lustre/obdclass/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 the Lustre file system, http://www.lustre.org
   8  *   Lustre is a trademark of Cluster File Systems, Inc.
   9  *
  10  *   You may have signed or agreed to another license before downloading
  11  *   this software.  If so, you are bound by the terms and conditions
  12  *   of that agreement, and the following does not apply to you.  See the
  13  *   LICENSE file included with this distribution for more information.
  14  *
  15  *   If you did not agree to a different license, then this copy of Lustre
  16  *   is open source software; you can redistribute it and/or modify it
  17  *   under the terms of version 2 of the GNU General Public License as
  18  *   published by the Free Software Foundation.
  19  *
  20  *   In either case, Lustre is distributed in the hope that it will be
  21  *   useful, but WITHOUT ANY WARRANTY; without even the implied warranty
  22  *   of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  23  *   license text for more details.
  24  *
  25  * OST<->MDS recovery logging infrastructure.
  26  *
  27  * Invariants in implementation:
  28  * - we do not share logs among different OST<->MDS connections, so that
  29  *   if an OST or MDS fails it need only look at log(s) relevant to itself
  30  */
  31
  32 #define DEBUG_SUBSYSTEM S_LOG
  33
  34 #ifndef EXPORT_SYMTAB
  35 #define EXPORT_SYMTAB
  36 #endif
  37
  38 #ifndef __KERNEL__
  39 #include <liblustre.h>
  40 #endif
  41
  42 #include <obd.h>
  43 #include <obd_class.h>
  44 #include <lustre_log.h>
  45 #include <obd_ost.h>
  46 #include <libcfs/list.h>
  47 #include <lvfs.h>
  48 #include <lustre_fsfilt.h>
  49 #include <lustre_disk.h>
  50 #include "llog_internal.h"
  51
  52 #if defined(__KERNEL__) && defined(LLOG_LVFS)
  53
  54 static int llog_lvfs_pad(struct obd_device *obd, struct l_file *file,
  55                                 int len, int index)
  56 {
  57         struct llog_rec_hdr rec = { 0 };
  58         struct llog_rec_tail tail;
  59         int rc;
  60         ENTRY;
  61
  62         LASSERT(len >= LLOG_MIN_REC_SIZE && (len & 0x7) == 0);
  63
  64         tail.lrt_len = rec.lrh_len = len;
  65         tail.lrt_index = rec.lrh_index = index;
  66         rec.lrh_type = LLOG_PAD_MAGIC;
  67
  68         rc = fsfilt_write_record(obd, file, &rec, sizeof(rec), &file->f_pos, 0);
  69         if (rc) {
  70                 CERROR("error writing padding record: rc %d\n", rc);
  71                 goto out;
  72         }
  73
  74         file->f_pos += len - sizeof(rec) - sizeof(tail);
  75         rc = fsfilt_write_record(obd, file, &tail, sizeof(tail),&file->f_pos,0);
  76         if (rc) {
  77                 CERROR("error writing padding record: rc %d\n", rc);
  78                 goto out;
  79         }
  80
  81  out:
  82         RETURN(rc);
  83 }
  84
  85 static int llog_lvfs_write_blob(struct obd_device *obd, struct l_file *file,
  86                                 struct llog_rec_hdr *rec, void *buf, loff_t off)
  87 {
  88         int rc;
  89         struct llog_rec_tail end;
  90         loff_t saved_off = file->f_pos;
  91         int buflen = rec->lrh_len;
  92
  93         ENTRY;
  94
  95         file->f_pos = off;
  96
  97         if (buflen == 0)
  98                 CWARN("0-length record\n");
  99
 100         if (!buf) {
 101                 rc = fsfilt_write_record(obd, file, rec, buflen,&file->f_pos,0);
 102                 if (rc) {
 103                         CERROR("error writing log record: rc %d\n", rc);
 104                         goto out;
 105                 }
 106                 GOTO(out, rc = 0);
 107         }
 108
 109         /* the buf case */
 110         rec->lrh_len = sizeof(*rec) + buflen + sizeof(end);
 111         rc = fsfilt_write_record(obd, file, rec, sizeof(*rec), &file->f_pos, 0);
 112         if (rc) {
 113                 CERROR("error writing log hdr: rc %d\n", rc);
 114                 goto out;
 115         }
 116
 117         rc = fsfilt_write_record(obd, file, buf, buflen, &file->f_pos, 0);
 118         if (rc) {
 119                 CERROR("error writing log buffer: rc %d\n", rc);
 120                 goto out;
 121         }
 122
 123         end.lrt_len = rec->lrh_len;
 124         end.lrt_index = rec->lrh_index;
 125         rc = fsfilt_write_record(obd, file, &end, sizeof(end), &file->f_pos, 0);
 126         if (rc) {
 127                 CERROR("error writing log tail: rc %d\n", rc);
 128                 goto out;
 129         }
 130
 131         rc = 0;
 132  out:
 133         if (saved_off > file->f_pos)
 134                 file->f_pos = saved_off;
 135         LASSERT(rc <= 0);
 136         RETURN(rc);
 137 }
 138
 139 static int llog_lvfs_read_blob(struct obd_device *obd, struct l_file *file,
 140                                 void *buf, int size, loff_t off)
 141 {
 142         loff_t offset = off;
 143         int rc;
 144         ENTRY;
 145
 146         rc = fsfilt_read_record(obd, file, buf, size, &offset);
 147         if (rc) {
 148                 CERROR("error reading log record: rc %d\n", rc);
 149                 RETURN(rc);
 150         }
 151         RETURN(0);
 152 }
 153
 154 static int llog_lvfs_read_header(struct llog_handle *handle)
 155 {
 156         struct obd_device *obd;
 157         int rc;
 158         ENTRY;
 159
 160         LASSERT(sizeof(*handle->lgh_hdr) == LLOG_CHUNK_SIZE);
 161
 162         obd = handle->lgh_ctxt->loc_exp->exp_obd;
 163
 164         if (i_size_read(handle->lgh_file->f_dentry->d_inode) == 0) {
 165                 CDEBUG(D_HA, "not reading header from 0-byte log\n");
 166                 RETURN(LLOG_EEMPTY);
 167         }
 168
 169         rc = llog_lvfs_read_blob(obd, handle->lgh_file, handle->lgh_hdr,
 170                                  LLOG_CHUNK_SIZE, 0);
 171         if (rc) {
 172                 CERROR("error reading log header from %.*s\n",
 173                        handle->lgh_file->f_dentry->d_name.len,
 174                        handle->lgh_file->f_dentry->d_name.name);
 175         } else {
 176                 struct llog_rec_hdr *llh_hdr = &handle->lgh_hdr->llh_hdr;
 177
 178                 if (LLOG_REC_HDR_NEEDS_SWABBING(llh_hdr))
 179                         lustre_swab_llog_hdr(handle->lgh_hdr);
 180
 181                 if (llh_hdr->lrh_type != LLOG_HDR_MAGIC) {
 182                         CERROR("bad log %.*s header magic: %#x (expected %#x)\n",
 183                                handle->lgh_file->f_dentry->d_name.len,
 184                                handle->lgh_file->f_dentry->d_name.name,
 185                                llh_hdr->lrh_type, LLOG_HDR_MAGIC);
 186                         rc = -EIO;
 187                 } else if (llh_hdr->lrh_len != LLOG_CHUNK_SIZE) {
 188                         CERROR("incorrectly sized log %.*s header: %#x "
 189                                "(expected %#x)\n",
 190                                handle->lgh_file->f_dentry->d_name.len,
 191                                handle->lgh_file->f_dentry->d_name.name,
 192                                llh_hdr->lrh_len, LLOG_CHUNK_SIZE);
 193                         CERROR("you may need to re-run lconf --write_conf.\n");
 194                         rc = -EIO;
 195                 }
 196         }
 197
 198         handle->lgh_last_idx = handle->lgh_hdr->llh_tail.lrt_index;
 199         handle->lgh_file->f_pos = i_size_read(handle->lgh_file->f_dentry->d_inode);
 200
 201         RETURN(rc);
 202 }
 203
 204 /* returns negative in on error; 0 if success && reccookie == 0; 1 otherwise */
 205 /* appends if idx == -1, otherwise overwrites record idx. */
 206 static int llog_lvfs_write_rec(struct llog_handle *loghandle,
 207                                struct llog_rec_hdr *rec,
 208                                struct llog_cookie *reccookie, int cookiecount,
 209                                void *buf, int idx)
 210 {
 211         struct llog_log_hdr *llh;
 212         int reclen = rec->lrh_len, index, rc;
 213         struct llog_rec_tail *lrt;
 214         struct obd_device *obd;
 215         struct file *file;
 216         size_t left;
 217         ENTRY;
 218
 219         llh = loghandle->lgh_hdr;
 220         file = loghandle->lgh_file;
 221         obd = loghandle->lgh_ctxt->loc_exp->exp_obd;
 222
 223         /* record length should not bigger than LLOG_CHUNK_SIZE */
 224         if (buf)
 225                 rc = (reclen > LLOG_CHUNK_SIZE - sizeof(struct llog_rec_hdr) -
 226                       sizeof(struct llog_rec_tail)) ? -E2BIG : 0;
 227         else
 228                 rc = (reclen > LLOG_CHUNK_SIZE) ? -E2BIG : 0;
 229         if (rc)
 230                 RETURN(rc);
 231
 232         if (buf)
 233                 /* write_blob adds header and tail to lrh_len. */
 234                 reclen = sizeof(*rec) + rec->lrh_len +
 235                          sizeof(struct llog_rec_tail);
 236
 237         if (idx != -1) {
 238                 loff_t saved_offset;
 239
 240                 /* no header: only allowed to insert record 1 */
 241                 if (idx != 1 && !i_size_read(file->f_dentry->d_inode)) {
 242                         CERROR("idx != -1 in empty log\n");
 243                         LBUG();
 244                 }
 245
 246                 if (idx && llh->llh_size && llh->llh_size != rec->lrh_len)
 247                         RETURN(-EINVAL);
 248
 249                 if (!ext2_test_bit(idx, llh->llh_bitmap))
 250                         CERROR("Modify unset record %u\n", idx);
 251                 if (idx != rec->lrh_index)
 252                         CERROR("Index mismatch %d %u\n", idx, rec->lrh_index);
 253
 254                 rc = llog_lvfs_write_blob(obd, file, &llh->llh_hdr, NULL, 0);
 255                 /* we are done if we only write the header or on error */
 256                 if (rc || idx == 0)
 257                         RETURN(rc);
 258
 259                 /* Assumes constant lrh_len */
 260                 saved_offset = sizeof(*llh) + (idx - 1) * reclen;
 261
 262                 if (buf) {
 263                         struct llog_rec_hdr check;
 264
 265                         /* We assume that caller has set lgh_cur_* */
 266                         saved_offset = loghandle->lgh_cur_offset;
 267                         CDEBUG(D_OTHER,
 268                                "modify record "LPX64": idx:%d/%u/%d, len:%u "
 269                                "offset %llu\n",
 270                                loghandle->lgh_id.lgl_oid, idx, rec->lrh_index,
 271                                loghandle->lgh_cur_idx, rec->lrh_len,
 272                                (long long)(saved_offset - sizeof(*llh)));
 273                         if (rec->lrh_index != loghandle->lgh_cur_idx) {
 274                                 CERROR("modify idx mismatch %u/%d\n",
 275                                        idx, loghandle->lgh_cur_idx);
 276                                 RETURN(-EFAULT);
 277                         }
 278 #if 1  /* FIXME remove this safety check at some point */
 279                         /* Verify that the record we're modifying is the
 280                            right one. */
 281                         rc = llog_lvfs_read_blob(obd, file, &check,
 282                                                  sizeof(check), saved_offset);
 283                         if (check.lrh_index != idx || check.lrh_len != reclen) {
 284                                 CERROR("Bad modify idx %u/%u size %u/%u (%d)\n",
 285                                        idx, check.lrh_index, reclen,
 286                                        check.lrh_len, rc);
 287                                 RETURN(-EFAULT);
 288                         }
 289 #endif
 290                 }
 291
 292                 rc = llog_lvfs_write_blob(obd, file, rec, buf, saved_offset);
 293                 if (rc == 0 && reccookie) {
 294                         reccookie->lgc_lgl = loghandle->lgh_id;
 295                         reccookie->lgc_index = idx;
 296                         rc = 1;
 297                 }
 298                 RETURN(rc);
 299         }
 300
 301         /* Make sure that records don't cross a chunk boundary, so we can
 302          * process them page-at-a-time if needed.  If it will cross a chunk
 303          * boundary, write in a fake (but referenced) entry to pad the chunk.
 304          *
 305          * We know that llog_current_log() will return a loghandle that is
 306          * big enough to hold reclen, so all we care about is padding here.
 307          */
 308         left = LLOG_CHUNK_SIZE - (file->f_pos & (LLOG_CHUNK_SIZE - 1));
 309
 310         /* NOTE: padding is a record, but no bit is set */
 311         if (left != 0 && left != reclen &&
 312             left < (reclen + LLOG_MIN_REC_SIZE)) {
 313                 loghandle->lgh_last_idx++;
 314                 rc = llog_lvfs_pad(obd, file, left, loghandle->lgh_last_idx);
 315                 if (rc)
 316                         RETURN(rc);
 317                 /* if it's the last idx in log file, then return -ENOSPC */
 318                 if (loghandle->lgh_last_idx == LLOG_BITMAP_SIZE(llh) - 1)
 319                         RETURN(-ENOSPC);
 320         }
 321
 322         loghandle->lgh_last_idx++;
 323         index = loghandle->lgh_last_idx;
 324         LASSERT(index < LLOG_BITMAP_SIZE(llh));
 325         rec->lrh_index = index;
 326         if (buf == NULL) {
 327                 lrt = (struct llog_rec_tail *)
 328                         ((char *)rec + rec->lrh_len - sizeof(*lrt));
 329                 lrt->lrt_len = rec->lrh_len;
 330                 lrt->lrt_index = rec->lrh_index;
 331         }
 332         if (ext2_set_bit(index, llh->llh_bitmap)) {
 333                 CERROR("argh, index %u already set in log bitmap?\n", index);
 334                 LBUG(); /* should never happen */
 335         }
 336         llh->llh_count++;
 337         llh->llh_tail.lrt_index = index;
 338
 339         rc = llog_lvfs_write_blob(obd, file, &llh->llh_hdr, NULL, 0);
 340         if (rc)
 341                 RETURN(rc);
 342
 343         rc = llog_lvfs_write_blob(obd, file, rec, buf, file->f_pos);
 344         if (rc)
 345                 RETURN(rc);
 346
 347         CDEBUG(D_HA, "added record "LPX64": idx: %u, %u bytes\n",
 348                loghandle->lgh_id.lgl_oid, index, rec->lrh_len);
 349         if (rc == 0 && reccookie) {
 350                 reccookie->lgc_lgl = loghandle->lgh_id;
 351                 reccookie->lgc_index = index;
 352                 if ((rec->lrh_type == MDS_UNLINK_REC) ||
 353                                 (rec->lrh_type == MDS_SETATTR_REC))
 354                         reccookie->lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
 355                 else if (rec->lrh_type == OST_SZ_REC)
 356                         reccookie->lgc_subsys = LLOG_SIZE_ORIG_CTXT;
 357                 else if (rec->lrh_type == OST_RAID1_REC)
 358                         reccookie->lgc_subsys = LLOG_RD1_ORIG_CTXT;
 359                 else
 360                         reccookie->lgc_subsys = -1;
 361                 rc = 1;
 362         }
 363         if (rc == 0 && rec->lrh_type == LLOG_GEN_REC)
 364                 rc = 1;
 365
 366         RETURN(rc);
 367 }
 368
 369 /* We can skip reading at least as many log blocks as the number of
 370 * minimum sized log records we are skipping.  If it turns out
 371 * that we are not far enough along the log (because the
 372 * actual records are larger than minimum size) we just skip
 373 * some more records. */
 374
 375 static void llog_skip_over(__u64 *off, int curr, int goal)
 376 {
 377         if (goal <= curr)
 378                 return;
 379         *off = (*off + (goal-curr-1) * LLOG_MIN_REC_SIZE) &
 380                 ~(LLOG_CHUNK_SIZE - 1);
 381 }
 382
 383
 384 /* sets:
 385  *  - cur_offset to the furthest point read in the log file
 386  *  - cur_idx to the log index preceeding cur_offset
 387  * returns -EIO/-EINVAL on error
 388  */
 389 static int llog_lvfs_next_block(struct llog_handle *loghandle, int *cur_idx,
 390                                 int next_idx, __u64 *cur_offset, void *buf,
 391                                 int len)
 392 {
 393         int rc;
 394         ENTRY;
 395
 396         if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
 397                 RETURN(-EINVAL);
 398
 399         CDEBUG(D_OTHER, "looking for log index %u (cur idx %u off "LPU64")\n",
 400                next_idx, *cur_idx, *cur_offset);
 401
 402         while (*cur_offset < i_size_read(loghandle->lgh_file->f_dentry->d_inode)) {
 403                 struct llog_rec_hdr *rec;
 404                 struct llog_rec_tail *tail;
 405                 loff_t ppos;
 406
 407                 llog_skip_over(cur_offset, *cur_idx, next_idx);
 408
 409                 ppos = *cur_offset;
 410                 rc = fsfilt_read_record(loghandle->lgh_ctxt->loc_exp->exp_obd,
 411                                         loghandle->lgh_file, buf, len,
 412                                         &ppos);
 413                 if (rc) {
 414                         CERROR("Cant read llog block at log id "LPU64
 415                                "/%u offset "LPU64"\n",
 416                                loghandle->lgh_id.lgl_oid,
 417                                loghandle->lgh_id.lgl_ogen,
 418                                *cur_offset);
 419                         RETURN(rc);
 420                 }
 421
 422                 /* put number of bytes read into rc to make code simpler */
 423                 rc = ppos - *cur_offset;
 424                 *cur_offset = ppos;
 425
 426                 if (rc < len) {
 427                         /* signal the end of the valid buffer to llog_process */
 428                         memset(buf + rc, 0, len - rc);
 429                 }
 430
 431                 if (rc == 0) /* end of file, nothing to do */
 432                         RETURN(0);
 433
 434                 if (rc < sizeof(*tail)) {
 435                         CERROR("Invalid llog block at log id "LPU64"/%u offset "
 436                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 437                                loghandle->lgh_id.lgl_ogen, *cur_offset);
 438                         RETURN(-EINVAL);
 439                 }
 440
 441                 rec = buf;
 442                 tail = (struct llog_rec_tail *)((char *)buf + rc -
 443                                                 sizeof(struct llog_rec_tail));
 444
 445                 if (LLOG_REC_HDR_NEEDS_SWABBING(rec)) {
 446                         lustre_swab_llog_rec(rec, tail);
 447                 }
 448
 449                 *cur_idx = tail->lrt_index;
 450
 451                 /* this shouldn't happen */
 452                 if (tail->lrt_index == 0) {
 453                         CERROR("Invalid llog tail at log id "LPU64"/%u offset "
 454                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 455                                loghandle->lgh_id.lgl_ogen, *cur_offset);
 456                         RETURN(-EINVAL);
 457                 }
 458                 if (tail->lrt_index < next_idx)
 459                         continue;
 460
 461                 /* sanity check that the start of the new buffer is no farther
 462                  * than the record that we wanted.  This shouldn't happen. */
 463                 if (rec->lrh_index > next_idx) {
 464                         CERROR("missed desired record? %u > %u\n",
 465                                rec->lrh_index, next_idx);
 466                         RETURN(-ENOENT);
 467                 }
 468                 RETURN(0);
 469         }
 470         RETURN(-EIO);
 471 }
 472
 473 static int llog_lvfs_prev_block(struct llog_handle *loghandle,
 474                                 int prev_idx, void *buf, int len)
 475 {
 476         __u64 cur_offset;
 477         int rc;
 478         ENTRY;
 479
 480         if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
 481                 RETURN(-EINVAL);
 482
 483         CDEBUG(D_OTHER, "looking for log index %u n", prev_idx);
 484
 485         cur_offset = LLOG_CHUNK_SIZE;
 486         llog_skip_over(&cur_offset, 0, prev_idx);
 487
 488         while (cur_offset < i_size_read(loghandle->lgh_file->f_dentry->d_inode)) {
 489                 struct llog_rec_hdr *rec;
 490                 struct llog_rec_tail *tail;
 491                 loff_t ppos;
 492
 493                 ppos = cur_offset;
 494
 495                 rc = fsfilt_read_record(loghandle->lgh_ctxt->loc_exp->exp_obd,
 496                                         loghandle->lgh_file, buf, len,
 497                                         &ppos);
 498                 if (rc) {
 499                         CERROR("Cant read llog block at log id "LPU64
 500                                "/%u offset "LPU64"\n",
 501                                loghandle->lgh_id.lgl_oid,
 502                                loghandle->lgh_id.lgl_ogen,
 503                                cur_offset);
 504                         RETURN(rc);
 505                 }
 506
 507                 /* put number of bytes read into rc to make code simpler */
 508                 rc = ppos - cur_offset;
 509                 cur_offset = ppos;
 510
 511                 if (rc == 0) /* end of file, nothing to do */
 512                         RETURN(0);
 513
 514                 if (rc < sizeof(*tail)) {
 515                         CERROR("Invalid llog block at log id "LPU64"/%u offset "
 516                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 517                                loghandle->lgh_id.lgl_ogen, cur_offset);
 518                         RETURN(-EINVAL);
 519                 }
 520
 521                 tail = buf + rc - sizeof(struct llog_rec_tail);
 522
 523                 /* this shouldn't happen */
 524                 if (tail->lrt_index == 0) {
 525                         CERROR("Invalid llog tail at log id "LPU64"/%u offset "
 526                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 527                                loghandle->lgh_id.lgl_ogen, cur_offset);
 528                         RETURN(-EINVAL);
 529                 }
 530                 if (le32_to_cpu(tail->lrt_index) < prev_idx)
 531                         continue;
 532
 533                 /* sanity check that the start of the new buffer is no farther
 534                  * than the record that we wanted.  This shouldn't happen. */
 535                 rec = buf;
 536                 if (le32_to_cpu(rec->lrh_index) > prev_idx) {
 537                         CERROR("missed desired record? %u > %u\n",
 538                                le32_to_cpu(rec->lrh_index), prev_idx);
 539                         RETURN(-ENOENT);
 540                 }
 541                 RETURN(0);
 542         }
 543         RETURN(-EIO);
 544 }
 545
 546 static struct file *llog_filp_open(char *dir, char *name, int flags, int mode)
 547 {
 548         char *logname;
 549         struct file *filp;
 550         int len;
 551
 552         OBD_ALLOC(logname, PATH_MAX);
 553         if (logname == NULL)
 554                 return ERR_PTR(-ENOMEM);
 555
 556         len = snprintf(logname, PATH_MAX, "%s/%s", dir, name);
 557         if (len >= PATH_MAX - 1) {
 558                 filp = ERR_PTR(-ENAMETOOLONG);
 559         } else {
 560                 filp = l_filp_open(logname, flags, mode);
 561                 if (IS_ERR(filp))
 562                         CERROR("logfile creation %s: %ld\n", logname,
 563                                PTR_ERR(filp));
 564         }
 565         OBD_FREE(logname, PATH_MAX);
 566         return filp;
 567 }
 568
 569 /* This is a callback from the llog_* functions.
 570  * Assumes caller has already pushed us into the kernel context. */
 571 static int llog_lvfs_create(struct llog_ctxt *ctxt, struct llog_handle **res,
 572                             struct llog_logid *logid, char *name)
 573 {
 574         struct llog_handle *handle;
 575         struct obd_device *obd;
 576         struct l_dentry *dchild = NULL;
 577         struct obdo *oa = NULL;
 578         int rc = 0, cleanup_phase = 1;
 579         int open_flags = O_RDWR | O_CREAT | O_LARGEFILE;
 580         ENTRY;
 581
 582         handle = llog_alloc_handle();
 583         if (handle == NULL)
 584                 RETURN(-ENOMEM);
 585         *res = handle;
 586
 587         LASSERT(ctxt);
 588         LASSERT(ctxt->loc_exp);
 589         obd = ctxt->loc_exp->exp_obd;
 590
 591         if (logid != NULL) {
 592                 dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, logid->lgl_oid,
 593                                              logid->lgl_ogen, logid->lgl_ogr);
 594
 595                 if (IS_ERR(dchild)) {
 596                         rc = PTR_ERR(dchild);
 597                         CERROR("error looking up logfile "LPX64":0x%x: rc %d\n",
 598                                logid->lgl_oid, logid->lgl_ogen, rc);
 599                         GOTO(cleanup, rc);
 600                 }
 601
 602                 cleanup_phase = 2;
 603                 if (dchild->d_inode == NULL) {
 604                         rc = -ENOENT;
 605                         CERROR("nonexistent log file "LPX64":"LPX64": rc %d\n",
 606                                logid->lgl_oid, logid->lgl_ogr, rc);
 607                         GOTO(cleanup, rc);
 608                 }
 609
 610                 handle->lgh_file = l_dentry_open(&obd->obd_lvfs_ctxt, dchild,
 611                                                     O_RDWR | O_LARGEFILE);
 612                 if (IS_ERR(handle->lgh_file)) {
 613                         rc = PTR_ERR(handle->lgh_file);
 614                         CERROR("error opening logfile "LPX64"0x%x: rc %d\n",
 615                                logid->lgl_oid, logid->lgl_ogen, rc);
 616                         GOTO(cleanup, rc);
 617                 }
 618
 619                 /* assign the value of lgh_id for handle directly */
 620                 handle->lgh_id = *logid;
 621
 622         } else if (name) {
 623                 /* COMPAT_146 */
 624                 if (strcmp(obd->obd_type->typ_name, LUSTRE_MDS_NAME) == 0) {
 625                         handle->lgh_file = llog_filp_open(MDT_LOGS_DIR, name,
 626                                                           open_flags, 0644);
 627                 } else {
 628                         /* end COMPAT_146 */
 629                         handle->lgh_file = llog_filp_open(MOUNT_CONFIGS_DIR,
 630                                                           name, open_flags,
 631                                                           0644);
 632                 }
 633                 if (IS_ERR(handle->lgh_file))
 634                         GOTO(cleanup, rc = PTR_ERR(handle->lgh_file));
 635
 636                 handle->lgh_id.lgl_ogr = 1;
 637                 handle->lgh_id.lgl_oid =
 638                         handle->lgh_file->f_dentry->d_inode->i_ino;
 639                 handle->lgh_id.lgl_ogen =
 640                         handle->lgh_file->f_dentry->d_inode->i_generation;
 641         } else {
 642                 OBDO_ALLOC(oa);
 643                 if (oa == NULL)
 644                         GOTO(cleanup, rc = -ENOMEM);
 645
 646                 oa->o_gr = FILTER_GROUP_LLOG;
 647                 oa->o_valid = OBD_MD_FLGENER | OBD_MD_FLGROUP;
 648
 649                 rc = obd_create(ctxt->loc_exp, oa, NULL, NULL);
 650                 if (rc)
 651                         GOTO(cleanup, rc);
 652
 653                 dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, oa->o_id,
 654                                              oa->o_generation, oa->o_gr);
 655
 656                 if (IS_ERR(dchild))
 657                         GOTO(cleanup, rc = PTR_ERR(dchild));
 658                 cleanup_phase = 2;
 659                 handle->lgh_file = l_dentry_open(&obd->obd_lvfs_ctxt, dchild,
 660                                                  open_flags);
 661                 if (IS_ERR(handle->lgh_file))
 662                         GOTO(cleanup, rc = PTR_ERR(handle->lgh_file));
 663
 664                 handle->lgh_id.lgl_ogr = oa->o_gr;
 665                 handle->lgh_id.lgl_oid = oa->o_id;
 666                 handle->lgh_id.lgl_ogen = oa->o_generation;
 667         }
 668
 669         handle->lgh_ctxt = ctxt;
 670  finish:
 671         if (oa)
 672                 OBDO_FREE(oa);
 673         RETURN(rc);
 674 cleanup:
 675         switch (cleanup_phase) {
 676         case 2:
 677                 l_dput(dchild);
 678         case 1:
 679                 llog_free_handle(handle);
 680         }
 681         goto finish;
 682 }
 683
 684 static int llog_lvfs_close(struct llog_handle *handle)
 685 {
 686         int rc;
 687         ENTRY;
 688
 689         rc = filp_close(handle->lgh_file, 0);
 690         if (rc)
 691                 CERROR("error closing log: rc %d\n", rc);
 692         RETURN(rc);
 693 }
 694
 695 static int llog_lvfs_destroy(struct llog_handle *handle)
 696 {
 697         struct dentry *fdentry;
 698         struct obdo *oa;
 699         struct obd_device *obd = handle->lgh_ctxt->loc_exp->exp_obd;
 700         char *dir;
 701         int rc;
 702         ENTRY;
 703
 704         /* COMPAT_146 */
 705         if (strcmp(obd->obd_type->typ_name, LUSTRE_MDS_NAME) == 0)
 706                 dir = MDT_LOGS_DIR;
 707         else
 708                 /* end COMPAT_146 */
 709                 dir = MOUNT_CONFIGS_DIR;
 710
 711         fdentry = handle->lgh_file->f_dentry;
 712         if (strcmp(fdentry->d_parent->d_name.name, dir) == 0) {
 713                 struct inode *inode = fdentry->d_parent->d_inode;
 714                 struct lvfs_run_ctxt saved;
 715
 716                 push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 717                 dget(fdentry);
 718                 rc = llog_lvfs_close(handle);
 719
 720                 if (rc == 0) {
 721                         LOCK_INODE_MUTEX(inode);
 722                         rc = vfs_unlink(inode, fdentry);
 723                         UNLOCK_INODE_MUTEX(inode);
 724                 }
 725
 726                 dput(fdentry);
 727                 pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 728                 RETURN(rc);
 729         }
 730
 731         OBDO_ALLOC(oa);
 732         if (oa == NULL)
 733                 RETURN(-ENOMEM);
 734
 735         oa->o_id = handle->lgh_id.lgl_oid;
 736         oa->o_gr = handle->lgh_id.lgl_ogr;
 737         oa->o_generation = handle->lgh_id.lgl_ogen;
 738         oa->o_valid = OBD_MD_FLID | OBD_MD_FLGROUP | OBD_MD_FLGENER;
 739
 740         rc = llog_lvfs_close(handle);
 741         if (rc)
 742                 GOTO(out, rc);
 743
 744         rc = obd_destroy(handle->lgh_ctxt->loc_exp, oa, NULL, NULL, NULL);
 745  out:
 746         OBDO_FREE(oa);
 747         RETURN(rc);
 748 }
 749
 750 /* reads the catalog list */
 751 int llog_get_cat_list(struct obd_device *obd, struct obd_device *disk_obd,
 752                       char *name, int count, struct llog_catid *idarray)
 753 {
 754         struct lvfs_run_ctxt saved;
 755         struct l_file *file;
 756         int rc;
 757         int size = sizeof(*idarray) * count;
 758         loff_t off = 0;
 759         ENTRY;
 760
 761         if (!count)
 762                 RETURN(0);
 763
 764         push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 765         file = filp_open(name, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
 766         if (!file || IS_ERR(file)) {
 767                 rc = PTR_ERR(file);
 768                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
 769                        name, rc);
 770                 GOTO(out, rc);
 771         }
 772
 773         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 774                 CERROR("%s is not a regular file!: mode = %o\n", name,
 775                        file->f_dentry->d_inode->i_mode);
 776                 GOTO(out, rc = -ENOENT);
 777         }
 778
 779         CDEBUG(D_CONFIG, "cat list: disk size=%d, read=%d\n",
 780                (int)i_size_read(file->f_dentry->d_inode), size);
 781
 782         rc = fsfilt_read_record(disk_obd, file, idarray, size, &off);
 783         if (rc) {
 784                 CERROR("OBD filter: error reading %s: rc %d\n", name, rc);
 785                 GOTO(out, rc);
 786         }
 787
 788         EXIT;
 789  out:
 790         pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 791         if (file && !IS_ERR(file))
 792                 rc = filp_close(file, 0);
 793         return rc;
 794 }
 795 EXPORT_SYMBOL(llog_get_cat_list);
 796
 797 /* writes the cat list */
 798 int llog_put_cat_list(struct obd_device *obd, struct obd_device *disk_obd,
 799                       char *name, int count, struct llog_catid *idarray)
 800 {
 801         struct lvfs_run_ctxt saved;
 802         struct l_file *file;
 803         int rc;
 804         int size = sizeof(*idarray) * count;
 805         loff_t off = 0;
 806
 807         if (!count)
 808                 return (0);
 809
 810         push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 811         file = filp_open(name, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
 812         if (!file || IS_ERR(file)) {
 813                 rc = PTR_ERR(file);
 814                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
 815                        name, rc);
 816                 GOTO(out, rc);
 817         }
 818
 819         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 820                 CERROR("%s is not a regular file!: mode = %o\n", name,
 821                        file->f_dentry->d_inode->i_mode);
 822                 GOTO(out, rc = -ENOENT);
 823         }
 824
 825         rc = fsfilt_write_record(disk_obd, file, idarray, size, &off, 1);
 826         if (rc) {
 827                 CDEBUG(D_INODE,"OBD filter: error reading %s: rc %d\n",
 828                        name, rc);
 829                 GOTO(out, rc);
 830         }
 831
 832  out:
 833         pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 834         if (file && !IS_ERR(file))
 835                 rc = filp_close(file, 0);
 836         RETURN(rc);
 837 }
 838
 839 struct llog_operations llog_lvfs_ops = {
 840         lop_write_rec:   llog_lvfs_write_rec,
 841         lop_next_block:  llog_lvfs_next_block,
 842         lop_prev_block:  llog_lvfs_prev_block,
 843         lop_read_header: llog_lvfs_read_header,
 844         lop_create:      llog_lvfs_create,
 845         lop_destroy:     llog_lvfs_destroy,
 846         lop_close:       llog_lvfs_close,
 847         //        lop_cancel: llog_lvfs_cancel,
 848 };
 849
 850 EXPORT_SYMBOL(llog_lvfs_ops);
 851
 852 #else /* !__KERNEL__ */
 853
 854 static int llog_lvfs_read_header(struct llog_handle *handle)
 855 {
 856         LBUG();
 857         return 0;
 858 }
 859
 860 static int llog_lvfs_write_rec(struct llog_handle *loghandle,
 861                                struct llog_rec_hdr *rec,
 862                                struct llog_cookie *reccookie, int cookiecount,
 863                                void *buf, int idx)
 864 {
 865         LBUG();
 866         return 0;
 867 }
 868
 869 static int llog_lvfs_next_block(struct llog_handle *loghandle, int *cur_idx,
 870                                 int next_idx, __u64 *cur_offset, void *buf,
 871                                 int len)
 872 {
 873         LBUG();
 874         return 0;
 875 }
 876
 877 static int llog_lvfs_prev_block(struct llog_handle *loghandle,
 878                                 int prev_idx, void *buf, int len)
 879 {
 880         LBUG();
 881         return 0;
 882 }
 883
 884 static int llog_lvfs_create(struct llog_ctxt *ctxt, struct llog_handle **res,
 885                             struct llog_logid *logid, char *name)
 886 {
 887         LBUG();
 888         return 0;
 889 }
 890
 891 static int llog_lvfs_close(struct llog_handle *handle)
 892 {
 893         LBUG();
 894         return 0;
 895 }
 896
 897 static int llog_lvfs_destroy(struct llog_handle *handle)
 898 {
 899         LBUG();
 900         return 0;
 901 }
 902
 903 int llog_get_cat_list(struct obd_device *obd, struct obd_device *disk_obd,
 904                       char *name, int count, struct llog_catid *idarray)
 905 {
 906         LBUG();
 907         return 0;
 908 }
 909
 910 int llog_put_cat_list(struct obd_device *obd, struct obd_device *disk_obd,
 911                       char *name, int count, struct llog_catid *idarray)
 912 {
 913         LBUG();
 914         return 0;
 915 }
 916
 917 struct llog_operations llog_lvfs_ops = {
 918         lop_write_rec:   llog_lvfs_write_rec,
 919         lop_next_block:  llog_lvfs_next_block,
 920         lop_prev_block:  llog_lvfs_prev_block,
 921         lop_read_header: llog_lvfs_read_header,
 922         lop_create:      llog_lvfs_create,
 923         lop_destroy:     llog_lvfs_destroy,
 924         lop_close:       llog_lvfs_close,
 925 //        lop_cancel:      llog_lvfs_cancel,
 926 };
 927 #endif