lustre/obdclass/llog_lvfs.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  */
  30 /*
  31  * This file is part of Lustre, http://www.lustre.org/
  32  * Lustre is a trademark of Sun Microsystems, Inc.
  33  *
  34  * lustre/obdclass/llog_lvfs.c
  35  *
  36  * OST<->MDS recovery logging infrastructure.
  37  * Invariants in implementation:
  38  * - we do not share logs among different OST<->MDS connections, so that
  39  *   if an OST or MDS fails it need only look at log(s) relevant to itself
  40  *
  41  * Author: Andreas Dilger <adilger@clusterfs.com>
  42  */
  43
  44 #define DEBUG_SUBSYSTEM S_LOG
  45
  46 #ifndef __KERNEL__
  47 #include <liblustre.h>
  48 #endif
  49
  50 #include <obd.h>
  51 #include <obd_class.h>
  52 #include <lustre_log.h>
  53 #include <obd_ost.h>
  54 #include <libcfs/list.h>
  55 #include <lvfs.h>
  56 #include <lustre_fsfilt.h>
  57 #include <lustre_disk.h>
  58 #include "llog_internal.h"
  59
  60 #if defined(__KERNEL__) && defined(LLOG_LVFS)
  61
  62 static int llog_lvfs_pad(struct obd_device *obd, struct l_file *file,
  63                                 int len, int index)
  64 {
  65         struct llog_rec_hdr rec = { 0 };
  66         struct llog_rec_tail tail;
  67         int rc;
  68         ENTRY;
  69
  70         LASSERT(len >= LLOG_MIN_REC_SIZE && (len & 0x7) == 0);
  71
  72         tail.lrt_len = rec.lrh_len = len;
  73         tail.lrt_index = rec.lrh_index = index;
  74         rec.lrh_type = LLOG_PAD_MAGIC;
  75
  76         rc = fsfilt_write_record(obd, file, &rec, sizeof(rec), &file->f_pos, 0);
  77         if (rc) {
  78                 CERROR("error writing padding record: rc %d\n", rc);
  79                 goto out;
  80         }
  81
  82         file->f_pos += len - sizeof(rec) - sizeof(tail);
  83         rc = fsfilt_write_record(obd, file, &tail, sizeof(tail),&file->f_pos,0);
  84         if (rc) {
  85                 CERROR("error writing padding record: rc %d\n", rc);
  86                 goto out;
  87         }
  88
  89  out:
  90         RETURN(rc);
  91 }
  92
  93 static int llog_lvfs_write_blob(struct obd_device *obd, struct l_file *file,
  94                                 struct llog_rec_hdr *rec, void *buf, loff_t off)
  95 {
  96         int rc;
  97         struct llog_rec_tail end;
  98         loff_t saved_off = file->f_pos;
  99         int buflen = rec->lrh_len;
 100
 101         ENTRY;
 102
 103         file->f_pos = off;
 104
 105         if (buflen == 0)
 106                 CWARN("0-length record\n");
 107
 108         if (!buf) {
 109                 rc = fsfilt_write_record(obd, file, rec, buflen,&file->f_pos,0);
 110                 if (rc) {
 111                         CERROR("error writing log record: rc %d\n", rc);
 112                         goto out;
 113                 }
 114                 GOTO(out, rc = 0);
 115         }
 116
 117         /* the buf case */
 118         rec->lrh_len = sizeof(*rec) + buflen + sizeof(end);
 119         rc = fsfilt_write_record(obd, file, rec, sizeof(*rec), &file->f_pos, 0);
 120         if (rc) {
 121                 CERROR("error writing log hdr: rc %d\n", rc);
 122                 goto out;
 123         }
 124
 125         rc = fsfilt_write_record(obd, file, buf, buflen, &file->f_pos, 0);
 126         if (rc) {
 127                 CERROR("error writing log buffer: rc %d\n", rc);
 128                 goto out;
 129         }
 130
 131         end.lrt_len = rec->lrh_len;
 132         end.lrt_index = rec->lrh_index;
 133         rc = fsfilt_write_record(obd, file, &end, sizeof(end), &file->f_pos, 0);
 134         if (rc) {
 135                 CERROR("error writing log tail: rc %d\n", rc);
 136                 goto out;
 137         }
 138
 139         rc = 0;
 140  out:
 141         if (saved_off > file->f_pos)
 142                 file->f_pos = saved_off;
 143         LASSERT(rc <= 0);
 144         RETURN(rc);
 145 }
 146
 147 static int llog_lvfs_read_blob(struct obd_device *obd, struct l_file *file,
 148                                 void *buf, int size, loff_t off)
 149 {
 150         loff_t offset = off;
 151         int rc;
 152         ENTRY;
 153
 154         rc = fsfilt_read_record(obd, file, buf, size, &offset);
 155         if (rc) {
 156                 CERROR("error reading log record: rc %d\n", rc);
 157                 RETURN(rc);
 158         }
 159         RETURN(0);
 160 }
 161
 162 static int llog_lvfs_read_header(struct llog_handle *handle)
 163 {
 164         struct obd_device *obd;
 165         int rc;
 166         ENTRY;
 167
 168         LASSERT(sizeof(*handle->lgh_hdr) == LLOG_CHUNK_SIZE);
 169
 170         obd = handle->lgh_ctxt->loc_exp->exp_obd;
 171
 172         if (i_size_read(handle->lgh_file->f_dentry->d_inode) == 0) {
 173                 CDEBUG(D_HA, "not reading header from 0-byte log\n");
 174                 RETURN(LLOG_EEMPTY);
 175         }
 176
 177         rc = llog_lvfs_read_blob(obd, handle->lgh_file, handle->lgh_hdr,
 178                                  LLOG_CHUNK_SIZE, 0);
 179         if (rc) {
 180                 CERROR("error reading log header from %.*s\n",
 181                        handle->lgh_file->f_dentry->d_name.len,
 182                        handle->lgh_file->f_dentry->d_name.name);
 183         } else {
 184                 struct llog_rec_hdr *llh_hdr = &handle->lgh_hdr->llh_hdr;
 185
 186                 if (LLOG_REC_HDR_NEEDS_SWABBING(llh_hdr))
 187                         lustre_swab_llog_hdr(handle->lgh_hdr);
 188
 189                 if (llh_hdr->lrh_type != LLOG_HDR_MAGIC) {
 190                         CERROR("bad log %.*s header magic: %#x (expected %#x)\n",
 191                                handle->lgh_file->f_dentry->d_name.len,
 192                                handle->lgh_file->f_dentry->d_name.name,
 193                                llh_hdr->lrh_type, LLOG_HDR_MAGIC);
 194                         rc = -EIO;
 195                 } else if (llh_hdr->lrh_len != LLOG_CHUNK_SIZE) {
 196                         CERROR("incorrectly sized log %.*s header: %#x "
 197                                "(expected %#x)\n",
 198                                handle->lgh_file->f_dentry->d_name.len,
 199                                handle->lgh_file->f_dentry->d_name.name,
 200                                llh_hdr->lrh_len, LLOG_CHUNK_SIZE);
 201                         CERROR("you may need to re-run lconf --write_conf.\n");
 202                         rc = -EIO;
 203                 }
 204         }
 205
 206         handle->lgh_last_idx = handle->lgh_hdr->llh_tail.lrt_index;
 207         handle->lgh_file->f_pos = i_size_read(handle->lgh_file->f_dentry->d_inode);
 208
 209         RETURN(rc);
 210 }
 211
 212 /* returns negative in on error; 0 if success && reccookie == 0; 1 otherwise */
 213 /* appends if idx == -1, otherwise overwrites record idx. */
 214 static int llog_lvfs_write_rec(struct llog_handle *loghandle,
 215                                struct llog_rec_hdr *rec,
 216                                struct llog_cookie *reccookie, int cookiecount,
 217                                void *buf, int idx)
 218 {
 219         struct llog_log_hdr *llh;
 220         int reclen = rec->lrh_len, index, rc;
 221         struct llog_rec_tail *lrt;
 222         struct obd_device *obd;
 223         struct file *file;
 224         size_t left;
 225         ENTRY;
 226
 227         llh = loghandle->lgh_hdr;
 228         file = loghandle->lgh_file;
 229         obd = loghandle->lgh_ctxt->loc_exp->exp_obd;
 230
 231         /* record length should not bigger than LLOG_CHUNK_SIZE */
 232         if (buf)
 233                 rc = (reclen > LLOG_CHUNK_SIZE - sizeof(struct llog_rec_hdr) -
 234                       sizeof(struct llog_rec_tail)) ? -E2BIG : 0;
 235         else
 236                 rc = (reclen > LLOG_CHUNK_SIZE) ? -E2BIG : 0;
 237         if (rc)
 238                 RETURN(rc);
 239
 240         if (buf)
 241                 /* write_blob adds header and tail to lrh_len. */
 242                 reclen = sizeof(*rec) + rec->lrh_len +
 243                          sizeof(struct llog_rec_tail);
 244
 245         if (idx != -1) {
 246                 loff_t saved_offset;
 247
 248                 /* no header: only allowed to insert record 1 */
 249                 if (idx != 1 && !i_size_read(file->f_dentry->d_inode)) {
 250                         CERROR("idx != -1 in empty log\n");
 251                         LBUG();
 252                 }
 253
 254                 if (idx && llh->llh_size && llh->llh_size != rec->lrh_len)
 255                         RETURN(-EINVAL);
 256
 257                 if (!ext2_test_bit(idx, llh->llh_bitmap))
 258                         CERROR("Modify unset record %u\n", idx);
 259                 if (idx != rec->lrh_index)
 260                         CERROR("Index mismatch %d %u\n", idx, rec->lrh_index);
 261
 262                 rc = llog_lvfs_write_blob(obd, file, &llh->llh_hdr, NULL, 0);
 263                 /* we are done if we only write the header or on error */
 264                 if (rc || idx == 0)
 265                         RETURN(rc);
 266
 267                 /* Assumes constant lrh_len */
 268                 saved_offset = sizeof(*llh) + (idx - 1) * reclen;
 269
 270                 if (buf) {
 271                         struct llog_rec_hdr check;
 272
 273                         /* We assume that caller has set lgh_cur_* */
 274                         saved_offset = loghandle->lgh_cur_offset;
 275                         CDEBUG(D_OTHER,
 276                                "modify record "LPX64": idx:%d/%u/%d, len:%u "
 277                                "offset %llu\n",
 278                                loghandle->lgh_id.lgl_oid, idx, rec->lrh_index,
 279                                loghandle->lgh_cur_idx, rec->lrh_len,
 280                                (long long)(saved_offset - sizeof(*llh)));
 281                         if (rec->lrh_index != loghandle->lgh_cur_idx) {
 282                                 CERROR("modify idx mismatch %u/%d\n",
 283                                        idx, loghandle->lgh_cur_idx);
 284                                 RETURN(-EFAULT);
 285                         }
 286 #if 1  /* FIXME remove this safety check at some point */
 287                         /* Verify that the record we're modifying is the
 288                            right one. */
 289                         rc = llog_lvfs_read_blob(obd, file, &check,
 290                                                  sizeof(check), saved_offset);
 291                         if (check.lrh_index != idx || check.lrh_len != reclen) {
 292                                 CERROR("Bad modify idx %u/%u size %u/%u (%d)\n",
 293                                        idx, check.lrh_index, reclen,
 294                                        check.lrh_len, rc);
 295                                 RETURN(-EFAULT);
 296                         }
 297 #endif
 298                 }
 299
 300                 rc = llog_lvfs_write_blob(obd, file, rec, buf, saved_offset);
 301                 if (rc == 0 && reccookie) {
 302                         reccookie->lgc_lgl = loghandle->lgh_id;
 303                         reccookie->lgc_index = idx;
 304                         rc = 1;
 305                 }
 306                 RETURN(rc);
 307         }
 308
 309         /* Make sure that records don't cross a chunk boundary, so we can
 310          * process them page-at-a-time if needed.  If it will cross a chunk
 311          * boundary, write in a fake (but referenced) entry to pad the chunk.
 312          *
 313          * We know that llog_current_log() will return a loghandle that is
 314          * big enough to hold reclen, so all we care about is padding here.
 315          */
 316         left = LLOG_CHUNK_SIZE - (file->f_pos & (LLOG_CHUNK_SIZE - 1));
 317
 318         /* NOTE: padding is a record, but no bit is set */
 319         if (left != 0 && left != reclen &&
 320             left < (reclen + LLOG_MIN_REC_SIZE)) {
 321                  index = loghandle->lgh_last_idx + 1;
 322                  rc = llog_lvfs_pad(obd, file, left, index);
 323                  if (rc)
 324                          RETURN(rc);
 325                  loghandle->lgh_last_idx++; /*for pad rec*/
 326          }
 327          /* if it's the last idx in log file, then return -ENOSPC */
 328          if (loghandle->lgh_last_idx >= LLOG_BITMAP_SIZE(llh) - 1)
 329                  RETURN(-ENOSPC);
 330         loghandle->lgh_last_idx++;
 331         index = loghandle->lgh_last_idx;
 332         LASSERT(index < LLOG_BITMAP_SIZE(llh));
 333         rec->lrh_index = index;
 334         if (buf == NULL) {
 335                 lrt = (struct llog_rec_tail *)
 336                         ((char *)rec + rec->lrh_len - sizeof(*lrt));
 337                 lrt->lrt_len = rec->lrh_len;
 338                 lrt->lrt_index = rec->lrh_index;
 339         }
 340         /*The caller should make sure only 1 process access the lgh_last_idx,
 341          *Otherwise it might hit the assert.*/
 342         LASSERT(index < LLOG_BITMAP_SIZE(llh));
 343         if (ext2_set_bit(index, llh->llh_bitmap)) {
 344                 CERROR("argh, index %u already set in log bitmap?\n", index);
 345                 LBUG(); /* should never happen */
 346         }
 347         llh->llh_count++;
 348         llh->llh_tail.lrt_index = index;
 349
 350         rc = llog_lvfs_write_blob(obd, file, &llh->llh_hdr, NULL, 0);
 351         if (rc)
 352                 RETURN(rc);
 353
 354         rc = llog_lvfs_write_blob(obd, file, rec, buf, file->f_pos);
 355         if (rc)
 356                 RETURN(rc);
 357
 358         CDEBUG(D_RPCTRACE, "added record "LPX64": idx: %u, %u \n",
 359                loghandle->lgh_id.lgl_oid, index, rec->lrh_len);
 360         if (rc == 0 && reccookie) {
 361                 reccookie->lgc_lgl = loghandle->lgh_id;
 362                 reccookie->lgc_index = index;
 363                 if ((rec->lrh_type == MDS_UNLINK_REC) ||
 364                     (rec->lrh_type == MDS_SETATTR64_REC))
 365                         reccookie->lgc_subsys = LLOG_MDS_OST_ORIG_CTXT;
 366                 else if (rec->lrh_type == OST_SZ_REC)
 367                         reccookie->lgc_subsys = LLOG_SIZE_ORIG_CTXT;
 368                 else
 369                         reccookie->lgc_subsys = -1;
 370                 rc = 1;
 371         }
 372         if (rc == 0 && rec->lrh_type == LLOG_GEN_REC)
 373                 rc = 1;
 374
 375         RETURN(rc);
 376 }
 377
 378 /* We can skip reading at least as many log blocks as the number of
 379 * minimum sized log records we are skipping.  If it turns out
 380 * that we are not far enough along the log (because the
 381 * actual records are larger than minimum size) we just skip
 382 * some more records. */
 383
 384 static void llog_skip_over(__u64 *off, int curr, int goal)
 385 {
 386         if (goal <= curr)
 387                 return;
 388         *off = (*off + (goal-curr-1) * LLOG_MIN_REC_SIZE) &
 389                 ~(LLOG_CHUNK_SIZE - 1);
 390 }
 391
 392
 393 /* sets:
 394  *  - cur_offset to the furthest point read in the log file
 395  *  - cur_idx to the log index preceeding cur_offset
 396  * returns -EIO/-EINVAL on error
 397  */
 398 static int llog_lvfs_next_block(struct llog_handle *loghandle, int *cur_idx,
 399                                 int next_idx, __u64 *cur_offset, void *buf,
 400                                 int len)
 401 {
 402         int rc;
 403         ENTRY;
 404
 405         if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
 406                 RETURN(-EINVAL);
 407
 408         CDEBUG(D_OTHER, "looking for log index %u (cur idx %u off "LPU64")\n",
 409                next_idx, *cur_idx, *cur_offset);
 410
 411         while (*cur_offset < i_size_read(loghandle->lgh_file->f_dentry->d_inode)) {
 412                 struct llog_rec_hdr *rec;
 413                 struct llog_rec_tail *tail;
 414                 loff_t ppos;
 415
 416                 llog_skip_over(cur_offset, *cur_idx, next_idx);
 417
 418                 ppos = *cur_offset;
 419                 rc = fsfilt_read_record(loghandle->lgh_ctxt->loc_exp->exp_obd,
 420                                         loghandle->lgh_file, buf, len,
 421                                         &ppos);
 422                 if (rc) {
 423                         CERROR("Cant read llog block at log id "LPU64
 424                                "/%u offset "LPU64"\n",
 425                                loghandle->lgh_id.lgl_oid,
 426                                loghandle->lgh_id.lgl_ogen,
 427                                *cur_offset);
 428                         RETURN(rc);
 429                 }
 430
 431                 /* put number of bytes read into rc to make code simpler */
 432                 rc = ppos - *cur_offset;
 433                 *cur_offset = ppos;
 434
 435                 if (rc < len) {
 436                         /* signal the end of the valid buffer to llog_process */
 437                         memset(buf + rc, 0, len - rc);
 438                 }
 439
 440                 if (rc == 0) /* end of file, nothing to do */
 441                         RETURN(0);
 442
 443                 if (rc < sizeof(*tail)) {
 444                         CERROR("Invalid llog block at log id "LPU64"/%u offset "
 445                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 446                                loghandle->lgh_id.lgl_ogen, *cur_offset);
 447                         RETURN(-EINVAL);
 448                 }
 449
 450                 rec = buf;
 451                 tail = (struct llog_rec_tail *)((char *)buf + rc -
 452                                                 sizeof(struct llog_rec_tail));
 453
 454                 if (LLOG_REC_HDR_NEEDS_SWABBING(rec))
 455                         lustre_swab_llog_rec(rec);
 456
 457                 *cur_idx = tail->lrt_index;
 458
 459                 /* this shouldn't happen */
 460                 if (tail->lrt_index == 0) {
 461                         CERROR("Invalid llog tail at log id "LPU64"/%u offset "
 462                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 463                                loghandle->lgh_id.lgl_ogen, *cur_offset);
 464                         RETURN(-EINVAL);
 465                 }
 466                 if (tail->lrt_index < next_idx)
 467                         continue;
 468
 469                 /* sanity check that the start of the new buffer is no farther
 470                  * than the record that we wanted.  This shouldn't happen. */
 471                 if (rec->lrh_index > next_idx) {
 472                         CERROR("missed desired record? %u > %u\n",
 473                                rec->lrh_index, next_idx);
 474                         RETURN(-ENOENT);
 475                 }
 476                 RETURN(0);
 477         }
 478         RETURN(-EIO);
 479 }
 480
 481 static int llog_lvfs_prev_block(struct llog_handle *loghandle,
 482                                 int prev_idx, void *buf, int len)
 483 {
 484         __u64 cur_offset;
 485         int rc;
 486         ENTRY;
 487
 488         if (len == 0 || len & (LLOG_CHUNK_SIZE - 1))
 489                 RETURN(-EINVAL);
 490
 491         CDEBUG(D_OTHER, "looking for log index %u\n", prev_idx);
 492
 493         cur_offset = LLOG_CHUNK_SIZE;
 494         llog_skip_over(&cur_offset, 0, prev_idx);
 495
 496         while (cur_offset < i_size_read(loghandle->lgh_file->f_dentry->d_inode)) {
 497                 struct llog_rec_hdr *rec;
 498                 struct llog_rec_tail *tail;
 499                 loff_t ppos;
 500
 501                 ppos = cur_offset;
 502
 503                 rc = fsfilt_read_record(loghandle->lgh_ctxt->loc_exp->exp_obd,
 504                                         loghandle->lgh_file, buf, len,
 505                                         &ppos);
 506                 if (rc) {
 507                         CERROR("Cant read llog block at log id "LPU64
 508                                "/%u offset "LPU64"\n",
 509                                loghandle->lgh_id.lgl_oid,
 510                                loghandle->lgh_id.lgl_ogen,
 511                                cur_offset);
 512                         RETURN(rc);
 513                 }
 514
 515                 /* put number of bytes read into rc to make code simpler */
 516                 rc = ppos - cur_offset;
 517                 cur_offset = ppos;
 518
 519                 if (rc == 0) /* end of file, nothing to do */
 520                         RETURN(0);
 521
 522                 if (rc < sizeof(*tail)) {
 523                         CERROR("Invalid llog block at log id "LPU64"/%u offset "
 524                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 525                                loghandle->lgh_id.lgl_ogen, cur_offset);
 526                         RETURN(-EINVAL);
 527                 }
 528
 529                 tail = buf + rc - sizeof(struct llog_rec_tail);
 530
 531                 /* this shouldn't happen */
 532                 if (tail->lrt_index == 0) {
 533                         CERROR("Invalid llog tail at log id "LPU64"/%u offset "
 534                                LPU64"\n", loghandle->lgh_id.lgl_oid,
 535                                loghandle->lgh_id.lgl_ogen, cur_offset);
 536                         RETURN(-EINVAL);
 537                 }
 538                 if (le32_to_cpu(tail->lrt_index) < prev_idx)
 539                         continue;
 540
 541                 /* sanity check that the start of the new buffer is no farther
 542                  * than the record that we wanted.  This shouldn't happen. */
 543                 rec = buf;
 544                 if (le32_to_cpu(rec->lrh_index) > prev_idx) {
 545                         CERROR("missed desired record? %u > %u\n",
 546                                le32_to_cpu(rec->lrh_index), prev_idx);
 547                         RETURN(-ENOENT);
 548                 }
 549                 RETURN(0);
 550         }
 551         RETURN(-EIO);
 552 }
 553
 554 static struct file *llog_filp_open(char *dir, char *name, int flags, int mode)
 555 {
 556         char *logname;
 557         struct file *filp;
 558         int len;
 559
 560         OBD_ALLOC(logname, PATH_MAX);
 561         if (logname == NULL)
 562                 return ERR_PTR(-ENOMEM);
 563
 564         len = snprintf(logname, PATH_MAX, "%s/%s", dir, name);
 565         if (len >= PATH_MAX - 1) {
 566                 filp = ERR_PTR(-ENAMETOOLONG);
 567         } else {
 568                 filp = l_filp_open(logname, flags, mode);
 569                 if (IS_ERR(filp))
 570                         CERROR("logfile creation %s: %ld\n", logname,
 571                                PTR_ERR(filp));
 572         }
 573         OBD_FREE(logname, PATH_MAX);
 574         return filp;
 575 }
 576
 577 /* This is a callback from the llog_* functions.
 578  * Assumes caller has already pushed us into the kernel context. */
 579 static int llog_lvfs_create(struct llog_ctxt *ctxt, struct llog_handle **res,
 580                             struct llog_logid *logid, char *name)
 581 {
 582         struct llog_handle *handle;
 583         struct obd_device *obd;
 584         struct l_dentry *dchild = NULL;
 585         struct obdo *oa = NULL;
 586         int rc = 0;
 587         int open_flags = O_RDWR | O_CREAT | O_LARGEFILE;
 588         ENTRY;
 589
 590         handle = llog_alloc_handle();
 591         if (handle == NULL)
 592                 RETURN(-ENOMEM);
 593         *res = handle;
 594
 595         LASSERT(ctxt);
 596         LASSERT(ctxt->loc_exp);
 597         obd = ctxt->loc_exp->exp_obd;
 598
 599         if (logid != NULL) {
 600                 dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, logid->lgl_oid,
 601                                              logid->lgl_ogen, logid->lgl_oseq);
 602
 603                 if (IS_ERR(dchild)) {
 604                         rc = PTR_ERR(dchild);
 605                         CERROR("error looking up logfile "LPX64":0x%x: rc %d\n",
 606                                logid->lgl_oid, logid->lgl_ogen, rc);
 607                         GOTO(out, rc);
 608                 }
 609
 610                 if (dchild->d_inode == NULL) {
 611                         l_dput(dchild);
 612                         rc = -ENOENT;
 613                         CERROR("nonexistent log file "LPX64":"LPX64": rc %d\n",
 614                                logid->lgl_oid, logid->lgl_oseq, rc);
 615                         GOTO(out, rc);
 616                 }
 617
 618                 /* l_dentry_open will call dput(dchild) if there is an error */
 619                 handle->lgh_file = l_dentry_open(&obd->obd_lvfs_ctxt, dchild,
 620                                                     O_RDWR | O_LARGEFILE);
 621                 if (IS_ERR(handle->lgh_file)) {
 622                         rc = PTR_ERR(handle->lgh_file);
 623                         CERROR("error opening logfile "LPX64"0x%x: rc %d\n",
 624                                logid->lgl_oid, logid->lgl_ogen, rc);
 625                         GOTO(out, rc);
 626                 }
 627
 628                 /* assign the value of lgh_id for handle directly */
 629                 handle->lgh_id = *logid;
 630
 631         } else if (name) {
 632                 handle->lgh_file = llog_filp_open(MOUNT_CONFIGS_DIR,
 633                                                   name, open_flags, 0644);
 634                 if (IS_ERR(handle->lgh_file))
 635                         GOTO(out, rc = PTR_ERR(handle->lgh_file));
 636
 637                 handle->lgh_id.lgl_oseq = 1;
 638                 handle->lgh_id.lgl_oid =
 639                         handle->lgh_file->f_dentry->d_inode->i_ino;
 640                 handle->lgh_id.lgl_ogen =
 641                         handle->lgh_file->f_dentry->d_inode->i_generation;
 642         } else {
 643                 OBDO_ALLOC(oa);
 644                 if (oa == NULL)
 645                         GOTO(out, rc = -ENOMEM);
 646
 647                 oa->o_seq = FID_SEQ_LLOG;
 648                 oa->o_valid = OBD_MD_FLGENER | OBD_MD_FLGROUP;
 649
 650                 rc = obd_create(NULL, ctxt->loc_exp, oa, NULL, NULL);
 651                 if (rc)
 652                         GOTO(out, rc);
 653
 654                 /* FIXME: rationalize the misuse of o_generation in
 655                  *        this API along with mds_obd_{create,destroy}.
 656                  *        Hopefully it is only an internal API issue. */
 657 #define o_generation o_parent_oid
 658                 dchild = obd_lvfs_fid2dentry(ctxt->loc_exp, oa->o_id,
 659                                              oa->o_generation, oa->o_seq);
 660
 661                 if (IS_ERR(dchild))
 662                         GOTO(out, rc = PTR_ERR(dchild));
 663
 664                 handle->lgh_file = l_dentry_open(&obd->obd_lvfs_ctxt, dchild,
 665                                                  open_flags);
 666                 if (IS_ERR(handle->lgh_file))
 667                         GOTO(out, rc = PTR_ERR(handle->lgh_file));
 668
 669                 handle->lgh_id.lgl_oseq = oa->o_seq;
 670                 handle->lgh_id.lgl_oid = oa->o_id;
 671                 handle->lgh_id.lgl_ogen = oa->o_generation;
 672         }
 673
 674         handle->lgh_ctxt = ctxt;
 675 out:
 676         if (rc)
 677                 llog_free_handle(handle);
 678
 679         if (oa)
 680                 OBDO_FREE(oa);
 681         RETURN(rc);
 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         void *th;
 702         struct inode *inode;
 703         int rc, rc1;
 704         ENTRY;
 705
 706         dir = MOUNT_CONFIGS_DIR;
 707
 708         fdentry = handle->lgh_file->f_dentry;
 709         inode = fdentry->d_parent->d_inode;
 710         if (strcmp(fdentry->d_parent->d_name.name, dir) == 0) {
 711                 struct lvfs_run_ctxt saved;
 712                 struct vfsmount *mnt = mntget(handle->lgh_file->f_vfsmnt);
 713
 714                 push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 715                 dget(fdentry);
 716                 rc = llog_lvfs_close(handle);
 717
 718                 if (rc == 0) {
 719                         mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
 720                         rc = ll_vfs_unlink(inode, fdentry, mnt);
 721                         mutex_unlock(&inode->i_mutex);
 722                 }
 723                 mntput(mnt);
 724
 725                 dput(fdentry);
 726                 pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
 727                 RETURN(rc);
 728         }
 729
 730         OBDO_ALLOC(oa);
 731         if (oa == NULL)
 732                 RETURN(-ENOMEM);
 733
 734         oa->o_id = handle->lgh_id.lgl_oid;
 735         oa->o_seq = handle->lgh_id.lgl_oseq;
 736         oa->o_generation = handle->lgh_id.lgl_ogen;
 737 #undef o_generation
 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         th = fsfilt_start_log(obd, inode, FSFILT_OP_UNLINK, NULL, 1);
 745         if (IS_ERR(th)) {
 746                 CERROR("fsfilt_start failed: %ld\n", PTR_ERR(th));
 747                 GOTO(out, rc = PTR_ERR(th));
 748         }
 749
 750         rc = obd_destroy(NULL, handle->lgh_ctxt->loc_exp, oa,
 751                          NULL, NULL, NULL, NULL);
 752
 753         rc1 = fsfilt_commit(obd, inode, th, 0);
 754         if (rc == 0 && rc1 != 0)
 755                 rc = rc1;
 756  out:
 757         OBDO_FREE(oa);
 758         RETURN(rc);
 759 }
 760
 761 /* reads the catalog list */
 762 int llog_get_cat_list(struct obd_device *disk_obd,
 763                       char *name, int idx, int count, struct llog_catid *idarray)
 764 {
 765         struct lvfs_run_ctxt saved;
 766         struct l_file *file;
 767         int rc, rc1 = 0;
 768         int size = sizeof(*idarray) * count;
 769         loff_t off = idx *  sizeof(*idarray);
 770         ENTRY;
 771
 772         if (!count)
 773                 RETURN(0);
 774
 775         push_ctxt(&saved, &disk_obd->obd_lvfs_ctxt, NULL);
 776         file = filp_open(name, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
 777         if (!file || IS_ERR(file)) {
 778                 rc = PTR_ERR(file);
 779                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
 780                        name, rc);
 781                 GOTO(out, rc);
 782         }
 783
 784         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 785                 CERROR("%s is not a regular file!: mode = %o\n", name,
 786                        file->f_dentry->d_inode->i_mode);
 787                 GOTO(out, rc = -ENOENT);
 788         }
 789
 790         CDEBUG(D_CONFIG, "cat list: disk size=%d, read=%d\n",
 791                (int)i_size_read(file->f_dentry->d_inode), size);
 792
 793         /* read for new ost index or for empty file */
 794         memset(idarray, 0, size);
 795         if (i_size_read(file->f_dentry->d_inode) < off)
 796                 GOTO(out, rc = 0);
 797
 798         rc = fsfilt_read_record(disk_obd, file, idarray, size, &off);
 799         if (rc) {
 800                 CERROR("OBD filter: error reading %s: rc %d\n", name, rc);
 801                 GOTO(out, rc);
 802         }
 803
 804         EXIT;
 805  out:
 806         pop_ctxt(&saved, &disk_obd->obd_lvfs_ctxt, NULL);
 807         if (file && !IS_ERR(file))
 808                 rc1 = filp_close(file, 0);
 809         if (rc == 0)
 810                 rc = rc1;
 811         return rc;
 812 }
 813 EXPORT_SYMBOL(llog_get_cat_list);
 814
 815 /* writes the cat list */
 816 int llog_put_cat_list(struct obd_device *disk_obd,
 817                       char *name, int idx, int count, struct llog_catid *idarray)
 818 {
 819         struct lvfs_run_ctxt saved;
 820         struct l_file *file;
 821         int rc, rc1 = 0;
 822         int size = sizeof(*idarray) * count;
 823         loff_t off = idx * sizeof(*idarray);
 824
 825         if (!count)
 826                 GOTO(out1, rc = 0);
 827
 828         push_ctxt(&saved, &disk_obd->obd_lvfs_ctxt, NULL);
 829         file = filp_open(name, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
 830         if (!file || IS_ERR(file)) {
 831                 rc = PTR_ERR(file);
 832                 CERROR("OBD filter: cannot open/create %s: rc = %d\n",
 833                        name, rc);
 834                 GOTO(out, rc);
 835         }
 836
 837         if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
 838                 CERROR("%s is not a regular file!: mode = %o\n", name,
 839                        file->f_dentry->d_inode->i_mode);
 840                 GOTO(out, rc = -ENOENT);
 841         }
 842
 843         rc = fsfilt_write_record(disk_obd, file, idarray, size, &off, 1);
 844         if (rc) {
 845                 CDEBUG(D_INODE,"OBD filter: error writeing %s: rc %d\n",
 846                        name, rc);
 847                 GOTO(out, rc);
 848         }
 849
 850 out:
 851         pop_ctxt(&saved, &disk_obd->obd_lvfs_ctxt, NULL);
 852         if (file && !IS_ERR(file))
 853                 rc1 = filp_close(file, 0);
 854
 855         if (rc == 0)
 856                 rc = rc1;
 857 out1:
 858         RETURN(rc);
 859 }
 860 EXPORT_SYMBOL(llog_put_cat_list);
 861
 862 struct llog_operations llog_lvfs_ops = {
 863         lop_write_rec:   llog_lvfs_write_rec,
 864         lop_next_block:  llog_lvfs_next_block,
 865         lop_prev_block:  llog_lvfs_prev_block,
 866         lop_read_header: llog_lvfs_read_header,
 867         lop_create:      llog_lvfs_create,
 868         lop_destroy:     llog_lvfs_destroy,
 869         lop_close:       llog_lvfs_close,
 870         //        lop_cancel: llog_lvfs_cancel,
 871 };
 872
 873 EXPORT_SYMBOL(llog_lvfs_ops);
 874
 875 #else /* !__KERNEL__ */
 876
 877 static int llog_lvfs_read_header(struct llog_handle *handle)
 878 {
 879         LBUG();
 880         return 0;
 881 }
 882
 883 static int llog_lvfs_write_rec(struct llog_handle *loghandle,
 884                                struct llog_rec_hdr *rec,
 885                                struct llog_cookie *reccookie, int cookiecount,
 886                                void *buf, int idx)
 887 {
 888         LBUG();
 889         return 0;
 890 }
 891
 892 static int llog_lvfs_next_block(struct llog_handle *loghandle, int *cur_idx,
 893                                 int next_idx, __u64 *cur_offset, void *buf,
 894                                 int len)
 895 {
 896         LBUG();
 897         return 0;
 898 }
 899
 900 static int llog_lvfs_prev_block(struct llog_handle *loghandle,
 901                                 int prev_idx, void *buf, int len)
 902 {
 903         LBUG();
 904         return 0;
 905 }
 906
 907 static int llog_lvfs_create(struct llog_ctxt *ctxt, struct llog_handle **res,
 908                             struct llog_logid *logid, char *name)
 909 {
 910         LBUG();
 911         return 0;
 912 }
 913
 914 static int llog_lvfs_close(struct llog_handle *handle)
 915 {
 916         LBUG();
 917         return 0;
 918 }
 919
 920 static int llog_lvfs_destroy(struct llog_handle *handle)
 921 {
 922         LBUG();
 923         return 0;
 924 }
 925
 926 int llog_get_cat_list(struct obd_device *disk_obd,
 927                       char *name, int idx, int count, struct llog_catid *idarray)
 928 {
 929         LBUG();
 930         return 0;
 931 }
 932
 933 int llog_put_cat_list(struct obd_device *disk_obd,
 934                       char *name, int idx, int count, struct llog_catid *idarray)
 935 {
 936         LBUG();
 937         return 0;
 938 }
 939
 940 struct llog_operations llog_lvfs_ops = {
 941         lop_write_rec:   llog_lvfs_write_rec,
 942         lop_next_block:  llog_lvfs_next_block,
 943         lop_prev_block:  llog_lvfs_prev_block,
 944         lop_read_header: llog_lvfs_read_header,
 945         lop_create:      llog_lvfs_create,
 946         lop_destroy:     llog_lvfs_destroy,
 947         lop_close:       llog_lvfs_close,
 948 //        lop_cancel:      llog_lvfs_cancel,
 949 };
 950 #endif