src/journal/journald.c

   1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
   2
   3 /***
   4   This file is part of systemd.
   5
   6   Copyright 2011 Lennart Poettering
   7
   8   systemd is free software; you can redistribute it and/or modify it
   9   under the terms of the GNU Lesser General Public License as published by
  10   the Free Software Foundation; either version 2.1 of the License, or
  11   (at your option) any later version.
  12
  13   systemd is distributed in the hope that it will be useful, but
  14   WITHOUT ANY WARRANTY; without even the implied warranty of
  15   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  16   Lesser General Public License for more details.
  17
  18   You should have received a copy of the GNU Lesser General Public License
  19   along with systemd; If not, see <http://www.gnu.org/licenses/>.
  20 ***/
  21
  22 #include <sys/epoll.h>
  23 #include <sys/socket.h>
  24 #include <errno.h>
  25 #include <sys/signalfd.h>
  26 #include <unistd.h>
  27 #include <fcntl.h>
  28 #include <stddef.h>
  29 #include <sys/ioctl.h>
  30 #include <linux/sockios.h>
  31 #include <sys/statvfs.h>
  32 #include <sys/mman.h>
  33
  34 #include <libudev.h>
  35 #include <systemd/sd-journal.h>
  36 #include <systemd/sd-messages.h>
  37 #include <systemd/sd-daemon.h>
  38
  39 #ifdef HAVE_LOGIND
  40 #include <systemd/sd-login.h>
  41 #endif
  42
  43 #include "mkdir.h"
  44 #include "hashmap.h"
  45 #include "journal-file.h"
  46 #include "socket-util.h"
  47 #include "cgroup-util.h"
  48 #include "list.h"
  49 #include "virt.h"
  50 #include "missing.h"
  51 #include "conf-parser.h"
  52 #include "journal-internal.h"
  53 #include "journal-vacuum.h"
  54 #include "journal-authenticate.h"
  55 #include "journald.h"
  56 #include "journald-rate-limit.h"
  57 #include "journald-kmsg.h"
  58 #include "journald-syslog.h"
  59 #include "journald-stream.h"
  60 #include "journald-console.h"
  61 #include "journald-native.h"
  62
  63 #ifdef HAVE_ACL
  64 #include <sys/acl.h>
  65 #include <acl/libacl.h>
  66 #include "acl-util.h"
  67 #endif
  68
  69 #ifdef HAVE_SELINUX
  70 #include <selinux/selinux.h>
  71 #endif
  72
  73 #define USER_JOURNALS_MAX 1024
  74
  75 #define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC)
  76 #define DEFAULT_RATE_LIMIT_BURST 200
  77
  78 #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
  79
  80 static const char* const storage_table[] = {
  81         [STORAGE_AUTO] = "auto",
  82         [STORAGE_VOLATILE] = "volatile",
  83         [STORAGE_PERSISTENT] = "persistent",
  84         [STORAGE_NONE] = "none"
  85 };
  86
  87 DEFINE_STRING_TABLE_LOOKUP(storage, Storage);
  88 DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting");
  89
  90 static uint64_t available_space(Server *s) {
  91         char ids[33], *p;
  92         const char *f;
  93         sd_id128_t machine;
  94         struct statvfs ss;
  95         uint64_t sum = 0, avail = 0, ss_avail = 0;
  96         int r;
  97         DIR *d;
  98         usec_t ts;
  99         JournalMetrics *m;
 100
 101         ts = now(CLOCK_MONOTONIC);
 102
 103         if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts)
 104                 return s->cached_available_space;
 105
 106         r = sd_id128_get_machine(&machine);
 107         if (r < 0)
 108                 return 0;
 109
 110         if (s->system_journal) {
 111                 f = "/var/log/journal/";
 112                 m = &s->system_metrics;
 113         } else {
 114                 f = "/run/log/journal/";
 115                 m = &s->runtime_metrics;
 116         }
 117
 118         assert(m);
 119
 120         p = strappend(f, sd_id128_to_string(machine, ids));
 121         if (!p)
 122                 return 0;
 123
 124         d = opendir(p);
 125         free(p);
 126
 127         if (!d)
 128                 return 0;
 129
 130         if (fstatvfs(dirfd(d), &ss) < 0)
 131                 goto finish;
 132
 133         for (;;) {
 134                 struct stat st;
 135                 struct dirent buf, *de;
 136
 137                 r = readdir_r(d, &buf, &de);
 138                 if (r != 0)
 139                         break;
 140
 141                 if (!de)
 142                         break;
 143
 144                 if (!endswith(de->d_name, ".journal") &&
 145                     !endswith(de->d_name, ".journal~"))
 146                         continue;
 147
 148                 if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
 149                         continue;
 150
 151                 if (!S_ISREG(st.st_mode))
 152                         continue;
 153
 154                 sum += (uint64_t) st.st_blocks * 512UL;
 155         }
 156
 157         avail = sum >= m->max_use ? 0 : m->max_use - sum;
 158
 159         ss_avail = ss.f_bsize * ss.f_bavail;
 160
 161         ss_avail = ss_avail < m->keep_free ? 0 : ss_avail - m->keep_free;
 162
 163         if (ss_avail < avail)
 164                 avail = ss_avail;
 165
 166         s->cached_available_space = avail;
 167         s->cached_available_space_timestamp = ts;
 168
 169 finish:
 170         closedir(d);
 171
 172         return avail;
 173 }
 174
 175 static void server_read_file_gid(Server *s) {
 176         const char *adm = "adm";
 177         int r;
 178
 179         assert(s);
 180
 181         if (s->file_gid_valid)
 182                 return;
 183
 184         r = get_group_creds(&adm, &s->file_gid);
 185         if (r < 0)
 186                 log_warning("Failed to resolve 'adm' group: %s", strerror(-r));
 187
 188         /* if we couldn't read the gid, then it will be 0, but that's
 189          * fine and we shouldn't try to resolve the group again, so
 190          * let's just pretend it worked right-away. */
 191         s->file_gid_valid = true;
 192 }
 193
 194 static void server_fix_perms(Server *s, JournalFile *f, uid_t uid) {
 195         int r;
 196 #ifdef HAVE_ACL
 197         acl_t acl;
 198         acl_entry_t entry;
 199         acl_permset_t permset;
 200 #endif
 201
 202         assert(f);
 203
 204         server_read_file_gid(s);
 205
 206         r = fchmod_and_fchown(f->fd, 0640, 0, s->file_gid);
 207         if (r < 0)
 208                 log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
 209
 210 #ifdef HAVE_ACL
 211         if (uid <= 0)
 212                 return;
 213
 214         acl = acl_get_fd(f->fd);
 215         if (!acl) {
 216                 log_warning("Failed to read ACL on %s, ignoring: %m", f->path);
 217                 return;
 218         }
 219
 220         r = acl_find_uid(acl, uid, &entry);
 221         if (r <= 0) {
 222
 223                 if (acl_create_entry(&acl, &entry) < 0 ||
 224                     acl_set_tag_type(entry, ACL_USER) < 0 ||
 225                     acl_set_qualifier(entry, &uid) < 0) {
 226                         log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
 227                         goto finish;
 228                 }
 229         }
 230
 231         if (acl_get_permset(entry, &permset) < 0 ||
 232             acl_add_perm(permset, ACL_READ) < 0 ||
 233             acl_calc_mask(&acl) < 0) {
 234                 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
 235                 goto finish;
 236         }
 237
 238         if (acl_set_fd(f->fd, acl) < 0)
 239                 log_warning("Failed to set ACL on %s, ignoring: %m", f->path);
 240
 241 finish:
 242         acl_free(acl);
 243 #endif
 244 }
 245
 246 static JournalFile* find_journal(Server *s, uid_t uid) {
 247         char *p;
 248         int r;
 249         JournalFile *f;
 250         sd_id128_t machine;
 251
 252         assert(s);
 253
 254         /* We split up user logs only on /var, not on /run. If the
 255          * runtime file is open, we write to it exclusively, in order
 256          * to guarantee proper order as soon as we flush /run to
 257          * /var and close the runtime file. */
 258
 259         if (s->runtime_journal)
 260                 return s->runtime_journal;
 261
 262         if (uid <= 0)
 263                 return s->system_journal;
 264
 265         r = sd_id128_get_machine(&machine);
 266         if (r < 0)
 267                 return s->system_journal;
 268
 269         f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid));
 270         if (f)
 271                 return f;
 272
 273         if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-%lu.journal",
 274                      SD_ID128_FORMAT_VAL(machine), (unsigned long) uid) < 0)
 275                 return s->system_journal;
 276
 277         while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
 278                 /* Too many open? Then let's close one */
 279                 f = hashmap_steal_first(s->user_journals);
 280                 assert(f);
 281                 journal_file_close(f);
 282         }
 283
 284         r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, s->system_journal, &f);
 285         free(p);
 286
 287         if (r < 0)
 288                 return s->system_journal;
 289
 290         server_fix_perms(s, f, uid);
 291
 292         r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
 293         if (r < 0) {
 294                 journal_file_close(f);
 295                 return s->system_journal;
 296         }
 297
 298         return f;
 299 }
 300
 301 static void server_rotate(Server *s) {
 302         JournalFile *f;
 303         void *k;
 304         Iterator i;
 305         int r;
 306
 307         log_info("Rotating...");
 308
 309         if (s->runtime_journal) {
 310                 r = journal_file_rotate(&s->runtime_journal, s->compress, false);
 311                 if (r < 0)
 312                         if (s->runtime_journal)
 313                                 log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
 314                         else
 315                                 log_error("Failed to create new runtime journal: %s", strerror(-r));
 316                 else
 317                         server_fix_perms(s, s->runtime_journal, 0);
 318         }
 319
 320         if (s->system_journal) {
 321                 r = journal_file_rotate(&s->system_journal, s->compress, s->seal);
 322                 if (r < 0)
 323                         if (s->system_journal)
 324                                 log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
 325                         else
 326                                 log_error("Failed to create new system journal: %s", strerror(-r));
 327
 328                 else
 329                         server_fix_perms(s, s->system_journal, 0);
 330         }
 331
 332         HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
 333                 r = journal_file_rotate(&f, s->compress, s->seal);
 334                 if (r < 0)
 335                         if (f->path)
 336                                 log_error("Failed to rotate %s: %s", f->path, strerror(-r));
 337                         else
 338                                 log_error("Failed to create user journal: %s", strerror(-r));
 339                 else {
 340                         hashmap_replace(s->user_journals, k, f);
 341                         server_fix_perms(s, s->system_journal, PTR_TO_UINT32(k));
 342                 }
 343         }
 344 }
 345
 346 static void server_vacuum(Server *s) {
 347         char *p;
 348         char ids[33];
 349         sd_id128_t machine;
 350         int r;
 351
 352         log_info("Vacuuming...");
 353
 354         r = sd_id128_get_machine(&machine);
 355         if (r < 0) {
 356                 log_error("Failed to get machine ID: %s", strerror(-r));
 357                 return;
 358         }
 359
 360         sd_id128_to_string(machine, ids);
 361
 362         if (s->system_journal) {
 363                 p = strappend("/var/log/journal/", ids);
 364                 if (!p) {
 365                         log_oom();
 366                         return;
 367                 }
 368
 369                 r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free);
 370                 if (r < 0 && r != -ENOENT)
 371                         log_error("Failed to vacuum %s: %s", p, strerror(-r));
 372                 free(p);
 373         }
 374
 375         if (s->runtime_journal) {
 376                 p = strappend("/run/log/journal/", ids);
 377                 if (!p) {
 378                         log_oom();
 379                         return;
 380                 }
 381
 382                 r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free);
 383                 if (r < 0 && r != -ENOENT)
 384                         log_error("Failed to vacuum %s: %s", p, strerror(-r));
 385                 free(p);
 386         }
 387
 388         s->cached_available_space_timestamp = 0;
 389 }
 390
 391 static char *shortened_cgroup_path(pid_t pid) {
 392         int r;
 393         char *process_path, *init_path, *path;
 394
 395         assert(pid > 0);
 396
 397         r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path);
 398         if (r < 0)
 399                 return NULL;
 400
 401         r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path);
 402         if (r < 0) {
 403                 free(process_path);
 404                 return NULL;
 405         }
 406
 407         if (endswith(init_path, "/system"))
 408                 init_path[strlen(init_path) - 7] = 0;
 409         else if (streq(init_path, "/"))
 410                 init_path[0] = 0;
 411
 412         if (startswith(process_path, init_path)) {
 413                 char *p;
 414
 415                 p = strdup(process_path + strlen(init_path));
 416                 if (!p) {
 417                         free(process_path);
 418                         free(init_path);
 419                         return NULL;
 420                 }
 421                 path = p;
 422         } else {
 423                 path = process_path;
 424                 process_path = NULL;
 425         }
 426
 427         free(process_path);
 428         free(init_path);
 429
 430         return path;
 431 }
 432
 433 static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n) {
 434         JournalFile *f;
 435         bool vacuumed = false;
 436         int r;
 437
 438         assert(s);
 439         assert(iovec);
 440         assert(n > 0);
 441
 442         f = find_journal(s, uid);
 443         if (!f)
 444                 return;
 445
 446         if (journal_file_rotate_suggested(f)) {
 447                 log_info("Journal header limits reached or header out-of-date, rotating.");
 448                 server_rotate(s);
 449                 server_vacuum(s);
 450                 vacuumed = true;
 451
 452                 f = find_journal(s, uid);
 453                 if (!f)
 454                         return;
 455         }
 456
 457         for (;;) {
 458                 r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
 459                 if (r >= 0)
 460                         return;
 461
 462                 if (vacuumed ||
 463                     (r != -E2BIG && /* hit limit */
 464                      r != -EFBIG && /* hit fs limit */
 465                      r != -EDQUOT && /* quota hit */
 466                      r != -ENOSPC && /* disk full */
 467                      r != -EBADMSG && /* corrupted */
 468                      r != -ENODATA && /* truncated */
 469                      r != -EHOSTDOWN && /* other machine */
 470                      r != -EPROTONOSUPPORT && /* unsupported feature */
 471                      r != -EBUSY && /* unclean shutdown */
 472                      r != -ESHUTDOWN /* already archived */)) {
 473                         log_error("Failed to write entry, ignoring: %s", strerror(-r));
 474                         return;
 475                 }
 476
 477                 if (r == -E2BIG || r == -EFBIG || r == EDQUOT || r == ENOSPC)
 478                         log_info("Allocation limit reached, rotating.");
 479                 else if (r == -EHOSTDOWN)
 480                         log_info("Journal file from other machine, rotating.");
 481                 else if (r == -EBUSY)
 482                         log_info("Unlcean shutdown, rotating.");
 483                 else
 484                         log_warning("Journal file corrupted, rotating.");
 485
 486                 server_rotate(s);
 487                 server_vacuum(s);
 488                 vacuumed = true;
 489
 490                 f = find_journal(s, uid);
 491                 if (!f)
 492                         return;
 493
 494                 log_info("Retrying write.");
 495         }
 496 }
 497
 498 static void dispatch_message_real(
 499                 Server *s,
 500                 struct iovec *iovec, unsigned n, unsigned m,
 501                 struct ucred *ucred,
 502                 struct timeval *tv,
 503                 const char *label, size_t label_len,
 504                 const char *unit_id) {
 505
 506         char *pid = NULL, *uid = NULL, *gid = NULL,
 507                 *source_time = NULL, *boot_id = NULL, *machine_id = NULL,
 508                 *comm = NULL, *cmdline = NULL, *hostname = NULL,
 509                 *audit_session = NULL, *audit_loginuid = NULL,
 510                 *exe = NULL, *cgroup = NULL, *session = NULL,
 511                 *owner_uid = NULL, *unit = NULL, *selinux_context = NULL;
 512
 513         char idbuf[33];
 514         sd_id128_t id;
 515         int r;
 516         char *t;
 517         uid_t loginuid = 0, realuid = 0;
 518
 519         assert(s);
 520         assert(iovec);
 521         assert(n > 0);
 522         assert(n + N_IOVEC_META_FIELDS <= m);
 523
 524         if (ucred) {
 525                 uint32_t audit;
 526 #ifdef HAVE_LOGIND
 527                 uid_t owner;
 528 #endif
 529
 530                 realuid = ucred->uid;
 531
 532                 if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
 533                         IOVEC_SET_STRING(iovec[n++], pid);
 534
 535                 if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
 536                         IOVEC_SET_STRING(iovec[n++], uid);
 537
 538                 if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
 539                         IOVEC_SET_STRING(iovec[n++], gid);
 540
 541                 r = get_process_comm(ucred->pid, &t);
 542                 if (r >= 0) {
 543                         comm = strappend("_COMM=", t);
 544                         free(t);
 545
 546                         if (comm)
 547                                 IOVEC_SET_STRING(iovec[n++], comm);
 548                 }
 549
 550                 r = get_process_exe(ucred->pid, &t);
 551                 if (r >= 0) {
 552                         exe = strappend("_EXE=", t);
 553                         free(t);
 554
 555                         if (exe)
 556                                 IOVEC_SET_STRING(iovec[n++], exe);
 557                 }
 558
 559                 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
 560                 if (r >= 0) {
 561                         cmdline = strappend("_CMDLINE=", t);
 562                         free(t);
 563
 564                         if (cmdline)
 565                                 IOVEC_SET_STRING(iovec[n++], cmdline);
 566                 }
 567
 568                 r = audit_session_from_pid(ucred->pid, &audit);
 569                 if (r >= 0)
 570                         if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit) >= 0)
 571                                 IOVEC_SET_STRING(iovec[n++], audit_session);
 572
 573                 r = audit_loginuid_from_pid(ucred->pid, &loginuid);
 574                 if (r >= 0)
 575                         if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
 576                                 IOVEC_SET_STRING(iovec[n++], audit_loginuid);
 577
 578                 t = shortened_cgroup_path(ucred->pid);
 579                 if (t) {
 580                         cgroup = strappend("_SYSTEMD_CGROUP=", t);
 581                         free(t);
 582
 583                         if (cgroup)
 584                                 IOVEC_SET_STRING(iovec[n++], cgroup);
 585                 }
 586
 587 #ifdef HAVE_LOGIND
 588                 if (sd_pid_get_session(ucred->pid, &t) >= 0) {
 589                         session = strappend("_SYSTEMD_SESSION=", t);
 590                         free(t);
 591
 592                         if (session)
 593                                 IOVEC_SET_STRING(iovec[n++], session);
 594                 }
 595
 596                 if (sd_pid_get_owner_uid(ucred->uid, &owner) >= 0)
 597                         if (asprintf(&owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner) >= 0)
 598                                 IOVEC_SET_STRING(iovec[n++], owner_uid);
 599 #endif
 600
 601                 if (cg_pid_get_unit(ucred->pid, &t) >= 0) {
 602                         unit = strappend("_SYSTEMD_UNIT=", t);
 603                         free(t);
 604                 } else if (unit_id)
 605                         unit = strappend("_SYSTEMD_UNIT=", unit_id);
 606
 607                 if (unit)
 608                         IOVEC_SET_STRING(iovec[n++], unit);
 609
 610 #ifdef HAVE_SELINUX
 611                 if (label) {
 612                         selinux_context = malloc(sizeof("_SELINUX_CONTEXT=") + label_len);
 613                         if (selinux_context) {
 614                                 memcpy(selinux_context, "_SELINUX_CONTEXT=", sizeof("_SELINUX_CONTEXT=")-1);
 615                                 memcpy(selinux_context+sizeof("_SELINUX_CONTEXT=")-1, label, label_len);
 616                                 selinux_context[sizeof("_SELINUX_CONTEXT=")-1+label_len] = 0;
 617                                 IOVEC_SET_STRING(iovec[n++], selinux_context);
 618                         }
 619                 } else {
 620                         security_context_t con;
 621
 622                         if (getpidcon(ucred->pid, &con) >= 0) {
 623                                 selinux_context = strappend("_SELINUX_CONTEXT=", con);
 624                                 if (selinux_context)
 625                                         IOVEC_SET_STRING(iovec[n++], selinux_context);
 626
 627                                 freecon(con);
 628                         }
 629                 }
 630 #endif
 631         }
 632
 633         if (tv) {
 634                 if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
 635                              (unsigned long long) timeval_load(tv)) >= 0)
 636                         IOVEC_SET_STRING(iovec[n++], source_time);
 637         }
 638
 639         /* Note that strictly speaking storing the boot id here is
 640          * redundant since the entry includes this in-line
 641          * anyway. However, we need this indexed, too. */
 642         r = sd_id128_get_boot(&id);
 643         if (r >= 0)
 644                 if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
 645                         IOVEC_SET_STRING(iovec[n++], boot_id);
 646
 647         r = sd_id128_get_machine(&id);
 648         if (r >= 0)
 649                 if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
 650                         IOVEC_SET_STRING(iovec[n++], machine_id);
 651
 652         t = gethostname_malloc();
 653         if (t) {
 654                 hostname = strappend("_HOSTNAME=", t);
 655                 free(t);
 656                 if (hostname)
 657                         IOVEC_SET_STRING(iovec[n++], hostname);
 658         }
 659
 660         assert(n <= m);
 661
 662         write_to_journal(s, realuid == 0 ? 0 : loginuid, iovec, n);
 663
 664         free(pid);
 665         free(uid);
 666         free(gid);
 667         free(comm);
 668         free(exe);
 669         free(cmdline);
 670         free(source_time);
 671         free(boot_id);
 672         free(machine_id);
 673         free(hostname);
 674         free(audit_session);
 675         free(audit_loginuid);
 676         free(cgroup);
 677         free(session);
 678         free(owner_uid);
 679         free(unit);
 680         free(selinux_context);
 681 }
 682
 683 void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) {
 684         char mid[11 + 32 + 1];
 685         char buffer[16 + LINE_MAX + 1];
 686         struct iovec iovec[N_IOVEC_META_FIELDS + 4];
 687         int n = 0;
 688         va_list ap;
 689         struct ucred ucred;
 690
 691         assert(s);
 692         assert(format);
 693
 694         IOVEC_SET_STRING(iovec[n++], "PRIORITY=6");
 695         IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver");
 696
 697         memcpy(buffer, "MESSAGE=", 8);
 698         va_start(ap, format);
 699         vsnprintf(buffer + 8, sizeof(buffer) - 8, format, ap);
 700         va_end(ap);
 701         char_array_0(buffer);
 702         IOVEC_SET_STRING(iovec[n++], buffer);
 703
 704         snprintf(mid, sizeof(mid), "MESSAGE_ID=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(message_id));
 705         char_array_0(mid);
 706         IOVEC_SET_STRING(iovec[n++], mid);
 707
 708         zero(ucred);
 709         ucred.pid = getpid();
 710         ucred.uid = getuid();
 711         ucred.gid = getgid();
 712
 713         dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL);
 714 }
 715
 716 void server_dispatch_message(
 717                 Server *s,
 718                 struct iovec *iovec, unsigned n, unsigned m,
 719                 struct ucred *ucred,
 720                 struct timeval *tv,
 721                 const char *label, size_t label_len,
 722                 const char *unit_id,
 723                 int priority) {
 724
 725         int rl;
 726         char *path = NULL, *c;
 727
 728         assert(s);
 729         assert(iovec || n == 0);
 730
 731         if (n == 0)
 732                 return;
 733
 734         if (LOG_PRI(priority) > s->max_level_store)
 735                 return;
 736
 737         if (!ucred)
 738                 goto finish;
 739
 740         path = shortened_cgroup_path(ucred->pid);
 741         if (!path)
 742                 goto finish;
 743
 744         /* example: /user/lennart/3/foobar
 745          *          /system/dbus.service/foobar
 746          *
 747          * So let's cut of everything past the third /, since that is
 748          * wher user directories start */
 749
 750         c = strchr(path, '/');
 751         if (c) {
 752                 c = strchr(c+1, '/');
 753                 if (c) {
 754                         c = strchr(c+1, '/');
 755                         if (c)
 756                                 *c = 0;
 757                 }
 758         }
 759
 760         rl = journal_rate_limit_test(s->rate_limit, path, priority & LOG_PRIMASK, available_space(s));
 761
 762         if (rl == 0) {
 763                 free(path);
 764                 return;
 765         }
 766
 767         /* Write a suppression message if we suppressed something */
 768         if (rl > 1)
 769                 server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, "Suppressed %u messages from %s", rl - 1, path);
 770
 771         free(path);
 772
 773 finish:
 774         dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id);
 775 }
 776
 777
 778 static int system_journal_open(Server *s) {
 779         int r;
 780         char *fn;
 781         sd_id128_t machine;
 782         char ids[33];
 783
 784         r = sd_id128_get_machine(&machine);
 785         if (r < 0)
 786                 return r;
 787
 788         sd_id128_to_string(machine, ids);
 789
 790         if (!s->system_journal &&
 791             (s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) &&
 792             access("/run/systemd/journal/flushed", F_OK) >= 0) {
 793
 794                 /* If in auto mode: first try to create the machine
 795                  * path, but not the prefix.
 796                  *
 797                  * If in persistent mode: create /var/log/journal and
 798                  * the machine path */
 799
 800                 if (s->storage == STORAGE_PERSISTENT)
 801                         (void) mkdir("/var/log/journal/", 0755);
 802
 803                 fn = strappend("/var/log/journal/", ids);
 804                 if (!fn)
 805                         return -ENOMEM;
 806
 807                 (void) mkdir(fn, 0755);
 808                 free(fn);
 809
 810                 fn = strjoin("/var/log/journal/", ids, "/system.journal", NULL);
 811                 if (!fn)
 812                         return -ENOMEM;
 813
 814                 r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &s->system_journal);
 815                 free(fn);
 816
 817                 if (r >= 0)
 818                         server_fix_perms(s, s->system_journal, 0);
 819                 else if (r < 0) {
 820
 821                         if (r != -ENOENT && r != -EROFS)
 822                                 log_warning("Failed to open system journal: %s", strerror(-r));
 823
 824                         r = 0;
 825                 }
 826         }
 827
 828         if (!s->runtime_journal &&
 829             (s->storage != STORAGE_NONE)) {
 830
 831                 fn = strjoin("/run/log/journal/", ids, "/system.journal", NULL);
 832                 if (!fn)
 833                         return -ENOMEM;
 834
 835                 if (s->system_journal) {
 836
 837                         /* Try to open the runtime journal, but only
 838                          * if it already exists, so that we can flush
 839                          * it into the system journal */
 840
 841                         r = journal_file_open(fn, O_RDWR, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal);
 842                         free(fn);
 843
 844                         if (r < 0) {
 845                                 if (r != -ENOENT)
 846                                         log_warning("Failed to open runtime journal: %s", strerror(-r));
 847
 848                                 r = 0;
 849                         }
 850
 851                 } else {
 852
 853                         /* OK, we really need the runtime journal, so create
 854                          * it if necessary. */
 855
 856                         (void) mkdir_parents(fn, 0755);
 857                         r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal);
 858                         free(fn);
 859
 860                         if (r < 0) {
 861                                 log_error("Failed to open runtime journal: %s", strerror(-r));
 862                                 return r;
 863                         }
 864                 }
 865
 866                 if (s->runtime_journal)
 867                         server_fix_perms(s, s->runtime_journal, 0);
 868         }
 869
 870         return r;
 871 }
 872
 873 static int server_flush_to_var(Server *s) {
 874         Object *o = NULL;
 875         int r;
 876         sd_id128_t machine;
 877         sd_journal *j;
 878
 879         assert(s);
 880
 881         if (s->storage != STORAGE_AUTO &&
 882             s->storage != STORAGE_PERSISTENT)
 883                 return 0;
 884
 885         if (!s->runtime_journal)
 886                 return 0;
 887
 888         system_journal_open(s);
 889
 890         if (!s->system_journal)
 891                 return 0;
 892
 893         log_info("Flushing to /var...");
 894
 895         r = sd_id128_get_machine(&machine);
 896         if (r < 0) {
 897                 log_error("Failed to get machine id: %s", strerror(-r));
 898                 return r;
 899         }
 900
 901         r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY);
 902         if (r < 0) {
 903                 log_error("Failed to read runtime journal: %s", strerror(-r));
 904                 return r;
 905         }
 906
 907         SD_JOURNAL_FOREACH(j) {
 908                 JournalFile *f;
 909
 910                 f = j->current_file;
 911                 assert(f && f->current_offset > 0);
 912
 913                 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
 914                 if (r < 0) {
 915                         log_error("Can't read entry: %s", strerror(-r));
 916                         goto finish;
 917                 }
 918
 919                 r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
 920                 if (r == -E2BIG) {
 921                         log_info("Allocation limit reached.");
 922
 923                         journal_file_post_change(s->system_journal);
 924                         server_rotate(s);
 925                         server_vacuum(s);
 926
 927                         r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
 928                 }
 929
 930                 if (r < 0) {
 931                         log_error("Can't write entry: %s", strerror(-r));
 932                         goto finish;
 933                 }
 934         }
 935
 936 finish:
 937         journal_file_post_change(s->system_journal);
 938
 939         journal_file_close(s->runtime_journal);
 940         s->runtime_journal = NULL;
 941
 942         if (r >= 0)
 943                 rm_rf("/run/log/journal", false, true, false);
 944
 945         return r;
 946 }
 947
 948 static int process_event(Server *s, struct epoll_event *ev) {
 949         assert(s);
 950         assert(ev);
 951
 952         if (ev->data.fd == s->signal_fd) {
 953                 struct signalfd_siginfo sfsi;
 954                 ssize_t n;
 955
 956                 if (ev->events != EPOLLIN) {
 957                         log_info("Got invalid event from epoll.");
 958                         return -EIO;
 959                 }
 960
 961                 n = read(s->signal_fd, &sfsi, sizeof(sfsi));
 962                 if (n != sizeof(sfsi)) {
 963
 964                         if (n >= 0)
 965                                 return -EIO;
 966
 967                         if (errno == EINTR || errno == EAGAIN)
 968                                 return 1;
 969
 970                         return -errno;
 971                 }
 972
 973                 log_info("Received SIG%s", signal_to_string(sfsi.ssi_signo));
 974
 975                 if (sfsi.ssi_signo == SIGUSR1) {
 976                         touch("/run/systemd/journal/flushed");
 977                         server_flush_to_var(s);
 978                         return 1;
 979                 }
 980
 981                 if (sfsi.ssi_signo == SIGUSR2) {
 982                         server_rotate(s);
 983                         server_vacuum(s);
 984                         return 1;
 985                 }
 986
 987                 return 0;
 988
 989         } else if (ev->data.fd == s->dev_kmsg_fd) {
 990                 int r;
 991
 992                 if (ev->events != EPOLLIN) {
 993                         log_info("Got invalid event from epoll.");
 994                         return -EIO;
 995                 }
 996
 997                 r = server_read_dev_kmsg(s);
 998                 if (r < 0)
 999                         return r;
1000
1001                 return 1;
1002
1003         } else if (ev->data.fd == s->native_fd ||
1004                    ev->data.fd == s->syslog_fd) {
1005
1006                 if (ev->events != EPOLLIN) {
1007                         log_info("Got invalid event from epoll.");
1008                         return -EIO;
1009                 }
1010
1011                 for (;;) {
1012                         struct msghdr msghdr;
1013                         struct iovec iovec;
1014                         struct ucred *ucred = NULL;
1015                         struct timeval *tv = NULL;
1016                         struct cmsghdr *cmsg;
1017                         char *label = NULL;
1018                         size_t label_len = 0;
1019                         union {
1020                                 struct cmsghdr cmsghdr;
1021
1022                                 /* We use NAME_MAX space for the
1023                                  * SELinux label here. The kernel
1024                                  * currently enforces no limit, but
1025                                  * according to suggestions from the
1026                                  * SELinux people this will change and
1027                                  * it will probably be identical to
1028                                  * NAME_MAX. For now we use that, but
1029                                  * this should be updated one day when
1030                                  * the final limit is known.*/
1031                                 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
1032                                             CMSG_SPACE(sizeof(struct timeval)) +
1033                                             CMSG_SPACE(sizeof(int)) + /* fd */
1034                                             CMSG_SPACE(NAME_MAX)]; /* selinux label */
1035                         } control;
1036                         ssize_t n;
1037                         int v;
1038                         int *fds = NULL;
1039                         unsigned n_fds = 0;
1040
1041                         if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
1042                                 log_error("SIOCINQ failed: %m");
1043                                 return -errno;
1044                         }
1045
1046                         if (s->buffer_size < (size_t) v) {
1047                                 void *b;
1048                                 size_t l;
1049
1050                                 l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
1051                                 b = realloc(s->buffer, l+1);
1052
1053                                 if (!b) {
1054                                         log_error("Couldn't increase buffer.");
1055                                         return -ENOMEM;
1056                                 }
1057
1058                                 s->buffer_size = l;
1059                                 s->buffer = b;
1060                         }
1061
1062                         zero(iovec);
1063                         iovec.iov_base = s->buffer;
1064                         iovec.iov_len = s->buffer_size;
1065
1066                         zero(control);
1067                         zero(msghdr);
1068                         msghdr.msg_iov = &iovec;
1069                         msghdr.msg_iovlen = 1;
1070                         msghdr.msg_control = &control;
1071                         msghdr.msg_controllen = sizeof(control);
1072
1073                         n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
1074                         if (n < 0) {
1075
1076                                 if (errno == EINTR || errno == EAGAIN)
1077                                         return 1;
1078
1079                                 log_error("recvmsg() failed: %m");
1080                                 return -errno;
1081                         }
1082
1083                         for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) {
1084
1085                                 if (cmsg->cmsg_level == SOL_SOCKET &&
1086                                     cmsg->cmsg_type == SCM_CREDENTIALS &&
1087                                     cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)))
1088                                         ucred = (struct ucred*) CMSG_DATA(cmsg);
1089                                 else if (cmsg->cmsg_level == SOL_SOCKET &&
1090                                          cmsg->cmsg_type == SCM_SECURITY) {
1091                                         label = (char*) CMSG_DATA(cmsg);
1092                                         label_len = cmsg->cmsg_len - CMSG_LEN(0);
1093                                 } else if (cmsg->cmsg_level == SOL_SOCKET &&
1094                                          cmsg->cmsg_type == SO_TIMESTAMP &&
1095                                          cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval)))
1096                                         tv = (struct timeval*) CMSG_DATA(cmsg);
1097                                 else if (cmsg->cmsg_level == SOL_SOCKET &&
1098                                          cmsg->cmsg_type == SCM_RIGHTS) {
1099                                         fds = (int*) CMSG_DATA(cmsg);
1100                                         n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int);
1101                                 }
1102                         }
1103
1104                         if (ev->data.fd == s->syslog_fd) {
1105                                 char *e;
1106
1107                                 if (n > 0 && n_fds == 0) {
1108                                         e = memchr(s->buffer, '\n', n);
1109                                         if (e)
1110                                                 *e = 0;
1111                                         else
1112                                                 s->buffer[n] = 0;
1113
1114                                         server_process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len);
1115                                 } else if (n_fds > 0)
1116                                         log_warning("Got file descriptors via syslog socket. Ignoring.");
1117
1118                         } else {
1119                                 if (n > 0 && n_fds == 0)
1120                                         server_process_native_message(s, s->buffer, n, ucred, tv, label, label_len);
1121                                 else if (n == 0 && n_fds == 1)
1122                                         server_process_native_file(s, fds[0], ucred, tv, label, label_len);
1123                                 else if (n_fds > 0)
1124                                         log_warning("Got too many file descriptors via native socket. Ignoring.");
1125                         }
1126
1127                         close_many(fds, n_fds);
1128                 }
1129
1130                 return 1;
1131
1132         } else if (ev->data.fd == s->stdout_fd) {
1133
1134                 if (ev->events != EPOLLIN) {
1135                         log_info("Got invalid event from epoll.");
1136                         return -EIO;
1137                 }
1138
1139                 stdout_stream_new(s);
1140                 return 1;
1141
1142         } else {
1143                 StdoutStream *stream;
1144
1145                 if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
1146                         log_info("Got invalid event from epoll.");
1147                         return -EIO;
1148                 }
1149
1150                 /* If it is none of the well-known fds, it must be an
1151                  * stdout stream fd. Note that this is a bit ugly here
1152                  * (since we rely that none of the well-known fds
1153                  * could be interpreted as pointer), but nonetheless
1154                  * safe, since the well-known fds would never get an
1155                  * fd > 4096, i.e. beyond the first memory page */
1156
1157                 stream = ev->data.ptr;
1158
1159                 if (stdout_stream_process(stream) <= 0)
1160                         stdout_stream_free(stream);
1161
1162                 return 1;
1163         }
1164
1165         log_error("Unknown event.");
1166         return 0;
1167 }
1168
1169 static int open_signalfd(Server *s) {
1170         sigset_t mask;
1171         struct epoll_event ev;
1172
1173         assert(s);
1174
1175         assert_se(sigemptyset(&mask) == 0);
1176         sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, -1);
1177         assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
1178
1179         s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
1180         if (s->signal_fd < 0) {
1181                 log_error("signalfd(): %m");
1182                 return -errno;
1183         }
1184
1185         zero(ev);
1186         ev.events = EPOLLIN;
1187         ev.data.fd = s->signal_fd;
1188
1189         if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) {
1190                 log_error("epoll_ctl(): %m");
1191                 return -errno;
1192         }
1193
1194         return 0;
1195 }
1196
1197 static int server_parse_proc_cmdline(Server *s) {
1198         char *line, *w, *state;
1199         int r;
1200         size_t l;
1201
1202         if (detect_container(NULL) > 0)
1203                 return 0;
1204
1205         r = read_one_line_file("/proc/cmdline", &line);
1206         if (r < 0) {
1207                 log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r));
1208                 return 0;
1209         }
1210
1211         FOREACH_WORD_QUOTED(w, l, line, state) {
1212                 char *word;
1213
1214                 word = strndup(w, l);
1215                 if (!word) {
1216                         r = -ENOMEM;
1217                         goto finish;
1218                 }
1219
1220                 if (startswith(word, "systemd.journald.forward_to_syslog=")) {
1221                         r = parse_boolean(word + 35);
1222                         if (r < 0)
1223                                 log_warning("Failed to parse forward to syslog switch %s. Ignoring.", word + 35);
1224                         else
1225                                 s->forward_to_syslog = r;
1226                 } else if (startswith(word, "systemd.journald.forward_to_kmsg=")) {
1227                         r = parse_boolean(word + 33);
1228                         if (r < 0)
1229                                 log_warning("Failed to parse forward to kmsg switch %s. Ignoring.", word + 33);
1230                         else
1231                                 s->forward_to_kmsg = r;
1232                 } else if (startswith(word, "systemd.journald.forward_to_console=")) {
1233                         r = parse_boolean(word + 36);
1234                         if (r < 0)
1235                                 log_warning("Failed to parse forward to console switch %s. Ignoring.", word + 36);
1236                         else
1237                                 s->forward_to_console = r;
1238                 } else if (startswith(word, "systemd.journald"))
1239                         log_warning("Invalid systemd.journald parameter. Ignoring.");
1240
1241                 free(word);
1242         }
1243
1244         r = 0;
1245
1246 finish:
1247         free(line);
1248         return r;
1249 }
1250
1251 static int server_parse_config_file(Server *s) {
1252         FILE *f;
1253         const char *fn;
1254         int r;
1255
1256         assert(s);
1257
1258         fn = "/etc/systemd/journald.conf";
1259         f = fopen(fn, "re");
1260         if (!f) {
1261                 if (errno == ENOENT)
1262                         return 0;
1263
1264                 log_warning("Failed to open configuration file %s: %m", fn);
1265                 return -errno;
1266         }
1267
1268         r = config_parse(fn, f, "Journal\0", config_item_perf_lookup, (void*) journald_gperf_lookup, false, s);
1269         if (r < 0)
1270                 log_warning("Failed to parse configuration file: %s", strerror(-r));
1271
1272         fclose(f);
1273
1274         return r;
1275 }
1276
1277 static int server_init(Server *s) {
1278         int n, r, fd;
1279
1280         assert(s);
1281
1282         zero(*s);
1283         s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = s->dev_kmsg_fd = -1;
1284         s->compress = true;
1285         s->seal = true;
1286
1287         s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL;
1288         s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST;
1289
1290         s->forward_to_syslog = true;
1291
1292         s->max_level_store = LOG_DEBUG;
1293         s->max_level_syslog = LOG_DEBUG;
1294         s->max_level_kmsg = LOG_NOTICE;
1295         s->max_level_console = LOG_INFO;
1296
1297         memset(&s->system_metrics, 0xFF, sizeof(s->system_metrics));
1298         memset(&s->runtime_metrics, 0xFF, sizeof(s->runtime_metrics));
1299
1300         server_parse_config_file(s);
1301         server_parse_proc_cmdline(s);
1302
1303         mkdir_p("/run/systemd/journal", 0755);
1304
1305         s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
1306         if (!s->user_journals)
1307                 return log_oom();
1308
1309         s->mmap = mmap_cache_new();
1310         if (!s->mmap)
1311                 return log_oom();
1312
1313         s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
1314         if (s->epoll_fd < 0) {
1315                 log_error("Failed to create epoll object: %m");
1316                 return -errno;
1317         }
1318
1319         n = sd_listen_fds(true);
1320         if (n < 0) {
1321                 log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
1322                 return n;
1323         }
1324
1325         for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) {
1326
1327                 if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) {
1328
1329                         if (s->native_fd >= 0) {
1330                                 log_error("Too many native sockets passed.");
1331                                 return -EINVAL;
1332                         }
1333
1334                         s->native_fd = fd;
1335
1336                 } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) {
1337
1338                         if (s->stdout_fd >= 0) {
1339                                 log_error("Too many stdout sockets passed.");
1340                                 return -EINVAL;
1341                         }
1342
1343                         s->stdout_fd = fd;
1344
1345                 } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
1346
1347                         if (s->syslog_fd >= 0) {
1348                                 log_error("Too many /dev/log sockets passed.");
1349                                 return -EINVAL;
1350                         }
1351
1352                         s->syslog_fd = fd;
1353
1354                 } else {
1355                         log_error("Unknown socket passed.");
1356                         return -EINVAL;
1357                 }
1358         }
1359
1360         r = server_open_syslog_socket(s);
1361         if (r < 0)
1362                 return r;
1363
1364         r = server_open_native_socket(s);
1365         if (r < 0)
1366                 return r;
1367
1368         r = server_open_stdout_socket(s);
1369         if (r < 0)
1370                 return r;
1371
1372         r = server_open_dev_kmsg(s);
1373         if (r < 0)
1374                 return r;
1375
1376         r = server_open_kernel_seqnum(s);
1377         if (r < 0)
1378                 return r;
1379
1380         r = open_signalfd(s);
1381         if (r < 0)
1382                 return r;
1383
1384         s->udev = udev_new();
1385         if (!s->udev)
1386                 return -ENOMEM;
1387
1388         s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst);
1389         if (!s->rate_limit)
1390                 return -ENOMEM;
1391
1392         r = system_journal_open(s);
1393         if (r < 0)
1394                 return r;
1395
1396         return 0;
1397 }
1398
1399 static void server_maybe_append_tags(Server *s) {
1400 #ifdef HAVE_GCRYPT
1401         JournalFile *f;
1402         Iterator i;
1403         usec_t n;
1404
1405         n = now(CLOCK_REALTIME);
1406
1407         if (s->system_journal)
1408                 journal_file_maybe_append_tag(s->system_journal, n);
1409
1410         HASHMAP_FOREACH(f, s->user_journals, i)
1411                 journal_file_maybe_append_tag(f, n);
1412 #endif
1413 }
1414
1415 static void server_done(Server *s) {
1416         JournalFile *f;
1417         assert(s);
1418
1419         while (s->stdout_streams)
1420                 stdout_stream_free(s->stdout_streams);
1421
1422         if (s->system_journal)
1423                 journal_file_close(s->system_journal);
1424
1425         if (s->runtime_journal)
1426                 journal_file_close(s->runtime_journal);
1427
1428         while ((f = hashmap_steal_first(s->user_journals)))
1429                 journal_file_close(f);
1430
1431         hashmap_free(s->user_journals);
1432
1433         if (s->epoll_fd >= 0)
1434                 close_nointr_nofail(s->epoll_fd);
1435
1436         if (s->signal_fd >= 0)
1437                 close_nointr_nofail(s->signal_fd);
1438
1439         if (s->syslog_fd >= 0)
1440                 close_nointr_nofail(s->syslog_fd);
1441
1442         if (s->native_fd >= 0)
1443                 close_nointr_nofail(s->native_fd);
1444
1445         if (s->stdout_fd >= 0)
1446                 close_nointr_nofail(s->stdout_fd);
1447
1448         if (s->dev_kmsg_fd >= 0)
1449                 close_nointr_nofail(s->dev_kmsg_fd);
1450
1451         if (s->rate_limit)
1452                 journal_rate_limit_free(s->rate_limit);
1453
1454         if (s->kernel_seqnum)
1455                 munmap(s->kernel_seqnum, sizeof(uint64_t));
1456
1457         free(s->buffer);
1458         free(s->tty_path);
1459
1460         if (s->mmap)
1461                 mmap_cache_unref(s->mmap);
1462
1463         if (s->udev)
1464                 udev_unref(s->udev);
1465 }
1466
1467 int main(int argc, char *argv[]) {
1468         Server server;
1469         int r;
1470
1471         /* if (getppid() != 1) { */
1472         /*         log_error("This program should be invoked by init only."); */
1473         /*         return EXIT_FAILURE; */
1474         /* } */
1475
1476         if (argc > 1) {
1477                 log_error("This program does not take arguments.");
1478                 return EXIT_FAILURE;
1479         }
1480
1481         log_set_target(LOG_TARGET_SAFE);
1482         log_set_facility(LOG_SYSLOG);
1483         log_set_max_level(LOG_DEBUG);
1484         log_parse_environment();
1485         log_open();
1486
1487         umask(0022);
1488
1489         r = server_init(&server);
1490         if (r < 0)
1491                 goto finish;
1492
1493         server_vacuum(&server);
1494         server_flush_to_var(&server);
1495         server_flush_dev_kmsg(&server);
1496
1497         log_debug("systemd-journald running as pid %lu", (unsigned long) getpid());
1498         server_driver_message(&server, SD_MESSAGE_JOURNAL_START, "Journal started");
1499
1500         sd_notify(false,
1501                   "READY=1\n"
1502                   "STATUS=Processing requests...");
1503
1504         for (;;) {
1505                 struct epoll_event event;
1506                 int t;
1507
1508 #ifdef HAVE_GCRYPT
1509                 usec_t u;
1510
1511                 if (server.system_journal &&
1512                     journal_file_next_evolve_usec(server.system_journal, &u)) {
1513                         usec_t n;
1514
1515                         n = now(CLOCK_REALTIME);
1516
1517                         if (n >= u)
1518                                 t = 0;
1519                         else
1520                                 t = (int) ((u - n + USEC_PER_MSEC - 1) / USEC_PER_MSEC);
1521                 } else
1522 #endif
1523                         t = -1;
1524
1525                 r = epoll_wait(server.epoll_fd, &event, 1, t);
1526                 if (r < 0) {
1527
1528                         if (errno == EINTR)
1529                                 continue;
1530
1531                         log_error("epoll_wait() failed: %m");
1532                         r = -errno;
1533                         goto finish;
1534                 }
1535
1536                 if (r > 0) {
1537                         r = process_event(&server, &event);
1538                         if (r < 0)
1539                                 goto finish;
1540                         else if (r == 0)
1541                                 break;
1542                 }
1543
1544                 server_maybe_append_tags(&server);
1545         }
1546
1547         log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
1548         server_driver_message(&server, SD_MESSAGE_JOURNAL_STOP, "Journal stopped");
1549
1550 finish:
1551         sd_notify(false,
1552                   "STATUS=Shutting down...");
1553
1554         server_done(&server);
1555
1556         return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
1557 }