sys/vfs/hammer/hammer_object.c

   1 /*
   2  * Copyright (c) 2007 The DragonFly Project.  All rights reserved.
   3  *
   4  * This code is derived from software contributed to The DragonFly Project
   5  * by Matthew Dillon <dillon@backplane.com>
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in
  15  *    the documentation and/or other materials provided with the
  16  *    distribution.
  17  * 3. Neither the name of The DragonFly Project nor the names of its
  18  *    contributors may be used to endorse or promote products derived
  19  *    from this software without specific, prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  29  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  30  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  31  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  * $DragonFly: src/sys/vfs/hammer/hammer_object.c,v 1.34 2008/02/24 19:48:45 dillon Exp $
  35  */
  36
  37 #include "hammer.h"
  38
  39 static int hammer_mem_add(hammer_transaction_t trans, hammer_record_t record);
  40 static int hammer_mem_lookup(hammer_cursor_t cursor, hammer_inode_t ip);
  41 static int hammer_mem_first(hammer_cursor_t cursor, hammer_inode_t ip);
  42
  43 /*
  44  * Red-black tree support.
  45  */
  46 static int
  47 hammer_rec_rb_compare(hammer_record_t rec1, hammer_record_t rec2)
  48 {
  49         if (rec1->rec.base.base.rec_type < rec2->rec.base.base.rec_type)
  50                 return(-1);
  51         if (rec1->rec.base.base.rec_type > rec2->rec.base.base.rec_type)
  52                 return(1);
  53
  54         if (rec1->rec.base.base.key < rec2->rec.base.base.key)
  55                 return(-1);
  56         if (rec1->rec.base.base.key > rec2->rec.base.base.key)
  57                 return(1);
  58
  59         if (rec1->rec.base.base.create_tid == 0) {
  60                 if (rec2->rec.base.base.create_tid == 0)
  61                         return(0);
  62                 return(1);
  63         }
  64         if (rec2->rec.base.base.create_tid == 0)
  65                 return(-1);
  66
  67         if (rec1->rec.base.base.create_tid < rec2->rec.base.base.create_tid)
  68                 return(-1);
  69         if (rec1->rec.base.base.create_tid > rec2->rec.base.base.create_tid)
  70                 return(1);
  71         return(0);
  72 }
  73
  74 static int
  75 hammer_rec_compare(hammer_base_elm_t info, hammer_record_t rec)
  76 {
  77         if (info->rec_type < rec->rec.base.base.rec_type)
  78                 return(-3);
  79         if (info->rec_type > rec->rec.base.base.rec_type)
  80                 return(3);
  81
  82         if (info->key < rec->rec.base.base.key)
  83                 return(-2);
  84         if (info->key > rec->rec.base.base.key)
  85                 return(2);
  86
  87         if (info->create_tid == 0) {
  88                 if (rec->rec.base.base.create_tid == 0)
  89                         return(0);
  90                 return(1);
  91         }
  92         if (rec->rec.base.base.create_tid == 0)
  93                 return(-1);
  94         if (info->create_tid < rec->rec.base.base.create_tid)
  95                 return(-1);
  96         if (info->create_tid > rec->rec.base.base.create_tid)
  97                 return(1);
  98         return(0);
  99 }
 100
 101 /*
 102  * RB_SCAN comparison code for hammer_mem_first().  The argument order
 103  * is reversed so the comparison result has to be negated.  key_beg and
 104  * key_end are both range-inclusive.
 105  *
 106  * The creation timestamp can cause hammer_rec_compare() to return -1 or +1.
 107  * These do not stop the scan.
 108  *
 109  * Localized deletions are not cached in-memory.
 110  */
 111 static
 112 int
 113 hammer_rec_scan_cmp(hammer_record_t rec, void *data)
 114 {
 115         hammer_cursor_t cursor = data;
 116         int r;
 117
 118         r = hammer_rec_compare(&cursor->key_beg, rec);
 119         if (r > 1)
 120                 return(-1);
 121         r = hammer_rec_compare(&cursor->key_end, rec);
 122         if (r < -1)
 123                 return(1);
 124         return(0);
 125 }
 126
 127 RB_GENERATE(hammer_rec_rb_tree, hammer_record, rb_node, hammer_rec_rb_compare);
 128 RB_GENERATE_XLOOKUP(hammer_rec_rb_tree, INFO, hammer_record, rb_node,
 129                     hammer_rec_compare, hammer_base_elm_t);
 130
 131 /*
 132  * Allocate a record for the caller to finish filling in.  The record is
 133  * returned referenced.
 134  */
 135 hammer_record_t
 136 hammer_alloc_mem_record(hammer_inode_t ip)
 137 {
 138         hammer_record_t record;
 139
 140         ++hammer_count_records;
 141         record = kmalloc(sizeof(*record), M_HAMMER, M_WAITOK|M_ZERO);
 142         record->ip = ip;
 143         record->rec.base.base.btype = HAMMER_BTREE_TYPE_RECORD;
 144         hammer_ref(&record->lock);
 145         return (record);
 146 }
 147
 148 /*
 149  * Release a memory record.  Records marked for deletion are immediately
 150  * removed from the RB-Tree but otherwise left intact until the last ref
 151  * goes away.
 152  */
 153 void
 154 hammer_rel_mem_record(struct hammer_record *record)
 155 {
 156         hammer_unref(&record->lock);
 157
 158         if (record->flags & HAMMER_RECF_DELETED) {
 159                 if (record->flags & HAMMER_RECF_ONRBTREE) {
 160                         RB_REMOVE(hammer_rec_rb_tree, &record->ip->rec_tree,
 161                                   record);
 162                         record->flags &= ~HAMMER_RECF_ONRBTREE;
 163                 }
 164                 if (record->lock.refs == 0) {
 165                         if (record->flags & HAMMER_RECF_ALLOCDATA) {
 166                                 --hammer_count_record_datas;
 167                                 kfree(record->data, M_HAMMER);
 168                                 record->flags &= ~HAMMER_RECF_ALLOCDATA;
 169                         }
 170                         record->data = NULL;
 171                         --hammer_count_records;
 172                         kfree(record, M_HAMMER);
 173                         return;
 174                 }
 175         }
 176
 177         /*
 178          * If someone wanted the record wake them up.
 179          */
 180         if (record->flags & HAMMER_RECF_WANTED) {
 181                 record->flags &= ~HAMMER_RECF_WANTED;
 182                 wakeup(record);
 183         }
 184 }
 185
 186 /*
 187  * Lookup an in-memory record given the key specified in the cursor.  Works
 188  * just like hammer_btree_lookup() but operates on an inode's in-memory
 189  * record list.
 190  *
 191  * The lookup must fail if the record is marked for deferred deletion.
 192  */
 193 static
 194 int
 195 hammer_mem_lookup(hammer_cursor_t cursor, hammer_inode_t ip)
 196 {
 197         int error;
 198
 199         if (cursor->iprec) {
 200                 hammer_rel_mem_record(cursor->iprec);
 201                 cursor->iprec = NULL;
 202         }
 203         if (cursor->ip) {
 204                 hammer_rec_rb_tree_scan_info_done(&cursor->scan,
 205                                                   &cursor->ip->rec_tree);
 206         }
 207         cursor->ip = ip;
 208         hammer_rec_rb_tree_scan_info_link(&cursor->scan, &ip->rec_tree);
 209         cursor->scan.node = NULL;
 210         cursor->iprec = hammer_rec_rb_tree_RB_LOOKUP_INFO(
 211                                 &ip->rec_tree, &cursor->key_beg);
 212         if (cursor->iprec == NULL) {
 213                 error = ENOENT;
 214         } else {
 215                 hammer_ref(&cursor->iprec->lock);
 216                 error = 0;
 217         }
 218         return(error);
 219 }
 220
 221 /*
 222  * hammer_mem_first() - locate the first in-memory record matching the
 223  * cursor.
 224  *
 225  * The RB_SCAN function we use is designed as a callback.  We terminate it
 226  * (return -1) as soon as we get a match.
 227  */
 228 static
 229 int
 230 hammer_rec_scan_callback(hammer_record_t rec, void *data)
 231 {
 232         hammer_cursor_t cursor = data;
 233
 234         /*
 235          * We terminate on success, so this should be NULL on entry.
 236          */
 237         KKASSERT(cursor->iprec == NULL);
 238
 239         /*
 240          * Skip if the record was marked deleted
 241          */
 242         if (rec->flags & HAMMER_RECF_DELETED)
 243                 return(0);
 244
 245         /*
 246          * Skip if not visible due to our as-of TID
 247          */
 248         if (cursor->flags & HAMMER_CURSOR_ASOF) {
 249                 if (cursor->asof < rec->rec.base.base.create_tid)
 250                         return(0);
 251                 if (rec->rec.base.base.delete_tid &&
 252                     cursor->asof >= rec->rec.base.base.delete_tid) {
 253                         return(0);
 254                 }
 255         }
 256
 257         /*
 258          * Block if currently being synchronized to disk, otherwise we
 259          * may get a duplicate.  Wakeup the syncer if it's stuck on
 260          * the record.
 261          */
 262         hammer_ref(&rec->lock);
 263         ++rec->blocked;
 264         while (rec->flags & HAMMER_RECF_SYNCING) {
 265                 rec->flags |= HAMMER_RECF_WANTED;
 266                 tsleep(rec, 0, "hmrrc2", 0);
 267         }
 268         --rec->blocked;
 269
 270         /*
 271          * The record may have been deleted while we were blocked.
 272          */
 273         if (rec->flags & HAMMER_RECF_DELETED) {
 274                 hammer_rel_mem_record(cursor->iprec);
 275                 return(0);
 276         }
 277
 278         /*
 279          * Set the matching record and stop the scan.
 280          */
 281         cursor->iprec = rec;
 282         return(-1);
 283 }
 284
 285 static
 286 int
 287 hammer_mem_first(hammer_cursor_t cursor, hammer_inode_t ip)
 288 {
 289         if (cursor->iprec) {
 290                 hammer_rel_mem_record(cursor->iprec);
 291                 cursor->iprec = NULL;
 292         }
 293         if (cursor->ip) {
 294                 hammer_rec_rb_tree_scan_info_done(&cursor->scan,
 295                                                   &cursor->ip->rec_tree);
 296         }
 297         cursor->ip = ip;
 298         hammer_rec_rb_tree_scan_info_link(&cursor->scan, &ip->rec_tree);
 299
 300         cursor->scan.node = NULL;
 301         hammer_rec_rb_tree_RB_SCAN(&ip->rec_tree, hammer_rec_scan_cmp,
 302                                    hammer_rec_scan_callback, cursor);
 303
 304         /*
 305          * Adjust scan.node and keep it linked into the RB-tree so we can
 306          * hold the cursor through third party modifications of the RB-tree.
 307          */
 308         if (cursor->iprec) {
 309                 cursor->scan.node = hammer_rec_rb_tree_RB_NEXT(cursor->iprec);
 310                 return(0);
 311         }
 312         return(ENOENT);
 313 }
 314
 315 void
 316 hammer_mem_done(hammer_cursor_t cursor)
 317 {
 318         if (cursor->ip) {
 319                 hammer_rec_rb_tree_scan_info_done(&cursor->scan,
 320                                                   &cursor->ip->rec_tree);
 321                 cursor->ip = NULL;
 322         }
 323         if (cursor->iprec) {
 324                 hammer_rel_mem_record(cursor->iprec);
 325                 cursor->iprec = NULL;
 326         }
 327 }
 328
 329 /************************************************************************
 330  *                   HAMMER IN-MEMORY RECORD FUNCTIONS                  *
 331  ************************************************************************
 332  *
 333  * These functions manipulate in-memory records.  Such records typically
 334  * exist prior to being committed to disk or indexed via the on-disk B-Tree.
 335  */
 336
 337 /*
 338  * Add a directory entry (dip,ncp) which references inode (ip).
 339  *
 340  * Note that the low 32 bits of the namekey are set temporarily to create
 341  * a unique in-memory record, and may be modified a second time when the
 342  * record is synchronized to disk.  In particular, the low 32 bits cannot be
 343  * all 0's when synching to disk, which is not handled here.
 344  */
 345 int
 346 hammer_ip_add_directory(struct hammer_transaction *trans,
 347                      struct hammer_inode *dip, struct namecache *ncp,
 348                      struct hammer_inode *ip)
 349 {
 350         hammer_record_t record;
 351         int error;
 352         int bytes;
 353
 354         record = hammer_alloc_mem_record(dip);
 355
 356         bytes = ncp->nc_nlen;   /* NOTE: terminating \0 is NOT included */
 357         if (++trans->hmp->namekey_iterator == 0)
 358                 ++trans->hmp->namekey_iterator;
 359
 360         record->rec.entry.base.base.obj_id = dip->obj_id;
 361         record->rec.entry.base.base.key =
 362                 hammer_directory_namekey(ncp->nc_name, bytes);
 363         record->rec.entry.base.base.key += trans->hmp->namekey_iterator;
 364         record->rec.entry.base.base.create_tid = trans->tid;
 365         record->rec.entry.base.base.rec_type = HAMMER_RECTYPE_DIRENTRY;
 366         record->rec.entry.base.base.obj_type = ip->ino_rec.base.base.obj_type;
 367         record->rec.entry.obj_id = ip->obj_id;
 368         record->data = (void *)ncp->nc_name;
 369         record->rec.entry.base.data_len = bytes;
 370         ++ip->ino_rec.ino_nlinks;
 371         hammer_modify_inode(trans, ip, HAMMER_INODE_RDIRTY);
 372         /* NOTE: copies record->data */
 373         error = hammer_mem_add(trans, record);
 374         return(error);
 375 }
 376
 377 /*
 378  * Delete the directory entry and update the inode link count.  The
 379  * cursor must be seeked to the directory entry record being deleted.
 380  *
 381  * NOTE: HAMMER_CURSOR_DELETE may not have been set.  XXX remove flag.
 382  *
 383  * This function can return EDEADLK requiring the caller to terminate
 384  * the cursor and retry.
 385  */
 386 int
 387 hammer_ip_del_directory(struct hammer_transaction *trans,
 388                      hammer_cursor_t cursor, struct hammer_inode *dip,
 389                      struct hammer_inode *ip)
 390 {
 391         int error;
 392
 393         error = hammer_ip_delete_record(cursor, trans->tid);
 394
 395         /*
 396          * One less link.  The file may still be open in the OS even after
 397          * all links have gone away so we only try to sync if the OS has
 398          * no references and nlinks falls to 0.
 399          *
 400          * We have to terminate the cursor before syncing the inode to
 401          * avoid deadlocking against ourselves.
 402          */
 403         if (error == 0) {
 404                 --ip->ino_rec.ino_nlinks;
 405                 hammer_modify_inode(trans, ip, HAMMER_INODE_RDIRTY);
 406                 if (ip->ino_rec.ino_nlinks == 0 &&
 407                     (ip->vp == NULL || (ip->vp->v_flag & VINACTIVE))) {
 408                         hammer_done_cursor(cursor);
 409                         hammer_sync_inode(ip, MNT_NOWAIT, 1);
 410                 }
 411
 412         }
 413         return(error);
 414 }
 415
 416 /*
 417  * Add a record to an inode.
 418  *
 419  * The caller must allocate the record with hammer_alloc_mem_record(ip) and
 420  * initialize the following additional fields:
 421  *
 422  * record->rec.entry.base.base.key
 423  * record->rec.entry.base.base.rec_type
 424  * record->rec.entry.base.base.data_len
 425  * record->data         (a copy will be kmalloc'd if it cannot be embedded)
 426  */
 427 int
 428 hammer_ip_add_record(struct hammer_transaction *trans, hammer_record_t record)
 429 {
 430         hammer_inode_t ip = record->ip;
 431         int error;
 432
 433         record->rec.base.base.obj_id = ip->obj_id;
 434         record->rec.base.base.create_tid = trans->tid;
 435         record->rec.base.base.obj_type = ip->ino_rec.base.base.obj_type;
 436
 437         hammer_modify_inode(trans, ip, HAMMER_INODE_RDIRTY);
 438         /* NOTE: copies record->data */
 439         error = hammer_mem_add(trans, record);
 440         return(error);
 441 }
 442
 443 /*
 444  * Sync data from a buffer cache buffer (typically) to the filesystem.  This
 445  * is called via the strategy called from a cached data source.  This code
 446  * is responsible for actually writing a data record out to the disk.
 447  *
 448  * This can only occur non-historically (i.e. 'current' data only).
 449  *
 450  * The file offset must be HAMMER_BUFSIZE aligned but the data length
 451  * can be truncated.  The record (currently) always represents a BUFSIZE
 452  * swath of space whether the data is truncated or not.
 453  */
 454 int
 455 hammer_ip_sync_data(hammer_transaction_t trans, hammer_inode_t ip,
 456                        int64_t offset, void *data, int bytes)
 457 {
 458         struct hammer_cursor cursor;
 459         hammer_record_ondisk_t rec;
 460         union hammer_btree_elm elm;
 461         hammer_off_t rec_offset;
 462         void *bdata;
 463         int error;
 464
 465         KKASSERT((offset & HAMMER_BUFMASK) == 0);
 466 retry:
 467         error = hammer_init_cursor_hmp(&cursor, &ip->cache[0], ip->hmp);
 468         if (error)
 469                 return(error);
 470         cursor.key_beg.obj_id = ip->obj_id;
 471         cursor.key_beg.key = offset + bytes;
 472         cursor.key_beg.create_tid = trans->tid;
 473         cursor.key_beg.delete_tid = 0;
 474         cursor.key_beg.rec_type = HAMMER_RECTYPE_DATA;
 475         cursor.asof = trans->tid;
 476         cursor.flags |= HAMMER_CURSOR_INSERT;
 477
 478         /*
 479          * Issue a lookup to position the cursor.
 480          */
 481         error = hammer_btree_lookup(&cursor);
 482         if (error == 0) {
 483                 kprintf("hammer_ip_sync_data: duplicate data at "
 484                         "(%lld,%d) tid %016llx\n",
 485                         offset, bytes, trans->tid);
 486                 hammer_print_btree_elm(&cursor.node->ondisk->elms[cursor.index],
 487                                        HAMMER_BTREE_TYPE_LEAF, cursor.index);
 488                 panic("Duplicate data");
 489                 error = EIO;
 490         }
 491         if (error != ENOENT)
 492                 goto done;
 493
 494         /*
 495          * Allocate record and data space.  HAMMER_RECTYPE_DATA records
 496          * can cross buffer boundaries so we may have to split our bcopy.
 497          */
 498         rec = hammer_alloc_record(ip->hmp, &rec_offset, HAMMER_RECTYPE_DATA,
 499                                   &cursor.record_buffer,
 500                                   bytes, &bdata,
 501                                   &cursor.data_buffer, &error);
 502         if (rec == NULL)
 503                 goto done;
 504         if (hammer_debug_general & 0x1000)
 505                 kprintf("OOB RECOR2 DATA REC %016llx DATA %016llx LEN=%d\n", rec_offset, rec->base.data_off, rec->base.data_len);
 506
 507         /*
 508          * Fill everything in and insert our B-Tree node.
 509          *
 510          * NOTE: hammer_alloc_record() has already marked the related
 511          * buffers as modified.  If we do it again we will generate
 512          * unnecessary undo elements.
 513          */
 514         rec->base.base.btype = HAMMER_BTREE_TYPE_RECORD;
 515         rec->base.base.obj_id = ip->obj_id;
 516         rec->base.base.key = offset + bytes;
 517         rec->base.base.create_tid = trans->tid;
 518         rec->base.base.delete_tid = 0;
 519         rec->base.base.rec_type = HAMMER_RECTYPE_DATA;
 520         rec->base.data_crc = crc32(data, bytes);
 521         KKASSERT(rec->base.data_len == bytes);
 522
 523         bcopy(data, bdata, bytes);
 524
 525         elm.leaf.base = rec->base.base;
 526         elm.leaf.rec_offset = rec_offset;
 527         elm.leaf.data_offset = rec->base.data_off;
 528         elm.leaf.data_len = bytes;
 529         elm.leaf.data_crc = rec->base.data_crc;
 530
 531         /*
 532          * Data records can wind up on-disk before the inode itself is
 533          * on-disk.  One must assume data records may be on-disk if either
 534          * HAMMER_INODE_DONDISK or HAMMER_INODE_ONDISK is set
 535          */
 536         ip->flags |= HAMMER_INODE_DONDISK;
 537
 538         error = hammer_btree_insert(&cursor, &elm);
 539         if (error == 0)
 540                 goto done;
 541
 542         hammer_blockmap_free(ip->hmp, rec_offset, HAMMER_RECORD_SIZE);
 543 done:
 544         hammer_done_cursor(&cursor);
 545         if (error == EDEADLK)
 546                 goto retry;
 547         return(error);
 548 }
 549
 550 /*
 551  * Sync an in-memory record to the disk.  This is typically called via fsync
 552  * from a cached record source.  This code is responsible for actually
 553  * writing a record out to the disk.
 554  */
 555 int
 556 hammer_ip_sync_record(hammer_record_t record)
 557 {
 558         struct hammer_cursor cursor;
 559         hammer_record_ondisk_t rec;
 560         hammer_mount_t hmp;
 561         union hammer_btree_elm elm;
 562         hammer_off_t rec_offset;
 563         void *bdata;
 564         int error;
 565
 566         hmp = record->ip->hmp;
 567 retry:
 568         /*
 569          * If the record has been deleted or is being synchronized, stop.
 570          * Interlock with the syncing flag.
 571          */
 572         if (record->flags & (HAMMER_RECF_DELETED | HAMMER_RECF_SYNCING))
 573                 return(0);
 574         record->flags |= HAMMER_RECF_SYNCING;
 575
 576         /*
 577          * If someone other then us is referencing the record and not
 578          * blocking waiting for us, we have to wait until they finish.
 579          *
 580          * It is possible the record got destroyed while we were blocked.
 581          */
 582         if (record->lock.refs > record->blocked + 1) {
 583                 while (record->lock.refs > record->blocked + 1) {
 584                         record->flags |= HAMMER_RECF_WANTED;
 585                         tsleep(record, 0, "hmrrc1", 0);
 586                 }
 587                 if (record->flags & HAMMER_RECF_DELETED)
 588                         return(0);
 589         }
 590
 591         /*
 592          * Get a cursor
 593          */
 594         error = hammer_init_cursor_hmp(&cursor, &record->ip->cache[0], hmp);
 595         if (error)
 596                 return(error);
 597         cursor.key_beg = record->rec.base.base;
 598         cursor.flags |= HAMMER_CURSOR_INSERT;
 599
 600         /*
 601          * Issue a lookup to position the cursor and locate the cluster.  The
 602          * target key should not exist.  If we are creating a directory entry
 603          * we may have to iterate the low 32 bits of the key to find an unused
 604          * key.
 605          */
 606         for (;;) {
 607                 error = hammer_btree_lookup(&cursor);
 608                 if (error)
 609                         break;
 610                 if (record->rec.base.base.rec_type != HAMMER_RECTYPE_DIRENTRY) {
 611                         kprintf("hammer_ip_sync_record: duplicate rec "
 612                                 "at (%016llx)\n", record->rec.base.base.key);
 613                         Debugger("duplicate record1");
 614                         error = EIO;
 615                         break;
 616                 }
 617                 if (++hmp->namekey_iterator == 0)
 618                         ++hmp->namekey_iterator;
 619                 record->rec.base.base.key &= ~(0xFFFFFFFFLL);
 620                 record->rec.base.base.key |= hmp->namekey_iterator;
 621                 cursor.key_beg.key = record->rec.base.base.key;
 622         }
 623         if (error != ENOENT)
 624                 goto done;
 625
 626         /*
 627          * Mark the record as undergoing synchronization.  Our cursor is
 628          * holding a locked B-Tree node for the insertion which interlocks
 629          * anyone trying to access this record.
 630          *
 631          * XXX There is still a race present related to iterations.  An
 632          * iteration may process the record, a sync may occur, and then
 633          * later process the B-Tree element for the same record.
 634          *
 635          * We do not try to synchronize a deleted record.
 636          */
 637         if (record->flags & HAMMER_RECF_DELETED) {
 638                 error = 0;
 639                 goto done;
 640         }
 641
 642         /*
 643          * Allocate the record and data.  The result buffers will be
 644          * marked as being modified and further calls to
 645          * hammer_modify_buffer() will result in unneeded UNDO records.
 646          *
 647          * Support zero-fill records (data == NULL and data_len != 0)
 648          */
 649         if (record->data == NULL) {
 650                 rec = hammer_alloc_record(hmp, &rec_offset,
 651                                           record->rec.base.base.rec_type,
 652                                           &cursor.record_buffer,
 653                                           0, &bdata,
 654                                           NULL, &error);
 655                 if (hammer_debug_general & 0x1000)
 656                         kprintf("NULL RECORD DATA\n");
 657         } else if (record->flags & HAMMER_RECF_INBAND) {
 658                 rec = hammer_alloc_record(hmp, &rec_offset,
 659                                           record->rec.base.base.rec_type,
 660                                           &cursor.record_buffer,
 661                                           record->rec.base.data_len, &bdata,
 662                                           NULL, &error);
 663                 if (hammer_debug_general & 0x1000)
 664                         kprintf("INBAND RECORD DATA %016llx DATA %016llx LEN=%d\n", rec_offset, rec->base.data_off, record->rec.base.data_len);
 665         } else {
 666                 rec = hammer_alloc_record(hmp, &rec_offset,
 667                                           record->rec.base.base.rec_type,
 668                                           &cursor.record_buffer,
 669                                           record->rec.base.data_len, &bdata,
 670                                           &cursor.data_buffer, &error);
 671                 if (hammer_debug_general & 0x1000)
 672                         kprintf("OOB RECORD DATA REC %016llx DATA %016llx LEN=%d\n", rec_offset, rec->base.data_off, record->rec.base.data_len);
 673         }
 674
 675         if (rec == NULL)
 676                 goto done;
 677
 678         /*
 679          * Fill in the remaining fields and insert our B-Tree node.
 680          */
 681         rec->base.base = record->rec.base.base;
 682         bcopy(&record->rec.base + 1, &rec->base + 1,
 683               HAMMER_RECORD_SIZE - sizeof(record->rec.base));
 684
 685         /*
 686          * Copy the data and deal with zero-fill support.
 687          */
 688         if (record->data) {
 689                 rec->base.data_crc = crc32(record->data, rec->base.data_len);
 690                 bcopy(record->data, bdata, rec->base.data_len);
 691         } else {
 692                 rec->base.data_len = record->rec.base.data_len;
 693         }
 694
 695         elm.leaf.base = record->rec.base.base;
 696         elm.leaf.rec_offset = rec_offset;
 697         elm.leaf.data_offset = rec->base.data_off;
 698         elm.leaf.data_len = rec->base.data_len;
 699         elm.leaf.data_crc = rec->base.data_crc;
 700
 701         error = hammer_btree_insert(&cursor, &elm);
 702
 703         /*
 704          * Clean up on success, or fall through on error.
 705          */
 706         if (error == 0) {
 707                 record->flags |= HAMMER_RECF_DELETED;
 708                 goto done;
 709         }
 710
 711         /*
 712          * Try to unwind the allocation
 713          */
 714         hammer_blockmap_free(hmp, rec_offset, HAMMER_RECORD_SIZE);
 715 done:
 716         record->flags &= ~HAMMER_RECF_SYNCING;
 717         hammer_done_cursor(&cursor);
 718         if (error == EDEADLK)
 719                 goto retry;
 720         return(error);
 721 }
 722
 723 /*
 724  * Add the record to the inode's rec_tree.  The low 32 bits of a directory
 725  * entry's key is used to deal with hash collisions in the upper 32 bits.
 726  * A unique 64 bit key is generated in-memory and may be regenerated a
 727  * second time when the directory record is flushed to the on-disk B-Tree.
 728  *
 729  * A referenced record is passed to this function.  This function
 730  * eats the reference.  If an error occurs the record will be deleted.
 731  *
 732  * A copy of the temporary record->data pointer provided by the caller
 733  * will be made.
 734  */
 735 static
 736 int
 737 hammer_mem_add(struct hammer_transaction *trans, hammer_record_t record)
 738 {
 739         void *data;
 740         int bytes;
 741         int reclen;
 742
 743         /*
 744          * Make a private copy of record->data
 745          */
 746         if (record->data) {
 747                 /*
 748                  * Try to embed the data in extra space in the record
 749                  * union, otherwise allocate a copy.
 750                  */
 751                 bytes = record->rec.base.data_len;
 752                 switch(record->rec.base.base.rec_type) {
 753                 case HAMMER_RECTYPE_DIRENTRY:
 754                         reclen = offsetof(struct hammer_entry_record, name[0]);
 755                         break;
 756                 case HAMMER_RECTYPE_DATA:
 757                         reclen = offsetof(struct hammer_data_record, data[0]);
 758                         break;
 759                 default:
 760                         reclen = sizeof(record->rec);
 761                         break;
 762                 }
 763                 if (reclen + bytes <= HAMMER_RECORD_SIZE) {
 764                         bcopy(record->data, (char *)&record->rec + reclen,
 765                               bytes);
 766                         record->data = (void *)((char *)&record->rec + reclen);
 767                         record->flags |= HAMMER_RECF_INBAND;
 768                 } else {
 769                         ++hammer_count_record_datas;
 770                         data = kmalloc(bytes, M_HAMMER, M_WAITOK);
 771                         record->flags |= HAMMER_RECF_ALLOCDATA;
 772                         bcopy(record->data, data, bytes);
 773                         record->data = data;
 774                 }
 775         }
 776
 777         /*
 778          * Insert into the RB tree, find an unused iterator if this is
 779          * a directory entry.
 780          */
 781         while (RB_INSERT(hammer_rec_rb_tree, &record->ip->rec_tree, record)) {
 782                 if (record->rec.base.base.rec_type != HAMMER_RECTYPE_DIRENTRY){
 783                         record->flags |= HAMMER_RECF_DELETED;
 784                         hammer_rel_mem_record(record);
 785                         return (EEXIST);
 786                 }
 787                 if (++trans->hmp->namekey_iterator == 0)
 788                         ++trans->hmp->namekey_iterator;
 789                 record->rec.base.base.key &= ~(0xFFFFFFFFLL);
 790                 record->rec.base.base.key |= trans->hmp->namekey_iterator;
 791         }
 792         record->flags |= HAMMER_RECF_ONRBTREE;
 793         hammer_modify_inode(trans, record->ip, HAMMER_INODE_XDIRTY);
 794         hammer_rel_mem_record(record);
 795         return(0);
 796 }
 797
 798 /************************************************************************
 799  *                   HAMMER INODE MERGED-RECORD FUNCTIONS               *
 800  ************************************************************************
 801  *
 802  * These functions augment the B-Tree scanning functions in hammer_btree.c
 803  * by merging in-memory records with on-disk records.
 804  */
 805
 806 /*
 807  * Locate a particular record either in-memory or on-disk.
 808  *
 809  * NOTE: This is basically a standalone routine, hammer_ip_next() may
 810  * NOT be called to iterate results.
 811  */
 812 int
 813 hammer_ip_lookup(hammer_cursor_t cursor, struct hammer_inode *ip)
 814 {
 815         int error;
 816
 817         /*
 818          * If the element is in-memory return it without searching the
 819          * on-disk B-Tree
 820          */
 821         error = hammer_mem_lookup(cursor, ip);
 822         if (error == 0) {
 823                 cursor->record = &cursor->iprec->rec;
 824                 return(error);
 825         }
 826         if (error != ENOENT)
 827                 return(error);
 828
 829         /*
 830          * If the inode has on-disk components search the on-disk B-Tree.
 831          */
 832         if ((ip->flags & (HAMMER_INODE_ONDISK|HAMMER_INODE_DONDISK)) == 0)
 833                 return(error);
 834         error = hammer_btree_lookup(cursor);
 835         if (error == 0)
 836                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_RECORD);
 837         return(error);
 838 }
 839
 840 /*
 841  * Locate the first record within the cursor's key_beg/key_end range,
 842  * restricted to a particular inode.  0 is returned on success, ENOENT
 843  * if no records matched the requested range, or some other error.
 844  *
 845  * When 0 is returned hammer_ip_next() may be used to iterate additional
 846  * records within the requested range.
 847  *
 848  * This function can return EDEADLK, requiring the caller to terminate
 849  * the cursor and try again.
 850  */
 851 int
 852 hammer_ip_first(hammer_cursor_t cursor, struct hammer_inode *ip)
 853 {
 854         int error;
 855
 856         /*
 857          * Clean up fields and setup for merged scan
 858          */
 859         cursor->flags &= ~HAMMER_CURSOR_DELBTREE;
 860         cursor->flags |= HAMMER_CURSOR_ATEDISK | HAMMER_CURSOR_ATEMEM;
 861         cursor->flags |= HAMMER_CURSOR_DISKEOF | HAMMER_CURSOR_MEMEOF;
 862         if (cursor->iprec) {
 863                 hammer_rel_mem_record(cursor->iprec);
 864                 cursor->iprec = NULL;
 865         }
 866
 867         /*
 868          * Search the on-disk B-Tree.  hammer_btree_lookup() only does an
 869          * exact lookup so if we get ENOENT we have to call the iterate
 870          * function to validate the first record after the begin key.
 871          *
 872          * The ATEDISK flag is used by hammer_btree_iterate to determine
 873          * whether it must index forwards or not.  It is also used here
 874          * to select the next record from in-memory or on-disk.
 875          *
 876          * EDEADLK can only occur if the lookup hit an empty internal
 877          * element and couldn't delete it.  Since this could only occur
 878          * in-range, we can just iterate from the failure point.
 879          */
 880         if (ip->flags & (HAMMER_INODE_ONDISK|HAMMER_INODE_DONDISK)) {
 881                 error = hammer_btree_lookup(cursor);
 882                 if (error == ENOENT || error == EDEADLK) {
 883                         cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
 884                         error = hammer_btree_iterate(cursor);
 885                 }
 886                 if (error && error != ENOENT)
 887                         return(error);
 888                 if (error == 0) {
 889                         cursor->flags &= ~HAMMER_CURSOR_DISKEOF;
 890                         cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
 891                 } else {
 892                         cursor->flags |= HAMMER_CURSOR_ATEDISK;
 893                 }
 894         }
 895
 896         /*
 897          * Search the in-memory record list (Red-Black tree).  Unlike the
 898          * B-Tree search, mem_first checks for records in the range.
 899          */
 900         error = hammer_mem_first(cursor, ip);
 901         if (error && error != ENOENT)
 902                 return(error);
 903         if (error == 0) {
 904                 cursor->flags &= ~HAMMER_CURSOR_MEMEOF;
 905                 cursor->flags &= ~HAMMER_CURSOR_ATEMEM;
 906         }
 907
 908         /*
 909          * This will return the first matching record.
 910          */
 911         return(hammer_ip_next(cursor));
 912 }
 913
 914 /*
 915  * Retrieve the next record in a merged iteration within the bounds of the
 916  * cursor.  This call may be made multiple times after the cursor has been
 917  * initially searched with hammer_ip_first().
 918  *
 919  * 0 is returned on success, ENOENT if no further records match the
 920  * requested range, or some other error code is returned.
 921  */
 922 int
 923 hammer_ip_next(hammer_cursor_t cursor)
 924 {
 925         hammer_btree_elm_t elm;
 926         hammer_record_t rec;
 927         int error;
 928         int r;
 929
 930         /*
 931          * Load the current on-disk and in-memory record.  If we ate any
 932          * records we have to get the next one.
 933          *
 934          * If we deleted the last on-disk record we had scanned ATEDISK will
 935          * be clear and DELBTREE will be set, forcing a call to iterate. The
 936          * fact that ATEDISK is clear causes iterate to re-test the 'current'
 937          * element.  If ATEDISK is set, iterate will skip the 'current'
 938          * element.
 939          *
 940          * Get the next on-disk record
 941          */
 942         if (cursor->flags & (HAMMER_CURSOR_ATEDISK|HAMMER_CURSOR_DELBTREE)) {
 943                 if ((cursor->flags & HAMMER_CURSOR_DISKEOF) == 0) {
 944                         error = hammer_btree_iterate(cursor);
 945                         cursor->flags &= ~HAMMER_CURSOR_DELBTREE;
 946                         if (error == 0)
 947                                 cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
 948                         else
 949                                 cursor->flags |= HAMMER_CURSOR_DISKEOF |
 950                                                  HAMMER_CURSOR_ATEDISK;
 951                 }
 952         }
 953
 954         /*
 955          * Get the next in-memory record.  The record can be ripped out
 956          * of the RB tree so we maintain a scan_info structure to track
 957          * the next node.
 958          *
 959          * hammer_rec_scan_cmp:  Is the record still in our general range,
 960          *                       (non-inclusive of snapshot exclusions)?
 961          * hammer_rec_scan_callback: Is the record in our snapshot?
 962          */
 963         if (cursor->flags & HAMMER_CURSOR_ATEMEM) {
 964                 if ((cursor->flags & HAMMER_CURSOR_MEMEOF) == 0) {
 965                         if (cursor->iprec) {
 966                                 hammer_rel_mem_record(cursor->iprec);
 967                                 cursor->iprec = NULL;
 968                         }
 969                         rec = cursor->scan.node;        /* next node */
 970                         while (rec) {
 971                                 if (hammer_rec_scan_cmp(rec, cursor) != 0)
 972                                         break;
 973                                 if (hammer_rec_scan_callback(rec, cursor) != 0)
 974                                         break;
 975                                 rec = hammer_rec_rb_tree_RB_NEXT(rec);
 976                         }
 977                         if (cursor->iprec) {
 978                                 KKASSERT(cursor->iprec == rec);
 979                                 cursor->flags &= ~HAMMER_CURSOR_ATEMEM;
 980                                 cursor->scan.node =
 981                                         hammer_rec_rb_tree_RB_NEXT(rec);
 982                         } else {
 983                                 cursor->flags |= HAMMER_CURSOR_MEMEOF;
 984                         }
 985                 }
 986         }
 987
 988         /*
 989          * Extract either the disk or memory record depending on their
 990          * relative position.
 991          */
 992         error = 0;
 993         switch(cursor->flags & (HAMMER_CURSOR_ATEDISK | HAMMER_CURSOR_ATEMEM)) {
 994         case 0:
 995                 /*
 996                  * Both entries valid
 997                  */
 998                 elm = &cursor->node->ondisk->elms[cursor->index];
 999                 r = hammer_btree_cmp(&elm->base, &cursor->iprec->rec.base.base);
1000                 if (r < 0) {
1001                         error = hammer_btree_extract(cursor,
1002                                                      HAMMER_CURSOR_GET_RECORD);
1003                         cursor->flags |= HAMMER_CURSOR_ATEDISK;
1004                         break;
1005                 }
1006                 /* fall through to the memory entry */
1007         case HAMMER_CURSOR_ATEDISK:
1008                 /*
1009                  * Only the memory entry is valid
1010                  */
1011                 cursor->record = &cursor->iprec->rec;
1012                 cursor->flags |= HAMMER_CURSOR_ATEMEM;
1013                 break;
1014         case HAMMER_CURSOR_ATEMEM:
1015                 /*
1016                  * Only the disk entry is valid
1017                  */
1018                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_RECORD);
1019                 cursor->flags |= HAMMER_CURSOR_ATEDISK;
1020                 break;
1021         default:
1022                 /*
1023                  * Neither entry is valid
1024                  *
1025                  * XXX error not set properly
1026                  */
1027                 cursor->record = NULL;
1028                 error = ENOENT;
1029                 break;
1030         }
1031         return(error);
1032 }
1033
1034 /*
1035  * Resolve the cursor->data pointer for the current cursor position in
1036  * a merged iteration.
1037  */
1038 int
1039 hammer_ip_resolve_data(hammer_cursor_t cursor)
1040 {
1041         int error;
1042
1043         if (cursor->iprec && cursor->record == &cursor->iprec->rec) {
1044                 cursor->data = cursor->iprec->data;
1045                 error = 0;
1046         } else {
1047                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_DATA);
1048         }
1049         return(error);
1050 }
1051
1052 int
1053 hammer_ip_resolve_record_and_data(hammer_cursor_t cursor)
1054 {
1055         int error;
1056
1057         if (cursor->iprec && cursor->record == &cursor->iprec->rec) {
1058                 cursor->data = cursor->iprec->data;
1059                 error = 0;
1060         } else {
1061                 error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_DATA |
1062                                                      HAMMER_CURSOR_GET_RECORD);
1063         }
1064         return(error);
1065 }
1066
1067 /*
1068  * Delete all records within the specified range for inode ip.
1069  *
1070  * NOTE: An unaligned range will cause new records to be added to cover
1071  * the edge cases. (XXX not implemented yet).
1072  *
1073  * NOTE: ran_end is inclusive (e.g. 0,1023 instead of 0,1024).
1074  *
1075  * NOTE: Record keys for regular file data have to be special-cased since
1076  * they indicate the end of the range (key = base + bytes).
1077  */
1078 int
1079 hammer_ip_delete_range(hammer_transaction_t trans, hammer_inode_t ip,
1080                        int64_t ran_beg, int64_t ran_end)
1081 {
1082         struct hammer_cursor cursor;
1083         hammer_record_ondisk_t rec;
1084         hammer_base_elm_t base;
1085         int error;
1086         int64_t off;
1087
1088 retry:
1089         hammer_init_cursor_hmp(&cursor, &ip->cache[0], ip->hmp);
1090
1091         cursor.key_beg.obj_id = ip->obj_id;
1092         cursor.key_beg.create_tid = 0;
1093         cursor.key_beg.delete_tid = 0;
1094         cursor.key_beg.obj_type = 0;
1095         cursor.asof = ip->obj_asof;
1096         cursor.flags |= HAMMER_CURSOR_ASOF;
1097
1098         cursor.key_end = cursor.key_beg;
1099         if (ip->ino_rec.base.base.obj_type == HAMMER_OBJTYPE_DBFILE) {
1100                 cursor.key_beg.key = ran_beg;
1101                 cursor.key_beg.rec_type = HAMMER_RECTYPE_DB;
1102                 cursor.key_end.rec_type = HAMMER_RECTYPE_DB;
1103                 cursor.key_end.key = ran_end;
1104         } else {
1105                 /*
1106                  * The key in the B-Tree is (base+bytes), so the first possible
1107                  * matching key is ran_beg + 1.
1108                  */
1109                 int64_t tmp64;
1110
1111                 cursor.key_beg.key = ran_beg + 1;
1112                 cursor.key_beg.rec_type = HAMMER_RECTYPE_DATA;
1113                 cursor.key_end.rec_type = HAMMER_RECTYPE_DATA;
1114
1115                 tmp64 = ran_end + MAXPHYS + 1;  /* work around GCC-4 bug */
1116                 if (tmp64 < ran_end)
1117                         cursor.key_end.key = 0x7FFFFFFFFFFFFFFFLL;
1118                 else
1119                         cursor.key_end.key = ran_end + MAXPHYS + 1;
1120         }
1121         cursor.flags |= HAMMER_CURSOR_END_INCLUSIVE;
1122
1123         error = hammer_ip_first(&cursor, ip);
1124
1125         /*
1126          * Iterate through matching records and mark them as deleted.
1127          */
1128         while (error == 0) {
1129                 rec = cursor.record;
1130                 base = &rec->base.base;
1131
1132                 KKASSERT(base->delete_tid == 0);
1133
1134                 /*
1135                  * There may be overlap cases for regular file data.  Also
1136                  * remember the key for a regular file record is the offset
1137                  * of the last byte of the record (base + len - 1), NOT the
1138                  * base offset.
1139                  */
1140 #if 0
1141                 kprintf("delete_range rec_type %02x\n", base->rec_type);
1142 #endif
1143                 if (base->rec_type == HAMMER_RECTYPE_DATA) {
1144 #if 0
1145                         kprintf("delete_range loop key %016llx\n",
1146                                 base->key - rec->base.data_len);
1147 #endif
1148                         off = base->key - rec->base.data_len;
1149                         /*
1150                          * Check the left edge case.  We currently do not
1151                          * split existing records.
1152                          */
1153                         if (off < ran_beg) {
1154                                 panic("hammer left edge case %016llx %d\n",
1155                                         base->key, rec->base.data_len);
1156                         }
1157
1158                         /*
1159                          * Check the right edge case.  Note that the
1160                          * record can be completely out of bounds, which
1161                          * terminates the search.
1162                          *
1163                          * base->key is exclusive of the right edge while
1164                          * ran_end is inclusive of the right edge.  The
1165                          * (key - data_len) left boundary is inclusive.
1166                          *
1167                          * XXX theory-check this test at some point, are
1168                          * we missing a + 1 somewhere?  Note that ran_end
1169                          * could overflow.
1170                          */
1171                         if (base->key - 1 > ran_end) {
1172                                 if (base->key - rec->base.data_len > ran_end)
1173                                         break;
1174                                 panic("hammer right edge case\n");
1175                         }
1176                 }
1177
1178                 /*
1179                  * Mark the record and B-Tree entry as deleted.  This will
1180                  * also physically delete the B-Tree entry, record, and
1181                  * data if the retention policy dictates.  The function
1182                  * will set HAMMER_CURSOR_DELBTREE which hammer_ip_next()
1183                  * uses to perform a fixup.
1184                  */
1185                 error = hammer_ip_delete_record(&cursor, trans->tid);
1186                 if (error)
1187                         break;
1188                 error = hammer_ip_next(&cursor);
1189         }
1190         hammer_done_cursor(&cursor);
1191         if (error == EDEADLK)
1192                 goto retry;
1193         if (error == ENOENT)
1194                 error = 0;
1195         return(error);
1196 }
1197
1198 /*
1199  * Delete all records associated with an inode except the inode record
1200  * itself.
1201  */
1202 int
1203 hammer_ip_delete_range_all(hammer_transaction_t trans, hammer_inode_t ip)
1204 {
1205         struct hammer_cursor cursor;
1206         hammer_record_ondisk_t rec;
1207         hammer_base_elm_t base;
1208         int error;
1209
1210 retry:
1211         hammer_init_cursor_hmp(&cursor, &ip->cache[0], ip->hmp);
1212
1213         cursor.key_beg.obj_id = ip->obj_id;
1214         cursor.key_beg.create_tid = 0;
1215         cursor.key_beg.delete_tid = 0;
1216         cursor.key_beg.obj_type = 0;
1217         cursor.key_beg.rec_type = HAMMER_RECTYPE_INODE + 1;
1218         cursor.key_beg.key = HAMMER_MIN_KEY;
1219
1220         cursor.key_end = cursor.key_beg;
1221         cursor.key_end.rec_type = 0xFFFF;
1222         cursor.key_end.key = HAMMER_MAX_KEY;
1223
1224         cursor.asof = ip->obj_asof;
1225         cursor.flags |= HAMMER_CURSOR_END_INCLUSIVE | HAMMER_CURSOR_ASOF;
1226
1227         error = hammer_ip_first(&cursor, ip);
1228
1229         /*
1230          * Iterate through matching records and mark them as deleted.
1231          */
1232         while (error == 0) {
1233                 rec = cursor.record;
1234                 base = &rec->base.base;
1235
1236                 KKASSERT(base->delete_tid == 0);
1237
1238                 /*
1239                  * Mark the record and B-Tree entry as deleted.  This will
1240                  * also physically delete the B-Tree entry, record, and
1241                  * data if the retention policy dictates.  The function
1242                  * will set HAMMER_CURSOR_DELBTREE which hammer_ip_next()
1243                  * uses to perform a fixup.
1244                  */
1245                 error = hammer_ip_delete_record(&cursor, trans->tid);
1246                 if (error)
1247                         break;
1248                 error = hammer_ip_next(&cursor);
1249         }
1250         hammer_done_cursor(&cursor);
1251         if (error == EDEADLK)
1252                 goto retry;
1253         if (error == ENOENT)
1254                 error = 0;
1255         return(error);
1256 }
1257
1258 /*
1259  * Delete the record at the current cursor.  On success the cursor will
1260  * be positioned appropriately for an iteration but may no longer be at
1261  * a leaf node.
1262  *
1263  * NOTE: This can return EDEADLK, requiring the caller to terminate the
1264  * cursor and retry.
1265  */
1266 int
1267 hammer_ip_delete_record(hammer_cursor_t cursor, hammer_tid_t tid)
1268 {
1269         hammer_btree_elm_t elm;
1270         hammer_mount_t hmp;
1271         int error;
1272         int dodelete;
1273
1274         /*
1275          * In-memory (unsynchronized) records can simply be freed.
1276          */
1277         if (cursor->record == &cursor->iprec->rec) {
1278                 cursor->iprec->flags |= HAMMER_RECF_DELETED;
1279                 return(0);
1280         }
1281
1282         /*
1283          * On-disk records are marked as deleted by updating their delete_tid.
1284          * This does not effect their position in the B-Tree (which is based
1285          * on their create_tid).
1286          */
1287         error = hammer_btree_extract(cursor, HAMMER_CURSOR_GET_RECORD);
1288         elm = NULL;
1289         hmp = cursor->node->hmp;
1290
1291         dodelete = 0;
1292         if (error == 0) {
1293                 error = hammer_cursor_upgrade(cursor);
1294                 if (error == 0) {
1295                         hammer_modify_node(cursor->node);
1296                         elm = &cursor->node->ondisk->elms[cursor->index];
1297                         elm->leaf.base.delete_tid = tid;
1298                         hammer_modify_buffer(cursor->record_buffer, &cursor->record->base.base.delete_tid, sizeof(hammer_tid_t));
1299                         cursor->record->base.base.delete_tid = tid;
1300                 }
1301         }
1302
1303         /*
1304          * If we were mounted with the nohistory option, we physically
1305          * delete the record.
1306          */
1307         if (hmp->hflags & HMNT_NOHISTORY)
1308                 dodelete = 1;
1309
1310         if (error == 0 && dodelete) {
1311                 error = hammer_delete_at_cursor(cursor, NULL);
1312                 if (error) {
1313                         panic("hammer_ip_delete_record: unable to physically delete the record!\n");
1314                         error = 0;
1315                 }
1316         }
1317         return(error);
1318 }
1319
1320 int
1321 hammer_delete_at_cursor(hammer_cursor_t cursor, int64_t *stat_bytes)
1322 {
1323         hammer_btree_elm_t elm;
1324         hammer_off_t rec_offset;
1325         hammer_off_t data_offset;
1326         int32_t data_len;
1327         u_int8_t rec_type;
1328         int error;
1329
1330         elm = &cursor->node->ondisk->elms[cursor->index];
1331         KKASSERT(elm->base.btype == HAMMER_BTREE_TYPE_RECORD);
1332
1333         rec_offset = elm->leaf.rec_offset;
1334         data_offset = elm->leaf.data_offset;
1335         data_len = elm->leaf.data_len;
1336         rec_type = elm->leaf.base.rec_type;
1337
1338         error = hammer_btree_delete(cursor);
1339         if (error == 0) {
1340                 /*
1341                  * This forces a fixup for the iteration because
1342                  * the cursor is now either sitting at the 'next'
1343                  * element or sitting at the end of a leaf.
1344                  */
1345                 if ((cursor->flags & HAMMER_CURSOR_DISKEOF) == 0) {
1346                         cursor->flags |= HAMMER_CURSOR_DELBTREE;
1347                         cursor->flags &= ~HAMMER_CURSOR_ATEDISK;
1348                 }
1349         }
1350         if (error == 0) {
1351                 hammer_blockmap_free(cursor->node->hmp, rec_offset,
1352                                      sizeof(union hammer_record_ondisk));
1353         }
1354         if (error == 0) {
1355                 switch(data_offset & HAMMER_OFF_ZONE_MASK) {
1356                 case HAMMER_ZONE_LARGE_DATA:
1357                 case HAMMER_ZONE_SMALL_DATA:
1358                         hammer_blockmap_free(cursor->node->hmp,
1359                                              data_offset, data_len);
1360                         break;
1361                 default:
1362                         break;
1363                 }
1364         }
1365 #if 0
1366         kprintf("hammer_delete_at_cursor: %d:%d:%08x %08x/%d "
1367                 "(%d remain in cluster)\n",
1368                 cluster->volume->vol_no, cluster->clu_no,
1369                 rec_offset, data_offset, data_len,
1370                 cluster->ondisk->stat_records);
1371 #endif
1372         return (error);
1373 }
1374
1375 /*
1376  * Determine whether a directory is empty or not.  Returns 0 if the directory
1377  * is empty, ENOTEMPTY if it isn't, plus other possible errors.
1378  */
1379 int
1380 hammer_ip_check_directory_empty(hammer_transaction_t trans, hammer_inode_t ip)
1381 {
1382         struct hammer_cursor cursor;
1383         int error;
1384
1385         hammer_init_cursor_hmp(&cursor, &ip->cache[0], ip->hmp);
1386
1387         cursor.key_beg.obj_id = ip->obj_id;
1388         cursor.key_beg.create_tid = 0;
1389         cursor.key_beg.delete_tid = 0;
1390         cursor.key_beg.obj_type = 0;
1391         cursor.key_beg.rec_type = HAMMER_RECTYPE_INODE + 1;
1392         cursor.key_beg.key = HAMMER_MIN_KEY;
1393
1394         cursor.key_end = cursor.key_beg;
1395         cursor.key_end.rec_type = 0xFFFF;
1396         cursor.key_end.key = HAMMER_MAX_KEY;
1397
1398         cursor.asof = ip->obj_asof;
1399         cursor.flags |= HAMMER_CURSOR_END_INCLUSIVE | HAMMER_CURSOR_ASOF;
1400
1401         error = hammer_ip_first(&cursor, ip);
1402         if (error == ENOENT)
1403                 error = 0;
1404         else if (error == 0)
1405                 error = ENOTEMPTY;
1406         hammer_done_cursor(&cursor);
1407         return(error);
1408 }
1409