src/journal/sd-journal.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2
   3 #include <errno.h>
   4 #include <fcntl.h>
   5 #include <inttypes.h>
   6 #include <linux/magic.h>
   7 #include <poll.h>
   8 #include <stddef.h>
   9 #include <sys/inotify.h>
  10 #include <sys/vfs.h>
  11 #include <unistd.h>
  12
  13 #include "sd-journal.h"
  14
  15 #include "alloc-util.h"
  16 #include "catalog.h"
  17 #include "compress.h"
  18 #include "dirent-util.h"
  19 #include "env-file.h"
  20 #include "escape.h"
  21 #include "fd-util.h"
  22 #include "fileio.h"
  23 #include "format-util.h"
  24 #include "fs-util.h"
  25 #include "hashmap.h"
  26 #include "hostname-util.h"
  27 #include "id128-util.h"
  28 #include "io-util.h"
  29 #include "journal-def.h"
  30 #include "journal-file.h"
  31 #include "journal-internal.h"
  32 #include "list.h"
  33 #include "lookup3.h"
  34 #include "missing.h"
  35 #include "nulstr-util.h"
  36 #include "path-util.h"
  37 #include "process-util.h"
  38 #include "replace-var.h"
  39 #include "stat-util.h"
  40 #include "stdio-util.h"
  41 #include "string-util.h"
  42 #include "strv.h"
  43
  44 #define JOURNAL_FILES_MAX 7168
  45
  46 #define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC)
  47
  48 #define REPLACE_VAR_MAX 256
  49
  50 #define DEFAULT_DATA_THRESHOLD (64*1024)
  51
  52 static void remove_file_real(sd_journal *j, JournalFile *f);
  53
  54 static bool journal_pid_changed(sd_journal *j) {
  55         assert(j);
  56
  57         /* We don't support people creating a journal object and
  58          * keeping it around over a fork(). Let's complain. */
  59
  60         return j->original_pid != getpid_cached();
  61 }
  62
  63 static int journal_put_error(sd_journal *j, int r, const char *path) {
  64         char *copy;
  65         int k;
  66
  67         /* Memorize an error we encountered, and store which
  68          * file/directory it was generated from. Note that we store
  69          * only *one* path per error code, as the error code is the
  70          * key into the hashmap, and the path is the value. This means
  71          * we keep track only of all error kinds, but not of all error
  72          * locations. This has the benefit that the hashmap cannot
  73          * grow beyond bounds.
  74          *
  75          * We return an error here only if we didn't manage to
  76          * memorize the real error. */
  77
  78         if (r >= 0)
  79                 return r;
  80
  81         k = hashmap_ensure_allocated(&j->errors, NULL);
  82         if (k < 0)
  83                 return k;
  84
  85         if (path) {
  86                 copy = strdup(path);
  87                 if (!copy)
  88                         return -ENOMEM;
  89         } else
  90                 copy = NULL;
  91
  92         k = hashmap_put(j->errors, INT_TO_PTR(r), copy);
  93         if (k < 0) {
  94                 free(copy);
  95
  96                 if (k == -EEXIST)
  97                         return 0;
  98
  99                 return k;
 100         }
 101
 102         return 0;
 103 }
 104
 105 static void detach_location(sd_journal *j) {
 106         Iterator i;
 107         JournalFile *f;
 108
 109         assert(j);
 110
 111         j->current_file = NULL;
 112         j->current_field = 0;
 113
 114         ORDERED_HASHMAP_FOREACH(f, j->files, i)
 115                 journal_file_reset_location(f);
 116 }
 117
 118 static void reset_location(sd_journal *j) {
 119         assert(j);
 120
 121         detach_location(j);
 122         zero(j->current_location);
 123 }
 124
 125 static void init_location(Location *l, LocationType type, JournalFile *f, Object *o) {
 126         assert(l);
 127         assert(IN_SET(type, LOCATION_DISCRETE, LOCATION_SEEK));
 128         assert(f);
 129         assert(o->object.type == OBJECT_ENTRY);
 130
 131         l->type = type;
 132         l->seqnum = le64toh(o->entry.seqnum);
 133         l->seqnum_id = f->header->seqnum_id;
 134         l->realtime = le64toh(o->entry.realtime);
 135         l->monotonic = le64toh(o->entry.monotonic);
 136         l->boot_id = o->entry.boot_id;
 137         l->xor_hash = le64toh(o->entry.xor_hash);
 138
 139         l->seqnum_set = l->realtime_set = l->monotonic_set = l->xor_hash_set = true;
 140 }
 141
 142 static void set_location(sd_journal *j, JournalFile *f, Object *o) {
 143         assert(j);
 144         assert(f);
 145         assert(o);
 146
 147         init_location(&j->current_location, LOCATION_DISCRETE, f, o);
 148
 149         j->current_file = f;
 150         j->current_field = 0;
 151
 152         /* Let f know its candidate entry was picked. */
 153         assert(f->location_type == LOCATION_SEEK);
 154         f->location_type = LOCATION_DISCRETE;
 155 }
 156
 157 static int match_is_valid(const void *data, size_t size) {
 158         const char *b, *p;
 159
 160         assert(data);
 161
 162         if (size < 2)
 163                 return false;
 164
 165         if (startswith(data, "__"))
 166                 return false;
 167
 168         b = data;
 169         for (p = b; p < b + size; p++) {
 170
 171                 if (*p == '=')
 172                         return p > b;
 173
 174                 if (*p == '_')
 175                         continue;
 176
 177                 if (*p >= 'A' && *p <= 'Z')
 178                         continue;
 179
 180                 if (*p >= '0' && *p <= '9')
 181                         continue;
 182
 183                 return false;
 184         }
 185
 186         return false;
 187 }
 188
 189 static bool same_field(const void *_a, size_t s, const void *_b, size_t t) {
 190         const uint8_t *a = _a, *b = _b;
 191         size_t j;
 192
 193         for (j = 0; j < s && j < t; j++) {
 194
 195                 if (a[j] != b[j])
 196                         return false;
 197
 198                 if (a[j] == '=')
 199                         return true;
 200         }
 201
 202         assert_not_reached("\"=\" not found");
 203 }
 204
 205 static Match *match_new(Match *p, MatchType t) {
 206         Match *m;
 207
 208         m = new0(Match, 1);
 209         if (!m)
 210                 return NULL;
 211
 212         m->type = t;
 213
 214         if (p) {
 215                 m->parent = p;
 216                 LIST_PREPEND(matches, p->matches, m);
 217         }
 218
 219         return m;
 220 }
 221
 222 static void match_free(Match *m) {
 223         assert(m);
 224
 225         while (m->matches)
 226                 match_free(m->matches);
 227
 228         if (m->parent)
 229                 LIST_REMOVE(matches, m->parent->matches, m);
 230
 231         free(m->data);
 232         free(m);
 233 }
 234
 235 static void match_free_if_empty(Match *m) {
 236         if (!m || m->matches)
 237                 return;
 238
 239         match_free(m);
 240 }
 241
 242 _public_ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
 243         Match *l3, *l4, *add_here = NULL, *m;
 244         le64_t le_hash;
 245
 246         assert_return(j, -EINVAL);
 247         assert_return(!journal_pid_changed(j), -ECHILD);
 248         assert_return(data, -EINVAL);
 249
 250         if (size == 0)
 251                 size = strlen(data);
 252
 253         assert_return(match_is_valid(data, size), -EINVAL);
 254
 255         /* level 0: AND term
 256          * level 1: OR terms
 257          * level 2: AND terms
 258          * level 3: OR terms
 259          * level 4: concrete matches */
 260
 261         if (!j->level0) {
 262                 j->level0 = match_new(NULL, MATCH_AND_TERM);
 263                 if (!j->level0)
 264                         return -ENOMEM;
 265         }
 266
 267         if (!j->level1) {
 268                 j->level1 = match_new(j->level0, MATCH_OR_TERM);
 269                 if (!j->level1)
 270                         return -ENOMEM;
 271         }
 272
 273         if (!j->level2) {
 274                 j->level2 = match_new(j->level1, MATCH_AND_TERM);
 275                 if (!j->level2)
 276                         return -ENOMEM;
 277         }
 278
 279         assert(j->level0->type == MATCH_AND_TERM);
 280         assert(j->level1->type == MATCH_OR_TERM);
 281         assert(j->level2->type == MATCH_AND_TERM);
 282
 283         le_hash = htole64(hash64(data, size));
 284
 285         LIST_FOREACH(matches, l3, j->level2->matches) {
 286                 assert(l3->type == MATCH_OR_TERM);
 287
 288                 LIST_FOREACH(matches, l4, l3->matches) {
 289                         assert(l4->type == MATCH_DISCRETE);
 290
 291                         /* Exactly the same match already? Then ignore
 292                          * this addition */
 293                         if (l4->le_hash == le_hash &&
 294                             l4->size == size &&
 295                             memcmp(l4->data, data, size) == 0)
 296                                 return 0;
 297
 298                         /* Same field? Then let's add this to this OR term */
 299                         if (same_field(data, size, l4->data, l4->size)) {
 300                                 add_here = l3;
 301                                 break;
 302                         }
 303                 }
 304
 305                 if (add_here)
 306                         break;
 307         }
 308
 309         if (!add_here) {
 310                 add_here = match_new(j->level2, MATCH_OR_TERM);
 311                 if (!add_here)
 312                         goto fail;
 313         }
 314
 315         m = match_new(add_here, MATCH_DISCRETE);
 316         if (!m)
 317                 goto fail;
 318
 319         m->le_hash = le_hash;
 320         m->size = size;
 321         m->data = memdup(data, size);
 322         if (!m->data)
 323                 goto fail;
 324
 325         detach_location(j);
 326
 327         return 0;
 328
 329 fail:
 330         match_free_if_empty(add_here);
 331         match_free_if_empty(j->level2);
 332         match_free_if_empty(j->level1);
 333         match_free_if_empty(j->level0);
 334
 335         return -ENOMEM;
 336 }
 337
 338 _public_ int sd_journal_add_conjunction(sd_journal *j) {
 339         assert_return(j, -EINVAL);
 340         assert_return(!journal_pid_changed(j), -ECHILD);
 341
 342         if (!j->level0)
 343                 return 0;
 344
 345         if (!j->level1)
 346                 return 0;
 347
 348         if (!j->level1->matches)
 349                 return 0;
 350
 351         j->level1 = NULL;
 352         j->level2 = NULL;
 353
 354         return 0;
 355 }
 356
 357 _public_ int sd_journal_add_disjunction(sd_journal *j) {
 358         assert_return(j, -EINVAL);
 359         assert_return(!journal_pid_changed(j), -ECHILD);
 360
 361         if (!j->level0)
 362                 return 0;
 363
 364         if (!j->level1)
 365                 return 0;
 366
 367         if (!j->level2)
 368                 return 0;
 369
 370         if (!j->level2->matches)
 371                 return 0;
 372
 373         j->level2 = NULL;
 374         return 0;
 375 }
 376
 377 static char *match_make_string(Match *m) {
 378         char *p = NULL, *r;
 379         Match *i;
 380         bool enclose = false;
 381
 382         if (!m)
 383                 return strdup("none");
 384
 385         if (m->type == MATCH_DISCRETE)
 386                 return cescape_length(m->data, m->size);
 387
 388         LIST_FOREACH(matches, i, m->matches) {
 389                 char *t, *k;
 390
 391                 t = match_make_string(i);
 392                 if (!t)
 393                         return mfree(p);
 394
 395                 if (p) {
 396                         k = strjoin(p, m->type == MATCH_OR_TERM ? " OR " : " AND ", t);
 397                         free(p);
 398                         free(t);
 399
 400                         if (!k)
 401                                 return NULL;
 402
 403                         p = k;
 404
 405                         enclose = true;
 406                 } else
 407                         p = t;
 408         }
 409
 410         if (enclose) {
 411                 r = strjoin("(", p, ")");
 412                 free(p);
 413                 return r;
 414         }
 415
 416         return p;
 417 }
 418
 419 char *journal_make_match_string(sd_journal *j) {
 420         assert(j);
 421
 422         return match_make_string(j->level0);
 423 }
 424
 425 _public_ void sd_journal_flush_matches(sd_journal *j) {
 426         if (!j)
 427                 return;
 428
 429         if (j->level0)
 430                 match_free(j->level0);
 431
 432         j->level0 = j->level1 = j->level2 = NULL;
 433
 434         detach_location(j);
 435 }
 436
 437 _pure_ static int compare_with_location(JournalFile *f, Location *l) {
 438         int r;
 439
 440         assert(f);
 441         assert(l);
 442         assert(f->location_type == LOCATION_SEEK);
 443         assert(IN_SET(l->type, LOCATION_DISCRETE, LOCATION_SEEK));
 444
 445         if (l->monotonic_set &&
 446             sd_id128_equal(f->current_boot_id, l->boot_id) &&
 447             l->realtime_set &&
 448             f->current_realtime == l->realtime &&
 449             l->xor_hash_set &&
 450             f->current_xor_hash == l->xor_hash)
 451                 return 0;
 452
 453         if (l->seqnum_set &&
 454             sd_id128_equal(f->header->seqnum_id, l->seqnum_id)) {
 455
 456                 r = CMP(f->current_seqnum, l->seqnum);
 457                 if (r != 0)
 458                         return r;
 459         }
 460
 461         if (l->monotonic_set &&
 462             sd_id128_equal(f->current_boot_id, l->boot_id)) {
 463
 464                 r = CMP(f->current_monotonic, l->monotonic);
 465                 if (r != 0)
 466                         return r;
 467         }
 468
 469         if (l->realtime_set) {
 470
 471                 r = CMP(f->current_realtime, l->realtime);
 472                 if (r != 0)
 473                         return r;
 474         }
 475
 476         if (l->xor_hash_set) {
 477
 478                 r = CMP(f->current_xor_hash, l->xor_hash);
 479                 if (r != 0)
 480                         return r;
 481         }
 482
 483         return 0;
 484 }
 485
 486 static int next_for_match(
 487                 sd_journal *j,
 488                 Match *m,
 489                 JournalFile *f,
 490                 uint64_t after_offset,
 491                 direction_t direction,
 492                 Object **ret,
 493                 uint64_t *offset) {
 494
 495         int r;
 496         uint64_t np = 0;
 497         Object *n;
 498
 499         assert(j);
 500         assert(m);
 501         assert(f);
 502
 503         if (m->type == MATCH_DISCRETE) {
 504                 uint64_t dp;
 505
 506                 r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp);
 507                 if (r <= 0)
 508                         return r;
 509
 510                 return journal_file_move_to_entry_by_offset_for_data(f, dp, after_offset, direction, ret, offset);
 511
 512         } else if (m->type == MATCH_OR_TERM) {
 513                 Match *i;
 514
 515                 /* Find the earliest match beyond after_offset */
 516
 517                 LIST_FOREACH(matches, i, m->matches) {
 518                         uint64_t cp;
 519
 520                         r = next_for_match(j, i, f, after_offset, direction, NULL, &cp);
 521                         if (r < 0)
 522                                 return r;
 523                         else if (r > 0) {
 524                                 if (np == 0 || (direction == DIRECTION_DOWN ? cp < np : cp > np))
 525                                         np = cp;
 526                         }
 527                 }
 528
 529                 if (np == 0)
 530                         return 0;
 531
 532         } else if (m->type == MATCH_AND_TERM) {
 533                 Match *i, *last_moved;
 534
 535                 /* Always jump to the next matching entry and repeat
 536                  * this until we find an offset that matches for all
 537                  * matches. */
 538
 539                 if (!m->matches)
 540                         return 0;
 541
 542                 r = next_for_match(j, m->matches, f, after_offset, direction, NULL, &np);
 543                 if (r <= 0)
 544                         return r;
 545
 546                 assert(direction == DIRECTION_DOWN ? np >= after_offset : np <= after_offset);
 547                 last_moved = m->matches;
 548
 549                 LIST_LOOP_BUT_ONE(matches, i, m->matches, last_moved) {
 550                         uint64_t cp;
 551
 552                         r = next_for_match(j, i, f, np, direction, NULL, &cp);
 553                         if (r <= 0)
 554                                 return r;
 555
 556                         assert(direction == DIRECTION_DOWN ? cp >= np : cp <= np);
 557                         if (direction == DIRECTION_DOWN ? cp > np : cp < np) {
 558                                 np = cp;
 559                                 last_moved = i;
 560                         }
 561                 }
 562         }
 563
 564         assert(np > 0);
 565
 566         r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n);
 567         if (r < 0)
 568                 return r;
 569
 570         if (ret)
 571                 *ret = n;
 572         if (offset)
 573                 *offset = np;
 574
 575         return 1;
 576 }
 577
 578 static int find_location_for_match(
 579                 sd_journal *j,
 580                 Match *m,
 581                 JournalFile *f,
 582                 direction_t direction,
 583                 Object **ret,
 584                 uint64_t *offset) {
 585
 586         int r;
 587
 588         assert(j);
 589         assert(m);
 590         assert(f);
 591
 592         if (m->type == MATCH_DISCRETE) {
 593                 uint64_t dp;
 594
 595                 r = journal_file_find_data_object_with_hash(f, m->data, m->size, le64toh(m->le_hash), NULL, &dp);
 596                 if (r <= 0)
 597                         return r;
 598
 599                 /* FIXME: missing: find by monotonic */
 600
 601                 if (j->current_location.type == LOCATION_HEAD)
 602                         return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_DOWN, ret, offset);
 603                 if (j->current_location.type == LOCATION_TAIL)
 604                         return journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_UP, ret, offset);
 605                 if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
 606                         return journal_file_move_to_entry_by_seqnum_for_data(f, dp, j->current_location.seqnum, direction, ret, offset);
 607                 if (j->current_location.monotonic_set) {
 608                         r = journal_file_move_to_entry_by_monotonic_for_data(f, dp, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset);
 609                         if (r != -ENOENT)
 610                                 return r;
 611                 }
 612                 if (j->current_location.realtime_set)
 613                         return journal_file_move_to_entry_by_realtime_for_data(f, dp, j->current_location.realtime, direction, ret, offset);
 614
 615                 return journal_file_next_entry_for_data(f, NULL, 0, dp, direction, ret, offset);
 616
 617         } else if (m->type == MATCH_OR_TERM) {
 618                 uint64_t np = 0;
 619                 Object *n;
 620                 Match *i;
 621
 622                 /* Find the earliest match */
 623
 624                 LIST_FOREACH(matches, i, m->matches) {
 625                         uint64_t cp;
 626
 627                         r = find_location_for_match(j, i, f, direction, NULL, &cp);
 628                         if (r < 0)
 629                                 return r;
 630                         else if (r > 0) {
 631                                 if (np == 0 || (direction == DIRECTION_DOWN ? np > cp : np < cp))
 632                                         np = cp;
 633                         }
 634                 }
 635
 636                 if (np == 0)
 637                         return 0;
 638
 639                 r = journal_file_move_to_object(f, OBJECT_ENTRY, np, &n);
 640                 if (r < 0)
 641                         return r;
 642
 643                 if (ret)
 644                         *ret = n;
 645                 if (offset)
 646                         *offset = np;
 647
 648                 return 1;
 649
 650         } else {
 651                 Match *i;
 652                 uint64_t np = 0;
 653
 654                 assert(m->type == MATCH_AND_TERM);
 655
 656                 /* First jump to the last match, and then find the
 657                  * next one where all matches match */
 658
 659                 if (!m->matches)
 660                         return 0;
 661
 662                 LIST_FOREACH(matches, i, m->matches) {
 663                         uint64_t cp;
 664
 665                         r = find_location_for_match(j, i, f, direction, NULL, &cp);
 666                         if (r <= 0)
 667                                 return r;
 668
 669                         if (np == 0 || (direction == DIRECTION_DOWN ? cp > np : cp < np))
 670                                 np = cp;
 671                 }
 672
 673                 return next_for_match(j, m, f, np, direction, ret, offset);
 674         }
 675 }
 676
 677 static int find_location_with_matches(
 678                 sd_journal *j,
 679                 JournalFile *f,
 680                 direction_t direction,
 681                 Object **ret,
 682                 uint64_t *offset) {
 683
 684         int r;
 685
 686         assert(j);
 687         assert(f);
 688         assert(ret);
 689         assert(offset);
 690
 691         if (!j->level0) {
 692                 /* No matches is simple */
 693
 694                 if (j->current_location.type == LOCATION_HEAD)
 695                         return journal_file_next_entry(f, 0, DIRECTION_DOWN, ret, offset);
 696                 if (j->current_location.type == LOCATION_TAIL)
 697                         return journal_file_next_entry(f, 0, DIRECTION_UP, ret, offset);
 698                 if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
 699                         return journal_file_move_to_entry_by_seqnum(f, j->current_location.seqnum, direction, ret, offset);
 700                 if (j->current_location.monotonic_set) {
 701                         r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset);
 702                         if (r != -ENOENT)
 703                                 return r;
 704                 }
 705                 if (j->current_location.realtime_set)
 706                         return journal_file_move_to_entry_by_realtime(f, j->current_location.realtime, direction, ret, offset);
 707
 708                 return journal_file_next_entry(f, 0, direction, ret, offset);
 709         } else
 710                 return find_location_for_match(j, j->level0, f, direction, ret, offset);
 711 }
 712
 713 static int next_with_matches(
 714                 sd_journal *j,
 715                 JournalFile *f,
 716                 direction_t direction,
 717                 Object **ret,
 718                 uint64_t *offset) {
 719
 720         assert(j);
 721         assert(f);
 722         assert(ret);
 723         assert(offset);
 724
 725         /* No matches is easy. We simple advance the file
 726          * pointer by one. */
 727         if (!j->level0)
 728                 return journal_file_next_entry(f, f->current_offset, direction, ret, offset);
 729
 730         /* If we have a match then we look for the next matching entry
 731          * with an offset at least one step larger */
 732         return next_for_match(j, j->level0, f,
 733                               direction == DIRECTION_DOWN ? f->current_offset + 1
 734                                                           : f->current_offset - 1,
 735                               direction, ret, offset);
 736 }
 737
 738 static int next_beyond_location(sd_journal *j, JournalFile *f, direction_t direction) {
 739         Object *c;
 740         uint64_t cp, n_entries;
 741         int r;
 742
 743         assert(j);
 744         assert(f);
 745
 746         n_entries = le64toh(f->header->n_entries);
 747
 748         /* If we hit EOF before, we don't need to look into this file again
 749          * unless direction changed or new entries appeared. */
 750         if (f->last_direction == direction && f->location_type == LOCATION_TAIL &&
 751             n_entries == f->last_n_entries)
 752                 return 0;
 753
 754         f->last_n_entries = n_entries;
 755
 756         if (f->last_direction == direction && f->current_offset > 0) {
 757                 /* LOCATION_SEEK here means we did the work in a previous
 758                  * iteration and the current location already points to a
 759                  * candidate entry. */
 760                 if (f->location_type != LOCATION_SEEK) {
 761                         r = next_with_matches(j, f, direction, &c, &cp);
 762                         if (r <= 0)
 763                                 return r;
 764
 765                         journal_file_save_location(f, c, cp);
 766                 }
 767         } else {
 768                 f->last_direction = direction;
 769
 770                 r = find_location_with_matches(j, f, direction, &c, &cp);
 771                 if (r <= 0)
 772                         return r;
 773
 774                 journal_file_save_location(f, c, cp);
 775         }
 776
 777         /* OK, we found the spot, now let's advance until an entry
 778          * that is actually different from what we were previously
 779          * looking at. This is necessary to handle entries which exist
 780          * in two (or more) journal files, and which shall all be
 781          * suppressed but one. */
 782
 783         for (;;) {
 784                 bool found;
 785
 786                 if (j->current_location.type == LOCATION_DISCRETE) {
 787                         int k;
 788
 789                         k = compare_with_location(f, &j->current_location);
 790
 791                         found = direction == DIRECTION_DOWN ? k > 0 : k < 0;
 792                 } else
 793                         found = true;
 794
 795                 if (found)
 796                         return 1;
 797
 798                 r = next_with_matches(j, f, direction, &c, &cp);
 799                 if (r <= 0)
 800                         return r;
 801
 802                 journal_file_save_location(f, c, cp);
 803         }
 804 }
 805
 806 static int real_journal_next(sd_journal *j, direction_t direction) {
 807         JournalFile *new_file = NULL;
 808         unsigned i, n_files;
 809         const void **files;
 810         Object *o;
 811         int r;
 812
 813         assert_return(j, -EINVAL);
 814         assert_return(!journal_pid_changed(j), -ECHILD);
 815
 816         r = iterated_cache_get(j->files_cache, NULL, &files, &n_files);
 817         if (r < 0)
 818                 return r;
 819
 820         for (i = 0; i < n_files; i++) {
 821                 JournalFile *f = (JournalFile *)files[i];
 822                 bool found;
 823
 824                 r = next_beyond_location(j, f, direction);
 825                 if (r < 0) {
 826                         log_debug_errno(r, "Can't iterate through %s, ignoring: %m", f->path);
 827                         remove_file_real(j, f);
 828                         continue;
 829                 } else if (r == 0) {
 830                         f->location_type = LOCATION_TAIL;
 831                         continue;
 832                 }
 833
 834                 if (!new_file)
 835                         found = true;
 836                 else {
 837                         int k;
 838
 839                         k = journal_file_compare_locations(f, new_file);
 840
 841                         found = direction == DIRECTION_DOWN ? k < 0 : k > 0;
 842                 }
 843
 844                 if (found)
 845                         new_file = f;
 846         }
 847
 848         if (!new_file)
 849                 return 0;
 850
 851         r = journal_file_move_to_object(new_file, OBJECT_ENTRY, new_file->current_offset, &o);
 852         if (r < 0)
 853                 return r;
 854
 855         set_location(j, new_file, o);
 856
 857         return 1;
 858 }
 859
 860 _public_ int sd_journal_next(sd_journal *j) {
 861         return real_journal_next(j, DIRECTION_DOWN);
 862 }
 863
 864 _public_ int sd_journal_previous(sd_journal *j) {
 865         return real_journal_next(j, DIRECTION_UP);
 866 }
 867
 868 static int real_journal_next_skip(sd_journal *j, direction_t direction, uint64_t skip) {
 869         int c = 0, r;
 870
 871         assert_return(j, -EINVAL);
 872         assert_return(!journal_pid_changed(j), -ECHILD);
 873
 874         if (skip == 0) {
 875                 /* If this is not a discrete skip, then at least
 876                  * resolve the current location */
 877                 if (j->current_location.type != LOCATION_DISCRETE) {
 878                         r = real_journal_next(j, direction);
 879                         if (r < 0)
 880                                 return r;
 881                 }
 882
 883                 return 0;
 884         }
 885
 886         do {
 887                 r = real_journal_next(j, direction);
 888                 if (r < 0)
 889                         return r;
 890
 891                 if (r == 0)
 892                         return c;
 893
 894                 skip--;
 895                 c++;
 896         } while (skip > 0);
 897
 898         return c;
 899 }
 900
 901 _public_ int sd_journal_next_skip(sd_journal *j, uint64_t skip) {
 902         return real_journal_next_skip(j, DIRECTION_DOWN, skip);
 903 }
 904
 905 _public_ int sd_journal_previous_skip(sd_journal *j, uint64_t skip) {
 906         return real_journal_next_skip(j, DIRECTION_UP, skip);
 907 }
 908
 909 _public_ int sd_journal_get_cursor(sd_journal *j, char **cursor) {
 910         Object *o;
 911         int r;
 912         char bid[33], sid[33];
 913
 914         assert_return(j, -EINVAL);
 915         assert_return(!journal_pid_changed(j), -ECHILD);
 916         assert_return(cursor, -EINVAL);
 917
 918         if (!j->current_file || j->current_file->current_offset <= 0)
 919                 return -EADDRNOTAVAIL;
 920
 921         r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
 922         if (r < 0)
 923                 return r;
 924
 925         sd_id128_to_string(j->current_file->header->seqnum_id, sid);
 926         sd_id128_to_string(o->entry.boot_id, bid);
 927
 928         if (asprintf(cursor,
 929                      "s=%s;i=%"PRIx64";b=%s;m=%"PRIx64";t=%"PRIx64";x=%"PRIx64,
 930                      sid, le64toh(o->entry.seqnum),
 931                      bid, le64toh(o->entry.monotonic),
 932                      le64toh(o->entry.realtime),
 933                      le64toh(o->entry.xor_hash)) < 0)
 934                 return -ENOMEM;
 935
 936         return 0;
 937 }
 938
 939 _public_ int sd_journal_seek_cursor(sd_journal *j, const char *cursor) {
 940         const char *word, *state;
 941         size_t l;
 942         unsigned long long seqnum, monotonic, realtime, xor_hash;
 943         bool
 944                 seqnum_id_set = false,
 945                 seqnum_set = false,
 946                 boot_id_set = false,
 947                 monotonic_set = false,
 948                 realtime_set = false,
 949                 xor_hash_set = false;
 950         sd_id128_t seqnum_id, boot_id;
 951
 952         assert_return(j, -EINVAL);
 953         assert_return(!journal_pid_changed(j), -ECHILD);
 954         assert_return(!isempty(cursor), -EINVAL);
 955
 956         FOREACH_WORD_SEPARATOR(word, l, cursor, ";", state) {
 957                 char *item;
 958                 int k = 0;
 959
 960                 if (l < 2 || word[1] != '=')
 961                         return -EINVAL;
 962
 963                 item = strndup(word, l);
 964                 if (!item)
 965                         return -ENOMEM;
 966
 967                 switch (word[0]) {
 968
 969                 case 's':
 970                         seqnum_id_set = true;
 971                         k = sd_id128_from_string(item+2, &seqnum_id);
 972                         break;
 973
 974                 case 'i':
 975                         seqnum_set = true;
 976                         if (sscanf(item+2, "%llx", &seqnum) != 1)
 977                                 k = -EINVAL;
 978                         break;
 979
 980                 case 'b':
 981                         boot_id_set = true;
 982                         k = sd_id128_from_string(item+2, &boot_id);
 983                         break;
 984
 985                 case 'm':
 986                         monotonic_set = true;
 987                         if (sscanf(item+2, "%llx", &monotonic) != 1)
 988                                 k = -EINVAL;
 989                         break;
 990
 991                 case 't':
 992                         realtime_set = true;
 993                         if (sscanf(item+2, "%llx", &realtime) != 1)
 994                                 k = -EINVAL;
 995                         break;
 996
 997                 case 'x':
 998                         xor_hash_set = true;
 999                         if (sscanf(item+2, "%llx", &xor_hash) != 1)
1000                                 k = -EINVAL;
1001                         break;
1002                 }
1003
1004                 free(item);
1005
1006                 if (k < 0)
1007                         return k;
1008         }
1009
1010         if ((!seqnum_set || !seqnum_id_set) &&
1011             (!monotonic_set || !boot_id_set) &&
1012             !realtime_set)
1013                 return -EINVAL;
1014
1015         reset_location(j);
1016
1017         j->current_location.type = LOCATION_SEEK;
1018
1019         if (realtime_set) {
1020                 j->current_location.realtime = (uint64_t) realtime;
1021                 j->current_location.realtime_set = true;
1022         }
1023
1024         if (seqnum_set && seqnum_id_set) {
1025                 j->current_location.seqnum = (uint64_t) seqnum;
1026                 j->current_location.seqnum_id = seqnum_id;
1027                 j->current_location.seqnum_set = true;
1028         }
1029
1030         if (monotonic_set && boot_id_set) {
1031                 j->current_location.monotonic = (uint64_t) monotonic;
1032                 j->current_location.boot_id = boot_id;
1033                 j->current_location.monotonic_set = true;
1034         }
1035
1036         if (xor_hash_set) {
1037                 j->current_location.xor_hash = (uint64_t) xor_hash;
1038                 j->current_location.xor_hash_set = true;
1039         }
1040
1041         return 0;
1042 }
1043
1044 _public_ int sd_journal_test_cursor(sd_journal *j, const char *cursor) {
1045         int r;
1046         Object *o;
1047
1048         assert_return(j, -EINVAL);
1049         assert_return(!journal_pid_changed(j), -ECHILD);
1050         assert_return(!isempty(cursor), -EINVAL);
1051
1052         if (!j->current_file || j->current_file->current_offset <= 0)
1053                 return -EADDRNOTAVAIL;
1054
1055         r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
1056         if (r < 0)
1057                 return r;
1058
1059         for (;;) {
1060                 _cleanup_free_ char *item = NULL;
1061                 unsigned long long ll;
1062                 sd_id128_t id;
1063                 int k = 0;
1064
1065                 r = extract_first_word(&cursor, &item, ";", EXTRACT_DONT_COALESCE_SEPARATORS);
1066                 if (r < 0)
1067                         return r;
1068
1069                 if (r == 0)
1070                         break;
1071
1072                 if (strlen(item) < 2 || item[1] != '=')
1073                         return -EINVAL;
1074
1075                 switch (item[0]) {
1076
1077                 case 's':
1078                         k = sd_id128_from_string(item+2, &id);
1079                         if (k < 0)
1080                                 return k;
1081                         if (!sd_id128_equal(id, j->current_file->header->seqnum_id))
1082                                 return 0;
1083                         break;
1084
1085                 case 'i':
1086                         if (sscanf(item+2, "%llx", &ll) != 1)
1087                                 return -EINVAL;
1088                         if (ll != le64toh(o->entry.seqnum))
1089                                 return 0;
1090                         break;
1091
1092                 case 'b':
1093                         k = sd_id128_from_string(item+2, &id);
1094                         if (k < 0)
1095                                 return k;
1096                         if (!sd_id128_equal(id, o->entry.boot_id))
1097                                 return 0;
1098                         break;
1099
1100                 case 'm':
1101                         if (sscanf(item+2, "%llx", &ll) != 1)
1102                                 return -EINVAL;
1103                         if (ll != le64toh(o->entry.monotonic))
1104                                 return 0;
1105                         break;
1106
1107                 case 't':
1108                         if (sscanf(item+2, "%llx", &ll) != 1)
1109                                 return -EINVAL;
1110                         if (ll != le64toh(o->entry.realtime))
1111                                 return 0;
1112                         break;
1113
1114                 case 'x':
1115                         if (sscanf(item+2, "%llx", &ll) != 1)
1116                                 return -EINVAL;
1117                         if (ll != le64toh(o->entry.xor_hash))
1118                                 return 0;
1119                         break;
1120                 }
1121         }
1122
1123         return 1;
1124 }
1125
1126 _public_ int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec) {
1127         assert_return(j, -EINVAL);
1128         assert_return(!journal_pid_changed(j), -ECHILD);
1129
1130         reset_location(j);
1131         j->current_location.type = LOCATION_SEEK;
1132         j->current_location.boot_id = boot_id;
1133         j->current_location.monotonic = usec;
1134         j->current_location.monotonic_set = true;
1135
1136         return 0;
1137 }
1138
1139 _public_ int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec) {
1140         assert_return(j, -EINVAL);
1141         assert_return(!journal_pid_changed(j), -ECHILD);
1142
1143         reset_location(j);
1144         j->current_location.type = LOCATION_SEEK;
1145         j->current_location.realtime = usec;
1146         j->current_location.realtime_set = true;
1147
1148         return 0;
1149 }
1150
1151 _public_ int sd_journal_seek_head(sd_journal *j) {
1152         assert_return(j, -EINVAL);
1153         assert_return(!journal_pid_changed(j), -ECHILD);
1154
1155         reset_location(j);
1156         j->current_location.type = LOCATION_HEAD;
1157
1158         return 0;
1159 }
1160
1161 _public_ int sd_journal_seek_tail(sd_journal *j) {
1162         assert_return(j, -EINVAL);
1163         assert_return(!journal_pid_changed(j), -ECHILD);
1164
1165         reset_location(j);
1166         j->current_location.type = LOCATION_TAIL;
1167
1168         return 0;
1169 }
1170
1171 static void check_network(sd_journal *j, int fd) {
1172         assert(j);
1173
1174         if (j->on_network)
1175                 return;
1176
1177         j->on_network = fd_is_network_fs(fd);
1178 }
1179
1180 static bool file_has_type_prefix(const char *prefix, const char *filename) {
1181         const char *full, *tilded, *atted;
1182
1183         full = strjoina(prefix, ".journal");
1184         tilded = strjoina(full, "~");
1185         atted = strjoina(prefix, "@");
1186
1187         return STR_IN_SET(filename, full, tilded) ||
1188                startswith(filename, atted);
1189 }
1190
1191 static bool file_type_wanted(int flags, const char *filename) {
1192         assert(filename);
1193
1194         if (!endswith(filename, ".journal") && !endswith(filename, ".journal~"))
1195                 return false;
1196
1197         /* no flags set → every type is OK */
1198         if (!(flags & (SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)))
1199                 return true;
1200
1201         if (flags & SD_JOURNAL_SYSTEM && file_has_type_prefix("system", filename))
1202                 return true;
1203
1204         if (flags & SD_JOURNAL_CURRENT_USER) {
1205                 char prefix[5 + DECIMAL_STR_MAX(uid_t) + 1];
1206
1207                 xsprintf(prefix, "user-"UID_FMT, getuid());
1208
1209                 if (file_has_type_prefix(prefix, filename))
1210                         return true;
1211         }
1212
1213         return false;
1214 }
1215
1216 static bool path_has_prefix(sd_journal *j, const char *path, const char *prefix) {
1217         assert(j);
1218         assert(path);
1219         assert(prefix);
1220
1221         if (j->toplevel_fd >= 0)
1222                 return false;
1223
1224         return path_startswith(path, prefix);
1225 }
1226
1227 static void track_file_disposition(sd_journal *j, JournalFile *f) {
1228         assert(j);
1229         assert(f);
1230
1231         if (!j->has_runtime_files && path_has_prefix(j, f->path, "/run"))
1232                 j->has_runtime_files = true;
1233         else if (!j->has_persistent_files && path_has_prefix(j, f->path, "/var"))
1234                 j->has_persistent_files = true;
1235 }
1236
1237 static const char *skip_slash(const char *p) {
1238
1239         if (!p)
1240                 return NULL;
1241
1242         while (*p == '/')
1243                 p++;
1244
1245         return p;
1246 }
1247
1248 static int add_any_file(
1249                 sd_journal *j,
1250                 int fd,
1251                 const char *path) {
1252
1253         bool close_fd = false;
1254         JournalFile *f;
1255         struct stat st;
1256         int r, k;
1257
1258         assert(j);
1259         assert(fd >= 0 || path);
1260
1261         if (fd < 0) {
1262                 if (j->toplevel_fd >= 0)
1263                         /* If there's a top-level fd defined make the path relative, explicitly, since otherwise
1264                          * openat() ignores the first argument. */
1265
1266                         fd = openat(j->toplevel_fd, skip_slash(path), O_RDONLY|O_CLOEXEC|O_NONBLOCK);
1267                 else
1268                         fd = open(path, O_RDONLY|O_CLOEXEC|O_NONBLOCK);
1269                 if (fd < 0) {
1270                         r = log_debug_errno(errno, "Failed to open journal file %s: %m", path);
1271                         goto finish;
1272                 }
1273
1274                 close_fd = true;
1275
1276                 r = fd_nonblock(fd, false);
1277                 if (r < 0) {
1278                         r = log_debug_errno(errno, "Failed to turn off O_NONBLOCK for %s: %m", path);
1279                         goto finish;
1280                 }
1281         }
1282
1283         if (fstat(fd, &st) < 0) {
1284                 r = log_debug_errno(errno, "Failed to fstat file '%s': %m", path);
1285                 goto finish;
1286         }
1287
1288         r = stat_verify_regular(&st);
1289         if (r < 0) {
1290                 log_debug_errno(r, "Refusing to open '%s', as it is not a regular file.", path);
1291                 goto finish;
1292         }
1293
1294         f = ordered_hashmap_get(j->files, path);
1295         if (f) {
1296                 if (f->last_stat.st_dev == st.st_dev &&
1297                     f->last_stat.st_ino == st.st_ino) {
1298
1299                         /* We already track this file, under the same path and with the same device/inode numbers, it's
1300                          * hence really the same. Mark this file as seen in this generation. This is used to GC old
1301                          * files in process_q_overflow() to detect journal files that are still there and discern them
1302                          * from those which are gone. */
1303
1304                         f->last_seen_generation = j->generation;
1305                         r = 0;
1306                         goto finish;
1307                 }
1308
1309                 /* So we tracked a file under this name, but it has a different inode/device. In that case, it got
1310                  * replaced (probably due to rotation?), let's drop it hence from our list. */
1311                 remove_file_real(j, f);
1312                 f = NULL;
1313         }
1314
1315         if (ordered_hashmap_size(j->files) >= JOURNAL_FILES_MAX) {
1316                 log_debug("Too many open journal files, not adding %s.", path);
1317                 r = -ETOOMANYREFS;
1318                 goto finish;
1319         }
1320
1321         r = journal_file_open(fd, path, O_RDONLY, 0, false, 0, false, NULL, j->mmap, NULL, NULL, &f);
1322         if (r < 0) {
1323                 log_debug_errno(r, "Failed to open journal file %s: %m", path);
1324                 goto finish;
1325         }
1326
1327         /* journal_file_dump(f); */
1328
1329         r = ordered_hashmap_put(j->files, f->path, f);
1330         if (r < 0) {
1331                 f->close_fd = false; /* make sure journal_file_close() doesn't close the caller's fd (or our own). We'll let the caller do that, or ourselves */
1332                 (void) journal_file_close(f);
1333                 goto finish;
1334         }
1335
1336         close_fd = false; /* the fd is now owned by the JournalFile object */
1337
1338         f->last_seen_generation = j->generation;
1339
1340         track_file_disposition(j, f);
1341         check_network(j, f->fd);
1342
1343         j->current_invalidate_counter++;
1344
1345         log_debug("File %s added.", f->path);
1346
1347         r = 0;
1348
1349 finish:
1350         if (close_fd)
1351                 safe_close(fd);
1352
1353         if (r < 0) {
1354                 k = journal_put_error(j, r, path);
1355                 if (k < 0)
1356                         return k;
1357         }
1358
1359         return r;
1360 }
1361
1362 static int add_file_by_name(
1363                 sd_journal *j,
1364                 const char *prefix,
1365                 const char *filename) {
1366
1367         const char *path;
1368
1369         assert(j);
1370         assert(prefix);
1371         assert(filename);
1372
1373         if (j->no_new_files)
1374                 return 0;
1375
1376         if (!file_type_wanted(j->flags, filename))
1377                 return 0;
1378
1379         path = strjoina(prefix, "/", filename);
1380         return add_any_file(j, -1, path);
1381 }
1382
1383 static void remove_file_by_name(
1384                 sd_journal *j,
1385                 const char *prefix,
1386                 const char *filename) {
1387
1388         const char *path;
1389         JournalFile *f;
1390
1391         assert(j);
1392         assert(prefix);
1393         assert(filename);
1394
1395         path = strjoina(prefix, "/", filename);
1396         f = ordered_hashmap_get(j->files, path);
1397         if (!f)
1398                 return;
1399
1400         remove_file_real(j, f);
1401 }
1402
1403 static void remove_file_real(sd_journal *j, JournalFile *f) {
1404         assert(j);
1405         assert(f);
1406
1407         (void) ordered_hashmap_remove(j->files, f->path);
1408
1409         log_debug("File %s removed.", f->path);
1410
1411         if (j->current_file == f) {
1412                 j->current_file = NULL;
1413                 j->current_field = 0;
1414         }
1415
1416         if (j->unique_file == f) {
1417                 /* Jump to the next unique_file or NULL if that one was last */
1418                 j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path);
1419                 j->unique_offset = 0;
1420                 if (!j->unique_file)
1421                         j->unique_file_lost = true;
1422         }
1423
1424         if (j->fields_file == f) {
1425                 j->fields_file = ordered_hashmap_next(j->files, j->fields_file->path);
1426                 j->fields_offset = 0;
1427                 if (!j->fields_file)
1428                         j->fields_file_lost = true;
1429         }
1430
1431         (void) journal_file_close(f);
1432
1433         j->current_invalidate_counter++;
1434 }
1435
1436 static int dirname_is_machine_id(const char *fn) {
1437         sd_id128_t id, machine;
1438         int r;
1439
1440         r = sd_id128_get_machine(&machine);
1441         if (r < 0)
1442                 return r;
1443
1444         r = sd_id128_from_string(fn, &id);
1445         if (r < 0)
1446                 return r;
1447
1448         return sd_id128_equal(id, machine);
1449 }
1450
1451 static bool dirent_is_journal_file(const struct dirent *de) {
1452         assert(de);
1453
1454         if (!IN_SET(de->d_type, DT_REG, DT_LNK, DT_UNKNOWN))
1455                 return false;
1456
1457         return endswith(de->d_name, ".journal") ||
1458                 endswith(de->d_name, ".journal~");
1459 }
1460
1461 static bool dirent_is_id128_subdir(const struct dirent *de) {
1462         assert(de);
1463
1464         if (!IN_SET(de->d_type, DT_DIR, DT_LNK, DT_UNKNOWN))
1465                 return false;
1466
1467         return id128_is_valid(de->d_name);
1468 }
1469
1470 static int directory_open(sd_journal *j, const char *path, DIR **ret) {
1471         DIR *d;
1472
1473         assert(j);
1474         assert(path);
1475         assert(ret);
1476
1477         if (j->toplevel_fd < 0)
1478                 d = opendir(path);
1479         else
1480                 /* Open the specified directory relative to the toplevel fd. Enforce that the path specified is
1481                  * relative, by dropping the initial slash */
1482                 d = xopendirat(j->toplevel_fd, skip_slash(path), 0);
1483         if (!d)
1484                 return -errno;
1485
1486         *ret = d;
1487         return 0;
1488 }
1489
1490 static int add_directory(sd_journal *j, const char *prefix, const char *dirname);
1491
1492 static void directory_enumerate(sd_journal *j, Directory *m, DIR *d) {
1493         struct dirent *de;
1494
1495         assert(j);
1496         assert(m);
1497         assert(d);
1498
1499         FOREACH_DIRENT_ALL(de, d, goto fail) {
1500
1501                 if (dirent_is_journal_file(de))
1502                         (void) add_file_by_name(j, m->path, de->d_name);
1503
1504                 if (m->is_root && dirent_is_id128_subdir(de))
1505                         (void) add_directory(j, m->path, de->d_name);
1506         }
1507
1508         return;
1509
1510 fail:
1511         log_debug_errno(errno, "Failed to enumerate directory %s, ignoring: %m", m->path);
1512 }
1513
1514 static void directory_watch(sd_journal *j, Directory *m, int fd, uint32_t mask) {
1515         int r;
1516
1517         assert(j);
1518         assert(m);
1519         assert(fd >= 0);
1520
1521         /* Watch this directory if that's enabled and if it not being watched yet. */
1522
1523         if (m->wd > 0) /* Already have a watch? */
1524                 return;
1525         if (j->inotify_fd < 0) /* Not watching at all? */
1526                 return;
1527
1528         m->wd = inotify_add_watch_fd(j->inotify_fd, fd, mask);
1529         if (m->wd < 0) {
1530                 log_debug_errno(errno, "Failed to watch journal directory '%s', ignoring: %m", m->path);
1531                 return;
1532         }
1533
1534         r = hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m);
1535         if (r == -EEXIST)
1536                 log_debug_errno(r, "Directory '%s' already being watched under a different path, ignoring: %m", m->path);
1537         if (r < 0) {
1538                 log_debug_errno(r, "Failed to add watch for journal directory '%s' to hashmap, ignoring: %m", m->path);
1539                 (void) inotify_rm_watch(j->inotify_fd, m->wd);
1540                 m->wd = -1;
1541         }
1542 }
1543
1544 static int add_directory(sd_journal *j, const char *prefix, const char *dirname) {
1545         _cleanup_free_ char *path = NULL;
1546         _cleanup_closedir_ DIR *d = NULL;
1547         Directory *m;
1548         int r, k;
1549
1550         assert(j);
1551         assert(prefix);
1552
1553         /* Adds a journal file directory to watch. If the directory is already tracked this updates the inotify watch
1554          * and reenumerates directory contents */
1555
1556         path = path_join(prefix, dirname);
1557         if (!path) {
1558                 r = -ENOMEM;
1559                 goto fail;
1560         }
1561
1562         log_debug("Considering directory '%s'.", path);
1563
1564         /* We consider everything local that is in a directory for the local machine ID, or that is stored in /run */
1565         if ((j->flags & SD_JOURNAL_LOCAL_ONLY) &&
1566             !((dirname && dirname_is_machine_id(dirname) > 0) || path_has_prefix(j, path, "/run")))
1567                 return 0;
1568
1569         r = directory_open(j, path, &d);
1570         if (r < 0) {
1571                 log_debug_errno(r, "Failed to open directory '%s': %m", path);
1572                 goto fail;
1573         }
1574
1575         m = hashmap_get(j->directories_by_path, path);
1576         if (!m) {
1577                 m = new0(Directory, 1);
1578                 if (!m) {
1579                         r = -ENOMEM;
1580                         goto fail;
1581                 }
1582
1583                 m->is_root = false;
1584                 m->path = path;
1585
1586                 if (hashmap_put(j->directories_by_path, m->path, m) < 0) {
1587                         free(m);
1588                         r = -ENOMEM;
1589                         goto fail;
1590                 }
1591
1592                 path = NULL; /* avoid freeing in cleanup */
1593                 j->current_invalidate_counter++;
1594
1595                 log_debug("Directory %s added.", m->path);
1596
1597         } else if (m->is_root)
1598                 return 0; /* Don't 'downgrade' from root directory */
1599
1600         m->last_seen_generation = j->generation;
1601
1602         directory_watch(j, m, dirfd(d),
1603                         IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE|
1604                         IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT|IN_MOVED_FROM|
1605                         IN_ONLYDIR);
1606
1607         if (!j->no_new_files)
1608                 directory_enumerate(j, m, d);
1609
1610         check_network(j, dirfd(d));
1611
1612         return 0;
1613
1614 fail:
1615         k = journal_put_error(j, r, path ?: prefix);
1616         if (k < 0)
1617                 return k;
1618
1619         return r;
1620 }
1621
1622 static int add_root_directory(sd_journal *j, const char *p, bool missing_ok) {
1623
1624         _cleanup_closedir_ DIR *d = NULL;
1625         Directory *m;
1626         int r, k;
1627
1628         assert(j);
1629
1630         /* Adds a root directory to our set of directories to use. If the root directory is already in the set, we
1631          * update the inotify logic, and renumerate the directory entries. This call may hence be called to initially
1632          * populate the set, as well as to update it later. */
1633
1634         if (p) {
1635                 /* If there's a path specified, use it. */
1636
1637                 log_debug("Considering root directory '%s'.", p);
1638
1639                 if ((j->flags & SD_JOURNAL_RUNTIME_ONLY) &&
1640                     !path_has_prefix(j, p, "/run"))
1641                         return -EINVAL;
1642
1643                 if (j->prefix)
1644                         p = strjoina(j->prefix, p);
1645
1646                 r = directory_open(j, p, &d);
1647                 if (r == -ENOENT && missing_ok)
1648                         return 0;
1649                 if (r < 0) {
1650                         log_debug_errno(r, "Failed to open root directory %s: %m", p);
1651                         goto fail;
1652                 }
1653         } else {
1654                 int dfd;
1655
1656                 /* If there's no path specified, then we use the top-level fd itself. We duplicate the fd here, since
1657                  * opendir() will take possession of the fd, and close it, which we don't want. */
1658
1659                 p = "."; /* store this as "." in the directories hashmap */
1660
1661                 dfd = fcntl(j->toplevel_fd, F_DUPFD_CLOEXEC, 3);
1662                 if (dfd < 0) {
1663                         r = -errno;
1664                         goto fail;
1665                 }
1666
1667                 d = fdopendir(dfd);
1668                 if (!d) {
1669                         r = -errno;
1670                         safe_close(dfd);
1671                         goto fail;
1672                 }
1673
1674                 rewinddir(d);
1675         }
1676
1677         m = hashmap_get(j->directories_by_path, p);
1678         if (!m) {
1679                 m = new0(Directory, 1);
1680                 if (!m) {
1681                         r = -ENOMEM;
1682                         goto fail;
1683                 }
1684
1685                 m->is_root = true;
1686
1687                 m->path = strdup(p);
1688                 if (!m->path) {
1689                         free(m);
1690                         r = -ENOMEM;
1691                         goto fail;
1692                 }
1693
1694                 if (hashmap_put(j->directories_by_path, m->path, m) < 0) {
1695                         free(m->path);
1696                         free(m);
1697                         r = -ENOMEM;
1698                         goto fail;
1699                 }
1700
1701                 j->current_invalidate_counter++;
1702
1703                 log_debug("Root directory %s added.", m->path);
1704
1705         } else if (!m->is_root)
1706                 return 0;
1707
1708         directory_watch(j, m, dirfd(d),
1709                         IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE|
1710                         IN_ONLYDIR);
1711
1712         if (!j->no_new_files)
1713                 directory_enumerate(j, m, d);
1714
1715         check_network(j, dirfd(d));
1716
1717         return 0;
1718
1719 fail:
1720         k = journal_put_error(j, r, p);
1721         if (k < 0)
1722                 return k;
1723
1724         return r;
1725 }
1726
1727 static void remove_directory(sd_journal *j, Directory *d) {
1728         assert(j);
1729
1730         if (d->wd > 0) {
1731                 hashmap_remove(j->directories_by_wd, INT_TO_PTR(d->wd));
1732
1733                 if (j->inotify_fd >= 0)
1734                         (void) inotify_rm_watch(j->inotify_fd, d->wd);
1735         }
1736
1737         hashmap_remove(j->directories_by_path, d->path);
1738
1739         if (d->is_root)
1740                 log_debug("Root directory %s removed.", d->path);
1741         else
1742                 log_debug("Directory %s removed.", d->path);
1743
1744         free(d->path);
1745         free(d);
1746 }
1747
1748 static int add_search_paths(sd_journal *j) {
1749
1750         static const char search_paths[] =
1751                 "/run/log/journal\0"
1752                 "/var/log/journal\0";
1753         const char *p;
1754
1755         assert(j);
1756
1757         /* We ignore most errors here, since the idea is to only open
1758          * what's actually accessible, and ignore the rest. */
1759
1760         NULSTR_FOREACH(p, search_paths)
1761                 (void) add_root_directory(j, p, true);
1762
1763         if (!(j->flags & SD_JOURNAL_LOCAL_ONLY))
1764                 (void) add_root_directory(j, "/var/log/journal/remote", true);
1765
1766         return 0;
1767 }
1768
1769 static int add_current_paths(sd_journal *j) {
1770         Iterator i;
1771         JournalFile *f;
1772
1773         assert(j);
1774         assert(j->no_new_files);
1775
1776         /* Simply adds all directories for files we have open as directories. We don't expect errors here, so we
1777          * treat them as fatal. */
1778
1779         ORDERED_HASHMAP_FOREACH(f, j->files, i) {
1780                 _cleanup_free_ char *dir;
1781                 int r;
1782
1783                 dir = dirname_malloc(f->path);
1784                 if (!dir)
1785                         return -ENOMEM;
1786
1787                 r = add_directory(j, dir, NULL);
1788                 if (r < 0)
1789                         return r;
1790         }
1791
1792         return 0;
1793 }
1794
1795 static int allocate_inotify(sd_journal *j) {
1796         assert(j);
1797
1798         if (j->inotify_fd < 0) {
1799                 j->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC);
1800                 if (j->inotify_fd < 0)
1801                         return -errno;
1802         }
1803
1804         return hashmap_ensure_allocated(&j->directories_by_wd, NULL);
1805 }
1806
1807 static sd_journal *journal_new(int flags, const char *path) {
1808         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
1809
1810         j = new0(sd_journal, 1);
1811         if (!j)
1812                 return NULL;
1813
1814         j->original_pid = getpid_cached();
1815         j->toplevel_fd = -1;
1816         j->inotify_fd = -1;
1817         j->flags = flags;
1818         j->data_threshold = DEFAULT_DATA_THRESHOLD;
1819
1820         if (path) {
1821                 char *t;
1822
1823                 t = strdup(path);
1824                 if (!t)
1825                         return NULL;
1826
1827                 if (flags & SD_JOURNAL_OS_ROOT)
1828                         j->prefix = t;
1829                 else
1830                         j->path = t;
1831         }
1832
1833         j->files = ordered_hashmap_new(&path_hash_ops);
1834         if (!j->files)
1835                 return NULL;
1836
1837         j->files_cache = ordered_hashmap_iterated_cache_new(j->files);
1838         j->directories_by_path = hashmap_new(&path_hash_ops);
1839         j->mmap = mmap_cache_new();
1840         if (!j->files_cache || !j->directories_by_path || !j->mmap)
1841                 return NULL;
1842
1843         return TAKE_PTR(j);
1844 }
1845
1846 #define OPEN_ALLOWED_FLAGS                              \
1847         (SD_JOURNAL_LOCAL_ONLY |                        \
1848          SD_JOURNAL_RUNTIME_ONLY |                      \
1849          SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER)
1850
1851 _public_ int sd_journal_open(sd_journal **ret, int flags) {
1852         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
1853         int r;
1854
1855         assert_return(ret, -EINVAL);
1856         assert_return((flags & ~OPEN_ALLOWED_FLAGS) == 0, -EINVAL);
1857
1858         j = journal_new(flags, NULL);
1859         if (!j)
1860                 return -ENOMEM;
1861
1862         r = add_search_paths(j);
1863         if (r < 0)
1864                 return r;
1865
1866         *ret = TAKE_PTR(j);
1867         return 0;
1868 }
1869
1870 #define OPEN_CONTAINER_ALLOWED_FLAGS                    \
1871         (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM)
1872
1873 _public_ int sd_journal_open_container(sd_journal **ret, const char *machine, int flags) {
1874         _cleanup_free_ char *root = NULL, *class = NULL;
1875         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
1876         char *p;
1877         int r;
1878
1879         /* This is pretty much deprecated, people should use machined's OpenMachineRootDirectory() call instead in
1880          * combination with sd_journal_open_directory_fd(). */
1881
1882         assert_return(machine, -EINVAL);
1883         assert_return(ret, -EINVAL);
1884         assert_return((flags & ~OPEN_CONTAINER_ALLOWED_FLAGS) == 0, -EINVAL);
1885         assert_return(machine_name_is_valid(machine), -EINVAL);
1886
1887         p = strjoina("/run/systemd/machines/", machine);
1888         r = parse_env_file(NULL, p,
1889                            "ROOT", &root,
1890                            "CLASS", &class);
1891         if (r == -ENOENT)
1892                 return -EHOSTDOWN;
1893         if (r < 0)
1894                 return r;
1895         if (!root)
1896                 return -ENODATA;
1897
1898         if (!streq_ptr(class, "container"))
1899                 return -EIO;
1900
1901         j = journal_new(flags, root);
1902         if (!j)
1903                 return -ENOMEM;
1904
1905         r = add_search_paths(j);
1906         if (r < 0)
1907                 return r;
1908
1909         *ret = TAKE_PTR(j);
1910         return 0;
1911 }
1912
1913 #define OPEN_DIRECTORY_ALLOWED_FLAGS                    \
1914         (SD_JOURNAL_OS_ROOT |                           \
1915          SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1916
1917 _public_ int sd_journal_open_directory(sd_journal **ret, const char *path, int flags) {
1918         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
1919         int r;
1920
1921         assert_return(ret, -EINVAL);
1922         assert_return(path, -EINVAL);
1923         assert_return((flags & ~OPEN_DIRECTORY_ALLOWED_FLAGS) == 0, -EINVAL);
1924
1925         j = journal_new(flags, path);
1926         if (!j)
1927                 return -ENOMEM;
1928
1929         if (flags & SD_JOURNAL_OS_ROOT)
1930                 r = add_search_paths(j);
1931         else
1932                 r = add_root_directory(j, path, false);
1933         if (r < 0)
1934                 return r;
1935
1936         *ret = TAKE_PTR(j);
1937         return 0;
1938 }
1939
1940 _public_ int sd_journal_open_files(sd_journal **ret, const char **paths, int flags) {
1941         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
1942         const char **path;
1943         int r;
1944
1945         assert_return(ret, -EINVAL);
1946         assert_return(flags == 0, -EINVAL);
1947
1948         j = journal_new(flags, NULL);
1949         if (!j)
1950                 return -ENOMEM;
1951
1952         STRV_FOREACH(path, paths) {
1953                 r = add_any_file(j, -1, *path);
1954                 if (r < 0)
1955                         return r;
1956         }
1957
1958         j->no_new_files = true;
1959
1960         *ret = TAKE_PTR(j);
1961         return 0;
1962 }
1963
1964 #define OPEN_DIRECTORY_FD_ALLOWED_FLAGS         \
1965         (SD_JOURNAL_OS_ROOT |                           \
1966          SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER )
1967
1968 _public_ int sd_journal_open_directory_fd(sd_journal **ret, int fd, int flags) {
1969         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
1970         struct stat st;
1971         int r;
1972
1973         assert_return(ret, -EINVAL);
1974         assert_return(fd >= 0, -EBADF);
1975         assert_return((flags & ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS) == 0, -EINVAL);
1976
1977         if (fstat(fd, &st) < 0)
1978                 return -errno;
1979
1980         if (!S_ISDIR(st.st_mode))
1981                 return -EBADFD;
1982
1983         j = journal_new(flags, NULL);
1984         if (!j)
1985                 return -ENOMEM;
1986
1987         j->toplevel_fd = fd;
1988
1989         if (flags & SD_JOURNAL_OS_ROOT)
1990                 r = add_search_paths(j);
1991         else
1992                 r = add_root_directory(j, NULL, false);
1993         if (r < 0)
1994                 return r;
1995
1996         *ret = TAKE_PTR(j);
1997         return 0;
1998 }
1999
2000 _public_ int sd_journal_open_files_fd(sd_journal **ret, int fds[], unsigned n_fds, int flags) {
2001         Iterator iterator;
2002         JournalFile *f;
2003         _cleanup_(sd_journal_closep) sd_journal *j = NULL;
2004         unsigned i;
2005         int r;
2006
2007         assert_return(ret, -EINVAL);
2008         assert_return(n_fds > 0, -EBADF);
2009         assert_return(flags == 0, -EINVAL);
2010
2011         j = journal_new(flags, NULL);
2012         if (!j)
2013                 return -ENOMEM;
2014
2015         for (i = 0; i < n_fds; i++) {
2016                 struct stat st;
2017
2018                 if (fds[i] < 0) {
2019                         r = -EBADF;
2020                         goto fail;
2021                 }
2022
2023                 if (fstat(fds[i], &st) < 0) {
2024                         r = -errno;
2025                         goto fail;
2026                 }
2027
2028                 r = stat_verify_regular(&st);
2029                 if (r < 0)
2030                         goto fail;
2031
2032                 r = add_any_file(j, fds[i], NULL);
2033                 if (r < 0)
2034                         goto fail;
2035         }
2036
2037         j->no_new_files = true;
2038         j->no_inotify = true;
2039
2040         *ret = TAKE_PTR(j);
2041         return 0;
2042
2043 fail:
2044         /* If we fail, make sure we don't take possession of the files we managed to make use of successfully, and they
2045          * remain open */
2046         ORDERED_HASHMAP_FOREACH(f, j->files, iterator)
2047                 f->close_fd = false;
2048
2049         return r;
2050 }
2051
2052 _public_ void sd_journal_close(sd_journal *j) {
2053         Directory *d;
2054
2055         if (!j)
2056                 return;
2057
2058         sd_journal_flush_matches(j);
2059
2060         ordered_hashmap_free_with_destructor(j->files, journal_file_close);
2061         iterated_cache_free(j->files_cache);
2062
2063         while ((d = hashmap_first(j->directories_by_path)))
2064                 remove_directory(j, d);
2065
2066         while ((d = hashmap_first(j->directories_by_wd)))
2067                 remove_directory(j, d);
2068
2069         hashmap_free(j->directories_by_path);
2070         hashmap_free(j->directories_by_wd);
2071
2072         safe_close(j->inotify_fd);
2073
2074         if (j->mmap) {
2075                 log_debug("mmap cache statistics: %u hit, %u miss", mmap_cache_get_hit(j->mmap), mmap_cache_get_missed(j->mmap));
2076                 mmap_cache_unref(j->mmap);
2077         }
2078
2079         hashmap_free_free(j->errors);
2080
2081         free(j->path);
2082         free(j->prefix);
2083         free(j->unique_field);
2084         free(j->fields_buffer);
2085         free(j);
2086 }
2087
2088 _public_ int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret) {
2089         Object *o;
2090         JournalFile *f;
2091         int r;
2092
2093         assert_return(j, -EINVAL);
2094         assert_return(!journal_pid_changed(j), -ECHILD);
2095         assert_return(ret, -EINVAL);
2096
2097         f = j->current_file;
2098         if (!f)
2099                 return -EADDRNOTAVAIL;
2100
2101         if (f->current_offset <= 0)
2102                 return -EADDRNOTAVAIL;
2103
2104         r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2105         if (r < 0)
2106                 return r;
2107
2108         *ret = le64toh(o->entry.realtime);
2109         return 0;
2110 }
2111
2112 _public_ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id) {
2113         Object *o;
2114         JournalFile *f;
2115         int r;
2116         sd_id128_t id;
2117
2118         assert_return(j, -EINVAL);
2119         assert_return(!journal_pid_changed(j), -ECHILD);
2120
2121         f = j->current_file;
2122         if (!f)
2123                 return -EADDRNOTAVAIL;
2124
2125         if (f->current_offset <= 0)
2126                 return -EADDRNOTAVAIL;
2127
2128         r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2129         if (r < 0)
2130                 return r;
2131
2132         if (ret_boot_id)
2133                 *ret_boot_id = o->entry.boot_id;
2134         else {
2135                 r = sd_id128_get_boot(&id);
2136                 if (r < 0)
2137                         return r;
2138
2139                 if (!sd_id128_equal(id, o->entry.boot_id))
2140                         return -ESTALE;
2141         }
2142
2143         if (ret)
2144                 *ret = le64toh(o->entry.monotonic);
2145
2146         return 0;
2147 }
2148
2149 static bool field_is_valid(const char *field) {
2150         const char *p;
2151
2152         assert(field);
2153
2154         if (isempty(field))
2155                 return false;
2156
2157         if (startswith(field, "__"))
2158                 return false;
2159
2160         for (p = field; *p; p++) {
2161
2162                 if (*p == '_')
2163                         continue;
2164
2165                 if (*p >= 'A' && *p <= 'Z')
2166                         continue;
2167
2168                 if (*p >= '0' && *p <= '9')
2169                         continue;
2170
2171                 return false;
2172         }
2173
2174         return true;
2175 }
2176
2177 _public_ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *size) {
2178         JournalFile *f;
2179         uint64_t i, n;
2180         size_t field_length;
2181         int r;
2182         Object *o;
2183
2184         assert_return(j, -EINVAL);
2185         assert_return(!journal_pid_changed(j), -ECHILD);
2186         assert_return(field, -EINVAL);
2187         assert_return(data, -EINVAL);
2188         assert_return(size, -EINVAL);
2189         assert_return(field_is_valid(field), -EINVAL);
2190
2191         f = j->current_file;
2192         if (!f)
2193                 return -EADDRNOTAVAIL;
2194
2195         if (f->current_offset <= 0)
2196                 return -EADDRNOTAVAIL;
2197
2198         r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2199         if (r < 0)
2200                 return r;
2201
2202         field_length = strlen(field);
2203
2204         n = journal_file_entry_n_items(o);
2205         for (i = 0; i < n; i++) {
2206                 uint64_t p, l;
2207                 le64_t le_hash;
2208                 size_t t;
2209                 int compression;
2210
2211                 p = le64toh(o->entry.items[i].object_offset);
2212                 le_hash = o->entry.items[i].hash;
2213                 r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
2214                 if (r < 0)
2215                         return r;
2216
2217                 if (le_hash != o->data.hash)
2218                         return -EBADMSG;
2219
2220                 l = le64toh(o->object.size) - offsetof(Object, data.payload);
2221
2222                 compression = o->object.flags & OBJECT_COMPRESSION_MASK;
2223                 if (compression) {
2224 #if HAVE_XZ || HAVE_LZ4
2225                         r = decompress_startswith(compression,
2226                                                   o->data.payload, l,
2227                                                   &f->compress_buffer, &f->compress_buffer_size,
2228                                                   field, field_length, '=');
2229                         if (r < 0)
2230                                 log_debug_errno(r, "Cannot decompress %s object of length %"PRIu64" at offset "OFSfmt": %m",
2231                                                 object_compressed_to_string(compression), l, p);
2232                         else if (r > 0) {
2233
2234                                 size_t rsize;
2235
2236                                 r = decompress_blob(compression,
2237                                                     o->data.payload, l,
2238                                                     &f->compress_buffer, &f->compress_buffer_size, &rsize,
2239                                                     j->data_threshold);
2240                                 if (r < 0)
2241                                         return r;
2242
2243                                 *data = f->compress_buffer;
2244                                 *size = (size_t) rsize;
2245
2246                                 return 0;
2247                         }
2248 #else
2249                         return -EPROTONOSUPPORT;
2250 #endif
2251                 } else if (l >= field_length+1 &&
2252                            memcmp(o->data.payload, field, field_length) == 0 &&
2253                            o->data.payload[field_length] == '=') {
2254
2255                         t = (size_t) l;
2256
2257                         if ((uint64_t) t != l)
2258                                 return -E2BIG;
2259
2260                         *data = o->data.payload;
2261                         *size = t;
2262
2263                         return 0;
2264                 }
2265
2266                 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2267                 if (r < 0)
2268                         return r;
2269         }
2270
2271         return -ENOENT;
2272 }
2273
2274 static int return_data(sd_journal *j, JournalFile *f, Object *o, const void **data, size_t *size) {
2275         size_t t;
2276         uint64_t l;
2277         int compression;
2278
2279         l = le64toh(o->object.size) - offsetof(Object, data.payload);
2280         t = (size_t) l;
2281
2282         /* We can't read objects larger than 4G on a 32bit machine */
2283         if ((uint64_t) t != l)
2284                 return -E2BIG;
2285
2286         compression = o->object.flags & OBJECT_COMPRESSION_MASK;
2287         if (compression) {
2288 #if HAVE_XZ || HAVE_LZ4
2289                 size_t rsize;
2290                 int r;
2291
2292                 r = decompress_blob(compression,
2293                                     o->data.payload, l, &f->compress_buffer,
2294                                     &f->compress_buffer_size, &rsize, j->data_threshold);
2295                 if (r < 0)
2296                         return r;
2297
2298                 *data = f->compress_buffer;
2299                 *size = (size_t) rsize;
2300 #else
2301                 return -EPROTONOSUPPORT;
2302 #endif
2303         } else {
2304                 *data = o->data.payload;
2305                 *size = t;
2306         }
2307
2308         return 0;
2309 }
2310
2311 _public_ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
2312         JournalFile *f;
2313         uint64_t p, n;
2314         le64_t le_hash;
2315         int r;
2316         Object *o;
2317
2318         assert_return(j, -EINVAL);
2319         assert_return(!journal_pid_changed(j), -ECHILD);
2320         assert_return(data, -EINVAL);
2321         assert_return(size, -EINVAL);
2322
2323         f = j->current_file;
2324         if (!f)
2325                 return -EADDRNOTAVAIL;
2326
2327         if (f->current_offset <= 0)
2328                 return -EADDRNOTAVAIL;
2329
2330         r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
2331         if (r < 0)
2332                 return r;
2333
2334         n = journal_file_entry_n_items(o);
2335         if (j->current_field >= n)
2336                 return 0;
2337
2338         p = le64toh(o->entry.items[j->current_field].object_offset);
2339         le_hash = o->entry.items[j->current_field].hash;
2340         r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
2341         if (r < 0)
2342                 return r;
2343
2344         if (le_hash != o->data.hash)
2345                 return -EBADMSG;
2346
2347         r = return_data(j, f, o, data, size);
2348         if (r < 0)
2349                 return r;
2350
2351         j->current_field++;
2352
2353         return 1;
2354 }
2355
2356 _public_ void sd_journal_restart_data(sd_journal *j) {
2357         if (!j)
2358                 return;
2359
2360         j->current_field = 0;
2361 }
2362
2363 static int reiterate_all_paths(sd_journal *j) {
2364         assert(j);
2365
2366         if (j->no_new_files)
2367                 return add_current_paths(j);
2368
2369         if (j->flags & SD_JOURNAL_OS_ROOT)
2370                 return add_search_paths(j);
2371
2372         if (j->toplevel_fd >= 0)
2373                 return add_root_directory(j, NULL, false);
2374
2375         if (j->path)
2376                 return add_root_directory(j, j->path, true);
2377
2378         return add_search_paths(j);
2379 }
2380
2381 _public_ int sd_journal_get_fd(sd_journal *j) {
2382         int r;
2383
2384         assert_return(j, -EINVAL);
2385         assert_return(!journal_pid_changed(j), -ECHILD);
2386
2387         if (j->no_inotify)
2388                 return -EMEDIUMTYPE;
2389
2390         if (j->inotify_fd >= 0)
2391                 return j->inotify_fd;
2392
2393         r = allocate_inotify(j);
2394         if (r < 0)
2395                 return r;
2396
2397         log_debug("Reiterating files to get inotify watches established.");
2398
2399         /* Iterate through all dirs again, to add them to the inotify */
2400         r = reiterate_all_paths(j);
2401         if (r < 0)
2402                 return r;
2403
2404         return j->inotify_fd;
2405 }
2406
2407 _public_ int sd_journal_get_events(sd_journal *j) {
2408         int fd;
2409
2410         assert_return(j, -EINVAL);
2411         assert_return(!journal_pid_changed(j), -ECHILD);
2412
2413         fd = sd_journal_get_fd(j);
2414         if (fd < 0)
2415                 return fd;
2416
2417         return POLLIN;
2418 }
2419
2420 _public_ int sd_journal_get_timeout(sd_journal *j, uint64_t *timeout_usec) {
2421         int fd;
2422
2423         assert_return(j, -EINVAL);
2424         assert_return(!journal_pid_changed(j), -ECHILD);
2425         assert_return(timeout_usec, -EINVAL);
2426
2427         fd = sd_journal_get_fd(j);
2428         if (fd < 0)
2429                 return fd;
2430
2431         if (!j->on_network) {
2432                 *timeout_usec = (uint64_t) -1;
2433                 return 0;
2434         }
2435
2436         /* If we are on the network we need to regularly check for
2437          * changes manually */
2438
2439         *timeout_usec = j->last_process_usec + JOURNAL_FILES_RECHECK_USEC;
2440         return 1;
2441 }
2442
2443 static void process_q_overflow(sd_journal *j) {
2444         JournalFile *f;
2445         Directory *m;
2446         Iterator i;
2447
2448         assert(j);
2449
2450         /* When the inotify queue overruns we need to enumerate and re-validate all journal files to bring our list
2451          * back in sync with what's on disk. For this we pick a new generation counter value. It'll be assigned to all
2452          * journal files we encounter. All journal files and all directories that don't carry it after reenumeration
2453          * are subject for unloading. */
2454
2455         log_debug("Inotify queue overrun, reiterating everything.");
2456
2457         j->generation++;
2458         (void) reiterate_all_paths(j);
2459
2460         ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2461
2462                 if (f->last_seen_generation == j->generation)
2463                         continue;
2464
2465                 log_debug("File '%s' hasn't been seen in this enumeration, removing.", f->path);
2466                 remove_file_real(j, f);
2467         }
2468
2469         HASHMAP_FOREACH(m, j->directories_by_path, i) {
2470
2471                 if (m->last_seen_generation == j->generation)
2472                         continue;
2473
2474                 if (m->is_root) /* Never GC root directories */
2475                         continue;
2476
2477                 log_debug("Directory '%s' hasn't been seen in this enumeration, removing.", f->path);
2478                 remove_directory(j, m);
2479         }
2480
2481         log_debug("Reiteration complete.");
2482 }
2483
2484 static void process_inotify_event(sd_journal *j, struct inotify_event *e) {
2485         Directory *d;
2486
2487         assert(j);
2488         assert(e);
2489
2490         if (e->mask & IN_Q_OVERFLOW) {
2491                 process_q_overflow(j);
2492                 return;
2493         }
2494
2495         /* Is this a subdirectory we watch? */
2496         d = hashmap_get(j->directories_by_wd, INT_TO_PTR(e->wd));
2497         if (d) {
2498                 if (!(e->mask & IN_ISDIR) && e->len > 0 &&
2499                     (endswith(e->name, ".journal") ||
2500                      endswith(e->name, ".journal~"))) {
2501
2502                         /* Event for a journal file */
2503
2504                         if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB))
2505                                 (void) add_file_by_name(j, d->path, e->name);
2506                         else if (e->mask & (IN_DELETE|IN_MOVED_FROM|IN_UNMOUNT))
2507                                 remove_file_by_name(j, d->path, e->name);
2508
2509                 } else if (!d->is_root && e->len == 0) {
2510
2511                         /* Event for a subdirectory */
2512
2513                         if (e->mask & (IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT))
2514                                 remove_directory(j, d);
2515
2516                 } else if (d->is_root && (e->mask & IN_ISDIR) && e->len > 0 && id128_is_valid(e->name)) {
2517
2518                         /* Event for root directory */
2519
2520                         if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB))
2521                                 (void) add_directory(j, d->path, e->name);
2522                 }
2523
2524                 return;
2525         }
2526
2527         if (e->mask & IN_IGNORED)
2528                 return;
2529
2530         log_debug("Unexpected inotify event.");
2531 }
2532
2533 static int determine_change(sd_journal *j) {
2534         bool b;
2535
2536         assert(j);
2537
2538         b = j->current_invalidate_counter != j->last_invalidate_counter;
2539         j->last_invalidate_counter = j->current_invalidate_counter;
2540
2541         return b ? SD_JOURNAL_INVALIDATE : SD_JOURNAL_APPEND;
2542 }
2543
2544 _public_ int sd_journal_process(sd_journal *j) {
2545         bool got_something = false;
2546
2547         assert_return(j, -EINVAL);
2548         assert_return(!journal_pid_changed(j), -ECHILD);
2549
2550         if (j->inotify_fd < 0) /* We have no inotify fd yet? Then there's noting to process. */
2551                 return 0;
2552
2553         j->last_process_usec = now(CLOCK_MONOTONIC);
2554         j->last_invalidate_counter = j->current_invalidate_counter;
2555
2556         for (;;) {
2557                 union inotify_event_buffer buffer;
2558                 struct inotify_event *e;
2559                 ssize_t l;
2560
2561                 l = read(j->inotify_fd, &buffer, sizeof(buffer));
2562                 if (l < 0) {
2563                         if (IN_SET(errno, EAGAIN, EINTR))
2564                                 return got_something ? determine_change(j) : SD_JOURNAL_NOP;
2565
2566                         return -errno;
2567                 }
2568
2569                 got_something = true;
2570
2571                 FOREACH_INOTIFY_EVENT(e, buffer, l)
2572                         process_inotify_event(j, e);
2573         }
2574 }
2575
2576 _public_ int sd_journal_wait(sd_journal *j, uint64_t timeout_usec) {
2577         int r;
2578         uint64_t t;
2579
2580         assert_return(j, -EINVAL);
2581         assert_return(!journal_pid_changed(j), -ECHILD);
2582
2583         if (j->inotify_fd < 0) {
2584
2585                 /* This is the first invocation, hence create the
2586                  * inotify watch */
2587                 r = sd_journal_get_fd(j);
2588                 if (r < 0)
2589                         return r;
2590
2591                 /* The journal might have changed since the context
2592                  * object was created and we weren't watching before,
2593                  * hence don't wait for anything, and return
2594                  * immediately. */
2595                 return determine_change(j);
2596         }
2597
2598         r = sd_journal_get_timeout(j, &t);
2599         if (r < 0)
2600                 return r;
2601
2602         if (t != (uint64_t) -1) {
2603                 usec_t n;
2604
2605                 n = now(CLOCK_MONOTONIC);
2606                 t = t > n ? t - n : 0;
2607
2608                 if (timeout_usec == (uint64_t) -1 || timeout_usec > t)
2609                         timeout_usec = t;
2610         }
2611
2612         do {
2613                 r = fd_wait_for_event(j->inotify_fd, POLLIN, timeout_usec);
2614         } while (r == -EINTR);
2615
2616         if (r < 0)
2617                 return r;
2618
2619         return sd_journal_process(j);
2620 }
2621
2622 _public_ int sd_journal_get_cutoff_realtime_usec(sd_journal *j, uint64_t *from, uint64_t *to) {
2623         Iterator i;
2624         JournalFile *f;
2625         bool first = true;
2626         uint64_t fmin = 0, tmax = 0;
2627         int r;
2628
2629         assert_return(j, -EINVAL);
2630         assert_return(!journal_pid_changed(j), -ECHILD);
2631         assert_return(from || to, -EINVAL);
2632         assert_return(from != to, -EINVAL);
2633
2634         ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2635                 usec_t fr, t;
2636
2637                 r = journal_file_get_cutoff_realtime_usec(f, &fr, &t);
2638                 if (r == -ENOENT)
2639                         continue;
2640                 if (r < 0)
2641                         return r;
2642                 if (r == 0)
2643                         continue;
2644
2645                 if (first) {
2646                         fmin = fr;
2647                         tmax = t;
2648                         first = false;
2649                 } else {
2650                         fmin = MIN(fr, fmin);
2651                         tmax = MAX(t, tmax);
2652                 }
2653         }
2654
2655         if (from)
2656                 *from = fmin;
2657         if (to)
2658                 *to = tmax;
2659
2660         return first ? 0 : 1;
2661 }
2662
2663 _public_ int sd_journal_get_cutoff_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t *from, uint64_t *to) {
2664         Iterator i;
2665         JournalFile *f;
2666         bool found = false;
2667         int r;
2668
2669         assert_return(j, -EINVAL);
2670         assert_return(!journal_pid_changed(j), -ECHILD);
2671         assert_return(from || to, -EINVAL);
2672         assert_return(from != to, -EINVAL);
2673
2674         ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2675                 usec_t fr, t;
2676
2677                 r = journal_file_get_cutoff_monotonic_usec(f, boot_id, &fr, &t);
2678                 if (r == -ENOENT)
2679                         continue;
2680                 if (r < 0)
2681                         return r;
2682                 if (r == 0)
2683                         continue;
2684
2685                 if (found) {
2686                         if (from)
2687                                 *from = MIN(fr, *from);
2688                         if (to)
2689                                 *to = MAX(t, *to);
2690                 } else {
2691                         if (from)
2692                                 *from = fr;
2693                         if (to)
2694                                 *to = t;
2695                         found = true;
2696                 }
2697         }
2698
2699         return found;
2700 }
2701
2702 void journal_print_header(sd_journal *j) {
2703         Iterator i;
2704         JournalFile *f;
2705         bool newline = false;
2706
2707         assert(j);
2708
2709         ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2710                 if (newline)
2711                         putchar('\n');
2712                 else
2713                         newline = true;
2714
2715                 journal_file_print_header(f);
2716         }
2717 }
2718
2719 _public_ int sd_journal_get_usage(sd_journal *j, uint64_t *bytes) {
2720         Iterator i;
2721         JournalFile *f;
2722         uint64_t sum = 0;
2723
2724         assert_return(j, -EINVAL);
2725         assert_return(!journal_pid_changed(j), -ECHILD);
2726         assert_return(bytes, -EINVAL);
2727
2728         ORDERED_HASHMAP_FOREACH(f, j->files, i) {
2729                 struct stat st;
2730
2731                 if (fstat(f->fd, &st) < 0)
2732                         return -errno;
2733
2734                 sum += (uint64_t) st.st_blocks * 512ULL;
2735         }
2736
2737         *bytes = sum;
2738         return 0;
2739 }
2740
2741 _public_ int sd_journal_query_unique(sd_journal *j, const char *field) {
2742         char *f;
2743
2744         assert_return(j, -EINVAL);
2745         assert_return(!journal_pid_changed(j), -ECHILD);
2746         assert_return(!isempty(field), -EINVAL);
2747         assert_return(field_is_valid(field), -EINVAL);
2748
2749         f = strdup(field);
2750         if (!f)
2751                 return -ENOMEM;
2752
2753         free(j->unique_field);
2754         j->unique_field = f;
2755         j->unique_file = NULL;
2756         j->unique_offset = 0;
2757         j->unique_file_lost = false;
2758
2759         return 0;
2760 }
2761
2762 _public_ int sd_journal_enumerate_unique(sd_journal *j, const void **data, size_t *l) {
2763         size_t k;
2764
2765         assert_return(j, -EINVAL);
2766         assert_return(!journal_pid_changed(j), -ECHILD);
2767         assert_return(data, -EINVAL);
2768         assert_return(l, -EINVAL);
2769         assert_return(j->unique_field, -EINVAL);
2770
2771         k = strlen(j->unique_field);
2772
2773         if (!j->unique_file) {
2774                 if (j->unique_file_lost)
2775                         return 0;
2776
2777                 j->unique_file = ordered_hashmap_first(j->files);
2778                 if (!j->unique_file)
2779                         return 0;
2780
2781                 j->unique_offset = 0;
2782         }
2783
2784         for (;;) {
2785                 JournalFile *of;
2786                 Iterator i;
2787                 Object *o;
2788                 const void *odata;
2789                 size_t ol;
2790                 bool found;
2791                 int r;
2792
2793                 /* Proceed to next data object in the field's linked list */
2794                 if (j->unique_offset == 0) {
2795                         r = journal_file_find_field_object(j->unique_file, j->unique_field, k, &o, NULL);
2796                         if (r < 0)
2797                                 return r;
2798
2799                         j->unique_offset = r > 0 ? le64toh(o->field.head_data_offset) : 0;
2800                 } else {
2801                         r = journal_file_move_to_object(j->unique_file, OBJECT_DATA, j->unique_offset, &o);
2802                         if (r < 0)
2803                                 return r;
2804
2805                         j->unique_offset = le64toh(o->data.next_field_offset);
2806                 }
2807
2808                 /* We reached the end of the list? Then start again, with the next file */
2809                 if (j->unique_offset == 0) {
2810                         j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path);
2811                         if (!j->unique_file)
2812                                 return 0;
2813
2814                         continue;
2815                 }
2816
2817                 /* We do not use OBJECT_DATA context here, but OBJECT_UNUSED
2818                  * instead, so that we can look at this data object at the same
2819                  * time as one on another file */
2820                 r = journal_file_move_to_object(j->unique_file, OBJECT_UNUSED, j->unique_offset, &o);
2821                 if (r < 0)
2822                         return r;
2823
2824                 /* Let's do the type check by hand, since we used 0 context above. */
2825                 if (o->object.type != OBJECT_DATA)
2826                         return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG),
2827                                                "%s:offset " OFSfmt ": object has type %d, expected %d",
2828                                                j->unique_file->path,
2829                                                j->unique_offset,
2830                                                o->object.type, OBJECT_DATA);
2831
2832                 r = return_data(j, j->unique_file, o, &odata, &ol);
2833                 if (r < 0)
2834                         return r;
2835
2836                 /* Check if we have at least the field name and "=". */
2837                 if (ol <= k)
2838                         return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG),
2839                                                "%s:offset " OFSfmt ": object has size %zu, expected at least %zu",
2840                                                j->unique_file->path,
2841                                                j->unique_offset, ol, k + 1);
2842
2843                 if (memcmp(odata, j->unique_field, k) || ((const char*) odata)[k] != '=')
2844                         return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG),
2845                                                "%s:offset " OFSfmt ": object does not start with \"%s=\"",
2846                                                j->unique_file->path,
2847                                                j->unique_offset,
2848                                                j->unique_field);
2849
2850                 /* OK, now let's see if we already returned this data
2851                  * object by checking if it exists in the earlier
2852                  * traversed files. */
2853                 found = false;
2854                 ORDERED_HASHMAP_FOREACH(of, j->files, i) {
2855                         if (of == j->unique_file)
2856                                 break;
2857
2858                         /* Skip this file it didn't have any fields indexed */
2859                         if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0)
2860                                 continue;
2861
2862                         r = journal_file_find_data_object_with_hash(of, odata, ol, le64toh(o->data.hash), NULL, NULL);
2863                         if (r < 0)
2864                                 return r;
2865                         if (r > 0) {
2866                                 found = true;
2867                                 break;
2868                         }
2869                 }
2870
2871                 if (found)
2872                         continue;
2873
2874                 r = return_data(j, j->unique_file, o, data, l);
2875                 if (r < 0)
2876                         return r;
2877
2878                 return 1;
2879         }
2880 }
2881
2882 _public_ void sd_journal_restart_unique(sd_journal *j) {
2883         if (!j)
2884                 return;
2885
2886         j->unique_file = NULL;
2887         j->unique_offset = 0;
2888         j->unique_file_lost = false;
2889 }
2890
2891 _public_ int sd_journal_enumerate_fields(sd_journal *j, const char **field) {
2892         int r;
2893
2894         assert_return(j, -EINVAL);
2895         assert_return(!journal_pid_changed(j), -ECHILD);
2896         assert_return(field, -EINVAL);
2897
2898         if (!j->fields_file) {
2899                 if (j->fields_file_lost)
2900                         return 0;
2901
2902                 j->fields_file = ordered_hashmap_first(j->files);
2903                 if (!j->fields_file)
2904                         return 0;
2905
2906                 j->fields_hash_table_index = 0;
2907                 j->fields_offset = 0;
2908         }
2909
2910         for (;;) {
2911                 JournalFile *f, *of;
2912                 Iterator i;
2913                 uint64_t m;
2914                 Object *o;
2915                 size_t sz;
2916                 bool found;
2917
2918                 f = j->fields_file;
2919
2920                 if (j->fields_offset == 0) {
2921                         bool eof = false;
2922
2923                         /* We are not yet positioned at any field. Let's pick the first one */
2924                         r = journal_file_map_field_hash_table(f);
2925                         if (r < 0)
2926                                 return r;
2927
2928                         m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem);
2929                         for (;;) {
2930                                 if (j->fields_hash_table_index >= m) {
2931                                         /* Reached the end of the hash table, go to the next file. */
2932                                         eof = true;
2933                                         break;
2934                                 }
2935
2936                                 j->fields_offset = le64toh(f->field_hash_table[j->fields_hash_table_index].head_hash_offset);
2937
2938                                 if (j->fields_offset != 0)
2939                                         break;
2940
2941                                 /* Empty hash table bucket, go to next one */
2942                                 j->fields_hash_table_index++;
2943                         }
2944
2945                         if (eof) {
2946                                 /* Proceed with next file */
2947                                 j->fields_file = ordered_hashmap_next(j->files, f->path);
2948                                 if (!j->fields_file) {
2949                                         *field = NULL;
2950                                         return 0;
2951                                 }
2952
2953                                 j->fields_offset = 0;
2954                                 j->fields_hash_table_index = 0;
2955                                 continue;
2956                         }
2957
2958                 } else {
2959                         /* We are already positioned at a field. If so, let's figure out the next field from it */
2960
2961                         r = journal_file_move_to_object(f, OBJECT_FIELD, j->fields_offset, &o);
2962                         if (r < 0)
2963                                 return r;
2964
2965                         j->fields_offset = le64toh(o->field.next_hash_offset);
2966                         if (j->fields_offset == 0) {
2967                                 /* Reached the end of the hash table chain */
2968                                 j->fields_hash_table_index++;
2969                                 continue;
2970                         }
2971                 }
2972
2973                 /* We use OBJECT_UNUSED here, so that the iterator below doesn't remove our mmap window */
2974                 r = journal_file_move_to_object(f, OBJECT_UNUSED, j->fields_offset, &o);
2975                 if (r < 0)
2976                         return r;
2977
2978                 /* Because we used OBJECT_UNUSED above, we need to do our type check manually */
2979                 if (o->object.type != OBJECT_FIELD)
2980                         return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG),
2981                                                "%s:offset " OFSfmt ": object has type %i, expected %i",
2982                                                f->path, j->fields_offset,
2983                                                o->object.type, OBJECT_FIELD);
2984
2985                 sz = le64toh(o->object.size) - offsetof(Object, field.payload);
2986
2987                 /* Let's see if we already returned this field name before. */
2988                 found = false;
2989                 ORDERED_HASHMAP_FOREACH(of, j->files, i) {
2990                         if (of == f)
2991                                 break;
2992
2993                         /* Skip this file it didn't have any fields indexed */
2994                         if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0)
2995                                 continue;
2996
2997                         r = journal_file_find_field_object_with_hash(of, o->field.payload, sz, le64toh(o->field.hash), NULL, NULL);
2998                         if (r < 0)
2999                                 return r;
3000                         if (r > 0) {
3001                                 found = true;
3002                                 break;
3003                         }
3004                 }
3005
3006                 if (found)
3007                         continue;
3008
3009                 /* Check if this is really a valid string containing no NUL byte */
3010                 if (memchr(o->field.payload, 0, sz))
3011                         return -EBADMSG;
3012
3013                 if (sz > j->data_threshold)
3014                         sz = j->data_threshold;
3015
3016                 if (!GREEDY_REALLOC(j->fields_buffer, j->fields_buffer_allocated, sz + 1))
3017                         return -ENOMEM;
3018
3019                 memcpy(j->fields_buffer, o->field.payload, sz);
3020                 j->fields_buffer[sz] = 0;
3021
3022                 if (!field_is_valid(j->fields_buffer))
3023                         return -EBADMSG;
3024
3025                 *field = j->fields_buffer;
3026                 return 1;
3027         }
3028 }
3029
3030 _public_ void sd_journal_restart_fields(sd_journal *j) {
3031         if (!j)
3032                 return;
3033
3034         j->fields_file = NULL;
3035         j->fields_hash_table_index = 0;
3036         j->fields_offset = 0;
3037         j->fields_file_lost = false;
3038 }
3039
3040 _public_ int sd_journal_reliable_fd(sd_journal *j) {
3041         assert_return(j, -EINVAL);
3042         assert_return(!journal_pid_changed(j), -ECHILD);
3043
3044         return !j->on_network;
3045 }
3046
3047 static char *lookup_field(const char *field, void *userdata) {
3048         sd_journal *j = userdata;
3049         const void *data;
3050         size_t size, d;
3051         int r;
3052
3053         assert(field);
3054         assert(j);
3055
3056         r = sd_journal_get_data(j, field, &data, &size);
3057         if (r < 0 ||
3058             size > REPLACE_VAR_MAX)
3059                 return strdup(field);
3060
3061         d = strlen(field) + 1;
3062
3063         return strndup((const char*) data + d, size - d);
3064 }
3065
3066 _public_ int sd_journal_get_catalog(sd_journal *j, char **ret) {
3067         const void *data;
3068         size_t size;
3069         sd_id128_t id;
3070         _cleanup_free_ char *text = NULL, *cid = NULL;
3071         char *t;
3072         int r;
3073
3074         assert_return(j, -EINVAL);
3075         assert_return(!journal_pid_changed(j), -ECHILD);
3076         assert_return(ret, -EINVAL);
3077
3078         r = sd_journal_get_data(j, "MESSAGE_ID", &data, &size);
3079         if (r < 0)
3080                 return r;
3081
3082         cid = strndup((const char*) data + 11, size - 11);
3083         if (!cid)
3084                 return -ENOMEM;
3085
3086         r = sd_id128_from_string(cid, &id);
3087         if (r < 0)
3088                 return r;
3089
3090         r = catalog_get(CATALOG_DATABASE, id, &text);
3091         if (r < 0)
3092                 return r;
3093
3094         t = replace_var(text, lookup_field, j);
3095         if (!t)
3096                 return -ENOMEM;
3097
3098         *ret = t;
3099         return 0;
3100 }
3101
3102 _public_ int sd_journal_get_catalog_for_message_id(sd_id128_t id, char **ret) {
3103         assert_return(ret, -EINVAL);
3104
3105         return catalog_get(CATALOG_DATABASE, id, ret);
3106 }
3107
3108 _public_ int sd_journal_set_data_threshold(sd_journal *j, size_t sz) {
3109         assert_return(j, -EINVAL);
3110         assert_return(!journal_pid_changed(j), -ECHILD);
3111
3112         j->data_threshold = sz;
3113         return 0;
3114 }
3115
3116 _public_ int sd_journal_get_data_threshold(sd_journal *j, size_t *sz) {
3117         assert_return(j, -EINVAL);
3118         assert_return(!journal_pid_changed(j), -ECHILD);
3119         assert_return(sz, -EINVAL);
3120
3121         *sz = j->data_threshold;
3122         return 0;
3123 }
3124
3125 _public_ int sd_journal_has_runtime_files(sd_journal *j) {
3126         assert_return(j, -EINVAL);
3127
3128         return j->has_runtime_files;
3129 }
3130
3131 _public_ int sd_journal_has_persistent_files(sd_journal *j) {
3132         assert_return(j, -EINVAL);
3133
3134         return j->has_persistent_files;
3135 }