871                 LASSERT(parent_inode == ctxt->loc_logs_dir->d_inode);
 872
 873                 namelen = ll_fid2str(fidname, fdentry->d_inode->i_ino,
 874                                      fdentry->d_inode->i_generation);
 875                 dget(fdentry);
 876                 rc = llog_lvfs_close(loghandle);
 877                 if (rc) {
 878                         dput(fdentry);
 879                         GOTO(out, rc);
 880                 }
 881
 882                 handle = llog_fsfilt_start(ctxt, parent_inode,
 883                                            FSFILT_OP_UNLINK, NULL);
 884                 if (IS_ERR(handle)) {
 885                         dput(fdentry);
 886                         GOTO(out, rc = PTR_ERR(handle));
 887                 }
 888
 889                 down(&parent_inode->i_sem);
 890                 rc = vfs_unlink(parent_inode, fdentry);
 891                 up(&parent_inode->i_sem);
 892                 dput(fdentry);
 893
 894                 if (!rc) {
 895                         down(&ctxt->loc_objects_dir->d_inode->i_sem);
 896                         fdentry = lookup_one_len(fidname, ctxt->loc_objects_dir,
 897                                                  namelen);
 898                         if (fdentry == NULL || fdentry->d_inode == NULL) {
 899                                 CERROR("destroy non_existent object %s\n", fidname);
 900                                 GOTO(out_err, rc = IS_ERR(fdentry) ?
 901                                      PTR_ERR(fdentry) : -ENOENT);
 902                         }
 903                         rc = vfs_unlink(ctxt->loc_objects_dir->d_inode, fdentry);
 904                         l_dput(fdentry);
 905 out_err:
 906                         up(&ctxt->loc_objects_dir->d_inode->i_sem);
 907                 }
 908                 err = llog_fsfilt_commit(ctxt, parent_inode, handle, 0);
 909                 if (err && !rc)
 910                         err = rc;
 911
 912                 GOTO(out, rc);
 913         }
 914         if (ctxt->loc_alone) {
 915                 if (!strcmp(fdentry->d_parent->d_name.name, "OBJECTS")) {
 916                         LASSERT(parent_inode == ctxt->loc_objects_dir->d_inode);
 917
 918                         dget(fdentry);
 919                         rc = llog_lvfs_close(loghandle);
 920                         if (rc == 0) {
 921                                 down(&parent_inode->i_sem);
 922                                 rc = vfs_unlink(parent_inode, fdentry);
 923                                 up(&parent_inode->i_sem);
 924                         }
 925                         dput(fdentry);
 926                 }
 927         } else {
 928                 struct obdo *oa = NULL;
 929
 930                 OBD_ALLOC(oa, sizeof(*oa));
 931                 if (!oa)
 932                         GOTO(out, rc = -ENOMEM);
 933
 934                 oa->o_id = loghandle->lgh_id.lgl_oid;
 935                 oa->o_gr = loghandle->lgh_id.lgl_ogr;
 936                 oa->o_generation = loghandle->lgh_id.lgl_ogen;
 937                 oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP | OBD_MD_FLGENER;
 938
 939                 rc = llog_lvfs_close(loghandle);
 940                 if (rc)
 941                         GOTO(out_free_oa, rc);
 942
 943                 rc = obd_destroy(loghandle->lgh_ctxt->loc_exp, oa, NULL, NULL);
 944 out_free_oa:
 945                 OBD_FREE(oa, sizeof(*oa));
 946         }
 947 out:
 948         if (ctxt->loc_lvfs_ctxt)
 949                 pop_ctxt(&saved, ctxt->loc_lvfs_ctxt, NULL);
 950         RETURN(rc);
 951 }
 952
 953 /* reads the catalog list */
 954 int llog_get_cat_list(struct lvfs_run_ctxt *ctxt,
 955                       struct fsfilt_operations *fsops, char *name,
 956                       int count, struct llog_catid *idarray)
 957 {
 958         struct lvfs_run_ctxt saved;
 959         struct l_file *file;
 960         int size = sizeof(*idarray) * count;
 961         loff_t off = 0;
 962         int rc;
 963
 964         LASSERT(count);
 965
 966         if (ctxt)
 967                 push_ctxt(&saved, ctxt, NULL);
 968         file = l_filp_open(name, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
 969         if (!file || IS_ERR(file)) {
 970                 rc = PTR_ERR(file);
 971                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
 972                        name, rc);
 973                 GOTO(out, rc);
 974         }
 975
 976         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 977                 CERROR("%s is not a regular file!: mode = %o\n", name,
 978                        file->f_dentry->d_inode->i_mode);
 979                 GOTO(out, rc = -ENOENT);
 980         }
 981
 982         rc = fsops->fs_read_record(file, idarray, size, &off);
 983         if (rc) {
 984                 CDEBUG(D_INODE,"OBD filter: error reading %s: rc %d\n",
 985                        name, rc);
 986                 GOTO(out, rc);
 987         }
 988
 989  out:
 990         if (file && !IS_ERR(file))
 991                 rc = filp_close(file, 0);
 992         if (ctxt)
 993                 pop_ctxt(&saved, ctxt, NULL);
 994         RETURN(rc);
 995 }
 996 EXPORT_SYMBOL(llog_get_cat_list);
 997
 998 /* writes the cat list */
 999 int llog_put_cat_list(struct lvfs_run_ctxt *ctxt,
1000                       struct fsfilt_operations *fsops, char *name,
1001                       int count, struct llog_catid *idarray)
1002 {
1003         struct lvfs_run_ctxt saved;
1004         struct l_file *file;
1005         int size = sizeof(*idarray) * count;
1006         loff_t off = 0;
1007         int rc;
1008
1009         LASSERT(count);
1010
1011         if (ctxt)
1012                 push_ctxt(&saved, ctxt, NULL);
1013         file = filp_open(name, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
1014         if (!file || IS_ERR(file)) {
1015                 rc = PTR_ERR(file);
1016                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
1017                        name, rc);
1018                 GOTO(out, rc);
1019         }
1020
1021         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
1022                 CERROR("%s is not a regular file!: mode = %o\n", name,
1023                        file->f_dentry->d_inode->i_mode);
1024                 GOTO(out, rc = -ENOENT);
1025         }
1026
1027         rc = fsops->fs_write_record(file, idarray, size, &off, 1);
1028         if (rc) {
1029                 CDEBUG(D_INODE,"OBD filter: error reading %s: rc %d\n",
1030                        name, rc);
1031                 GOTO(out, rc);
1032         }
1033
1034  out:
1035         if (file && !IS_ERR(file))
1036                 rc = filp_close(file, 0);
1037         if (ctxt)
1038                 pop_ctxt(&saved, ctxt, NULL);
1039         RETURN(rc);
1040 }
1041 EXPORT_SYMBOL(llog_put_cat_list);
1042
1043 struct llog_operations llog_lvfs_ops = {
1044         lop_open:        llog_lvfs_open,
1045         lop_destroy:     llog_lvfs_destroy,
1046         lop_close:       llog_lvfs_close,
1047         lop_read_header: llog_lvfs_read_header,
1048         lop_write_rec:   llog_lvfs_write_rec,
1049         lop_next_block:  llog_lvfs_next_block,
1050         lop_prev_block:  llog_lvfs_prev_block,
1051 };
1052 EXPORT_SYMBOL(llog_lvfs_ops);
1053
1054 #else /* !__KERNEL__ */
1055
1056 static int llog_lvfs_read_header(struct llog_handle *handle)
1057 {
1058         LBUG();
1059         return 0;
1060 }
1061
1062 static int llog_lvfs_write_rec(struct llog_handle *loghandle,
1063                                struct llog_rec_hdr *rec,
1064                                struct llog_cookie *reccookie, int cookiecount,
1065                                void *buf, int idx)
1066 {
1067         LBUG();
1068         return 0;
1069 }
1070
1071 static int llog_lvfs_open(struct llog_ctxt *ctxt, struct llog_handle **res,
1072                           struct llog_logid *logid, char *name, int flags)
1073 {
1074         LBUG();
1075         return 0;
1076 }
1077
1078 static int llog_lvfs_close(struct llog_handle *handle)
1079 {
1080         LBUG();
1081         return 0;
1082 }
1083
1084 static int llog_lvfs_destroy(struct llog_handle *handle)
1085 {
1086         LBUG();
1087         return 0;
1088 }
1089
1090 int llog_get_cat_list(struct lvfs_run_ctxt *ctxt,
1091                       struct fsfilt_operations *fsops, char *name,
1092                       int count, struct llog_catid *idarray)
1093 {
1094         LBUG();
1095         return 0;
1096 }
1097
1098 int llog_put_cat_list(struct lvfs_run_ctxt *ctxt,
1099                       struct fsfilt_operations *fsops, char *name,
1100                       int count, struct llog_catid *idarray)
1101 {
1102         LBUG();
1103         return 0;
1104 }
1105
1106 int llog_lvfs_prev_block(struct llog_handle *loghandle,
1107                          int prev_idx, void *buf, int len)
1108 {
1109         LBUG();
1110         return 0;
1111 }
1112
1113 int llog_lvfs_next_block(struct llog_handle *loghandle, int *curr_idx,
1114                          int next_idx, __u64 *offset, void *buf, int len)
1115 {
1116         LBUG();
1117         return 0;
1118 }
1119
1120 struct llog_operations llog_lvfs_ops = {
1121         lop_open:        llog_lvfs_open,
1122         lop_destroy:     llog_lvfs_destroy,
1123         lop_close:       llog_lvfs_close,
1124         lop_read_header: llog_lvfs_read_header,
1125         lop_write_rec:   llog_lvfs_write_rec,
1126         lop_next_block:  llog_lvfs_next_block,
1127         lop_prev_block:  llog_lvfs_prev_block,
1128 };
1129 #endif