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 General Public License as published by
  10   the Free Software Foundation; either version 2 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   General Public License for more details.
  17
  18   You should have received a copy of the GNU 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 <sys/acl.h>
  29 #include <acl/libacl.h>
  30 #include <stddef.h>
  31 #include <sys/ioctl.h>
  32 #include <linux/sockios.h>
  33 #include <sys/statvfs.h>
  34
  35 #include "hashmap.h"
  36 #include "journal-file.h"
  37 #include "sd-daemon.h"
  38 #include "socket-util.h"
  39 #include "acl-util.h"
  40 #include "cgroup-util.h"
  41 #include "list.h"
  42 #include "journal-rate-limit.h"
  43 #include "sd-journal.h"
  44 #include "journal-internal.h"
  45
  46 #define USER_JOURNALS_MAX 1024
  47 #define STDOUT_STREAMS_MAX 4096
  48
  49 #define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC)
  50 #define DEFAULT_RATE_LIMIT_BURST 200
  51
  52 #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
  53
  54 #define RECHECK_VAR_AVAILABLE_USEC (30*USEC_PER_SEC)
  55
  56 #define SYSLOG_TIMEOUT_USEC (5*USEC_PER_SEC)
  57
  58 typedef struct StdoutStream StdoutStream;
  59
  60 typedef struct Server {
  61         int epoll_fd;
  62         int signal_fd;
  63         int syslog_fd;
  64         int native_fd;
  65         int stdout_fd;
  66
  67         JournalFile *runtime_journal;
  68         JournalFile *system_journal;
  69         Hashmap *user_journals;
  70
  71         uint64_t seqnum;
  72
  73         char *buffer;
  74         size_t buffer_size;
  75
  76         JournalRateLimit *rate_limit;
  77
  78         JournalMetrics metrics;
  79         uint64_t max_use;
  80         bool compress;
  81
  82         uint64_t cached_available_space;
  83         usec_t cached_available_space_timestamp;
  84
  85         uint64_t var_available_timestamp;
  86
  87         LIST_HEAD(StdoutStream, stdout_streams);
  88         unsigned n_stdout_streams;
  89 } Server;
  90
  91 typedef enum StdoutStreamState {
  92         STDOUT_STREAM_TAG,
  93         STDOUT_STREAM_PRIORITY,
  94         STDOUT_STREAM_PRIORITY_PREFIX,
  95         STDOUT_STREAM_TEE_CONSOLE,
  96         STDOUT_STREAM_RUNNING
  97 } StdoutStreamState;
  98
  99 struct StdoutStream {
 100         Server *server;
 101         StdoutStreamState state;
 102
 103         int fd;
 104
 105         struct ucred ucred;
 106
 107         char *tag;
 108         int priority;
 109         bool priority_prefix:1;
 110         bool tee_console:1;
 111
 112         char buffer[LINE_MAX+1];
 113         size_t length;
 114
 115         LIST_FIELDS(StdoutStream, stdout_stream);
 116 };
 117
 118 static int server_flush_to_var(Server *s);
 119
 120 static uint64_t available_space(Server *s) {
 121         char ids[33];
 122         sd_id128_t machine;
 123         char *p;
 124         const char *f;
 125         struct statvfs ss;
 126         uint64_t sum = 0, avail = 0, ss_avail = 0;
 127         int r;
 128         DIR *d;
 129         usec_t ts = now(CLOCK_MONOTONIC);
 130
 131         if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts)
 132                 return s->cached_available_space;
 133
 134         r = sd_id128_get_machine(&machine);
 135         if (r < 0)
 136                 return 0;
 137
 138         if (s->system_journal)
 139                 f = "/var/log/journal/";
 140         else
 141                 f = "/run/log/journal/";
 142
 143         p = strappend(f, sd_id128_to_string(machine, ids));
 144         if (!p)
 145                 return 0;
 146
 147         d = opendir(p);
 148         free(p);
 149
 150         if (!d)
 151                 return 0;
 152
 153         if (fstatvfs(dirfd(d), &ss) < 0)
 154                 goto finish;
 155
 156         for (;;) {
 157                 struct stat st;
 158                 struct dirent buf, *de;
 159                 int k;
 160
 161                 k = readdir_r(d, &buf, &de);
 162                 if (k != 0) {
 163                         r = -k;
 164                         goto finish;
 165                 }
 166
 167                 if (!de)
 168                         break;
 169
 170                 if (!dirent_is_file_with_suffix(de, ".journal"))
 171                         continue;
 172
 173                 if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
 174                         continue;
 175
 176                 sum += (uint64_t) st.st_blocks * (uint64_t) st.st_blksize;
 177         }
 178
 179         avail = sum >= s->max_use ? 0 : s->max_use - sum;
 180
 181         ss_avail = ss.f_bsize * ss.f_bavail;
 182
 183         ss_avail = ss_avail < s->metrics.keep_free ? 0 : ss_avail - s->metrics.keep_free;
 184
 185         if (ss_avail < avail)
 186                 avail = ss_avail;
 187
 188         s->cached_available_space = avail;
 189         s->cached_available_space_timestamp = ts;
 190
 191 finish:
 192         closedir(d);
 193
 194         return avail;
 195 }
 196
 197 static void fix_perms(JournalFile *f, uid_t uid) {
 198         acl_t acl;
 199         acl_entry_t entry;
 200         acl_permset_t permset;
 201         int r;
 202
 203         assert(f);
 204
 205         r = fchmod_and_fchown(f->fd, 0640, 0, 0);
 206         if (r < 0)
 207                 log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
 208
 209         if (uid <= 0)
 210                 return;
 211
 212         acl = acl_get_fd(f->fd);
 213         if (!acl) {
 214                 log_warning("Failed to read ACL on %s, ignoring: %m", f->path);
 215                 return;
 216         }
 217
 218         r = acl_find_uid(acl, uid, &entry);
 219         if (r <= 0) {
 220
 221                 if (acl_create_entry(&acl, &entry) < 0 ||
 222                     acl_set_tag_type(entry, ACL_USER) < 0 ||
 223                     acl_set_qualifier(entry, &uid) < 0) {
 224                         log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
 225                         goto finish;
 226                 }
 227         }
 228
 229         if (acl_get_permset(entry, &permset) < 0 ||
 230             acl_add_perm(permset, ACL_READ) < 0 ||
 231             acl_calc_mask(&acl) < 0) {
 232                 log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);
 233                 goto finish;
 234         }
 235
 236         if (acl_set_fd(f->fd, acl) < 0)
 237                 log_warning("Failed to set ACL on %s, ignoring: %m", f->path);
 238
 239 finish:
 240         acl_free(acl);
 241 }
 242
 243 static JournalFile* find_journal(Server *s, uid_t uid) {
 244         char *p;
 245         int r;
 246         JournalFile *f;
 247         char ids[33];
 248         sd_id128_t machine;
 249
 250         assert(s);
 251
 252         /* We split up user logs only on /var, not on /run. If the
 253          * runtime file is open, we write to it exclusively, in order
 254          * to guarantee proper order as soon as we flush /run to
 255          * /var and close the runtime file. */
 256
 257         if (s->runtime_journal)
 258                 return s->runtime_journal;
 259
 260         if (uid <= 0)
 261                 return s->system_journal;
 262
 263         r = sd_id128_get_machine(&machine);
 264         if (r < 0)
 265                 return s->system_journal;
 266
 267         f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid));
 268         if (f)
 269                 return f;
 270
 271         if (asprintf(&p, "/var/log/journal/%s/user-%lu.journal", sd_id128_to_string(machine, ids), (unsigned long) uid) < 0)
 272                 return s->system_journal;
 273
 274         while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
 275                 /* Too many open? Then let's close one */
 276                 f = hashmap_steal_first(s->user_journals);
 277                 assert(f);
 278                 journal_file_close(f);
 279         }
 280
 281         r = journal_file_open(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
 282         free(p);
 283
 284         if (r < 0)
 285                 return s->system_journal;
 286
 287         fix_perms(f, uid);
 288         f->metrics = s->metrics;
 289         f->compress = s->compress;
 290
 291         r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
 292         if (r < 0) {
 293                 journal_file_close(f);
 294                 return s->system_journal;
 295         }
 296
 297         return f;
 298 }
 299
 300 static void server_vacuum(Server *s) {
 301         Iterator i;
 302         void *k;
 303         char *p;
 304         char ids[33];
 305         sd_id128_t machine;
 306         int r;
 307         JournalFile *f;
 308
 309         log_info("Rotating...");
 310
 311         if (s->runtime_journal) {
 312                 r = journal_file_rotate(&s->runtime_journal);
 313                 if (r < 0)
 314                         log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
 315         }
 316
 317         if (s->system_journal) {
 318                 r = journal_file_rotate(&s->system_journal);
 319                 if (r < 0)
 320                         log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
 321         }
 322
 323         HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
 324                 r = journal_file_rotate(&f);
 325                 if (r < 0)
 326                         log_error("Failed to rotate %s: %s", f->path, strerror(-r));
 327                 else
 328                         hashmap_replace(s->user_journals, k, f);
 329         }
 330
 331         log_info("Vacuuming...");
 332
 333         r = sd_id128_get_machine(&machine);
 334         if (r < 0) {
 335                 log_error("Failed to get machine ID: %s", strerror(-r));
 336                 return;
 337         }
 338
 339         if (asprintf(&p, "/var/log/journal/%s", sd_id128_to_string(machine, ids)) < 0) {
 340                 log_error("Out of memory.");
 341                 return;
 342         }
 343
 344         r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
 345         if (r < 0 && r != -ENOENT)
 346                 log_error("Failed to vacuum %s: %s", p, strerror(-r));
 347         free(p);
 348
 349         if (asprintf(&p, "/run/log/journal/%s", ids) < 0) {
 350                 log_error("Out of memory.");
 351                 return;
 352         }
 353
 354         r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free);
 355         if (r < 0 && r != -ENOENT)
 356                 log_error("Failed to vacuum %s: %s", p, strerror(-r));
 357         free(p);
 358
 359         s->cached_available_space_timestamp = 0;
 360 }
 361
 362 static char *shortened_cgroup_path(pid_t pid) {
 363         int r;
 364         char *process_path, *init_path, *path;
 365
 366         assert(pid > 0);
 367
 368         r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path);
 369         if (r < 0)
 370                 return NULL;
 371
 372         r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path);
 373         if (r < 0) {
 374                 free(process_path);
 375                 return NULL;
 376         }
 377
 378         if (streq(init_path, "/"))
 379                 init_path[0] = 0;
 380
 381         if (startswith(process_path, init_path))
 382                 path = process_path + strlen(init_path);
 383         else
 384                 path = process_path;
 385
 386         free(init_path);
 387
 388         return path;
 389 }
 390
 391 static void dispatch_message_real(Server *s,
 392                              struct iovec *iovec, unsigned n, unsigned m,
 393                              struct ucred *ucred,
 394                              struct timeval *tv) {
 395
 396         char *pid = NULL, *uid = NULL, *gid = NULL,
 397                 *source_time = NULL, *boot_id = NULL, *machine_id = NULL,
 398                 *comm = NULL, *cmdline = NULL, *hostname = NULL,
 399                 *audit_session = NULL, *audit_loginuid = NULL,
 400                 *exe = NULL, *cgroup = NULL;
 401
 402         char idbuf[33];
 403         sd_id128_t id;
 404         int r;
 405         char *t;
 406         uid_t loginuid = 0, realuid = 0;
 407         JournalFile *f;
 408         bool vacuumed = false;
 409
 410         assert(s);
 411         assert(iovec);
 412         assert(n > 0);
 413         assert(n + 13 <= m);
 414
 415         if (ucred) {
 416                 uint32_t session;
 417                 char *path;
 418
 419                 realuid = ucred->uid;
 420
 421                 if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
 422                         IOVEC_SET_STRING(iovec[n++], pid);
 423
 424                 if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
 425                         IOVEC_SET_STRING(iovec[n++], uid);
 426
 427                 if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
 428                         IOVEC_SET_STRING(iovec[n++], gid);
 429
 430                 r = get_process_comm(ucred->pid, &t);
 431                 if (r >= 0) {
 432                         comm = strappend("_COMM=", t);
 433                         if (comm)
 434                                 IOVEC_SET_STRING(iovec[n++], comm);
 435                         free(t);
 436                 }
 437
 438                 r = get_process_exe(ucred->pid, &t);
 439                 if (r >= 0) {
 440                         exe = strappend("_EXE=", t);
 441                         if (comm)
 442                                 IOVEC_SET_STRING(iovec[n++], exe);
 443                         free(t);
 444                 }
 445
 446                 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
 447                 if (r >= 0) {
 448                         cmdline = strappend("_CMDLINE=", t);
 449                         if (cmdline)
 450                                 IOVEC_SET_STRING(iovec[n++], cmdline);
 451                         free(t);
 452                 }
 453
 454                 r = audit_session_from_pid(ucred->pid, &session);
 455                 if (r >= 0)
 456                         if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
 457                                 IOVEC_SET_STRING(iovec[n++], audit_session);
 458
 459                 r = audit_loginuid_from_pid(ucred->pid, &loginuid);
 460                 if (r >= 0)
 461                         if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
 462                                 IOVEC_SET_STRING(iovec[n++], audit_loginuid);
 463
 464                 path = shortened_cgroup_path(ucred->pid);
 465                 if (path) {
 466                         cgroup = strappend("_SYSTEMD_CGROUP=", path);
 467                         if (cgroup)
 468                                 IOVEC_SET_STRING(iovec[n++], cgroup);
 469
 470                         free(path);
 471                 }
 472         }
 473
 474         if (tv) {
 475                 if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
 476                              (unsigned long long) timeval_load(tv)) >= 0)
 477                         IOVEC_SET_STRING(iovec[n++], source_time);
 478         }
 479
 480         /* Note that strictly speaking storing the boot id here is
 481          * redundant since the entry includes this in-line
 482          * anyway. However, we need this indexed, too. */
 483         r = sd_id128_get_boot(&id);
 484         if (r >= 0)
 485                 if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
 486                         IOVEC_SET_STRING(iovec[n++], boot_id);
 487
 488         r = sd_id128_get_machine(&id);
 489         if (r >= 0)
 490                 if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
 491                         IOVEC_SET_STRING(iovec[n++], machine_id);
 492
 493         t = gethostname_malloc();
 494         if (t) {
 495                 hostname = strappend("_HOSTNAME=", t);
 496                 if (hostname)
 497                         IOVEC_SET_STRING(iovec[n++], hostname);
 498                 free(t);
 499         }
 500
 501         assert(n <= m);
 502
 503         server_flush_to_var(s);
 504
 505 retry:
 506         f = find_journal(s, realuid == 0 ? 0 : loginuid);
 507         if (!f)
 508                 log_warning("Dropping message, as we can't find a place to store the data.");
 509         else {
 510                 r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
 511
 512                 if (r == -E2BIG && !vacuumed) {
 513                         log_info("Allocation limit reached.");
 514
 515                         server_vacuum(s);
 516                         vacuumed = true;
 517
 518                         log_info("Retrying write.");
 519                         goto retry;
 520                 }
 521
 522                 if (r < 0)
 523                         log_error("Failed to write entry, ignoring: %s", strerror(-r));
 524         }
 525
 526         free(pid);
 527         free(uid);
 528         free(gid);
 529         free(comm);
 530         free(exe);
 531         free(cmdline);
 532         free(source_time);
 533         free(boot_id);
 534         free(machine_id);
 535         free(hostname);
 536         free(audit_session);
 537         free(audit_loginuid);
 538         free(cgroup);
 539 }
 540
 541 static void dispatch_message(Server *s,
 542                              struct iovec *iovec, unsigned n, unsigned m,
 543                              struct ucred *ucred,
 544                              struct timeval *tv,
 545                              int priority) {
 546         int rl;
 547         char *path, *c;
 548
 549         assert(s);
 550         assert(iovec || n == 0);
 551
 552         if (n == 0)
 553                 return;
 554
 555         if (!ucred)
 556                 goto finish;
 557
 558         path = shortened_cgroup_path(ucred->pid);
 559         if (!path)
 560                 goto finish;
 561
 562         /* example: /user/lennart/3/foobar
 563          *          /system/dbus.service/foobar
 564          *
 565          * So let's cut of everything past the third /, since that is
 566          * wher user directories start */
 567
 568         c = strchr(path, '/');
 569         if (c) {
 570                 c = strchr(c+1, '/');
 571                 if (c) {
 572                         c = strchr(c+1, '/');
 573                         if (c)
 574                                 *c = 0;
 575                 }
 576         }
 577
 578         rl = journal_rate_limit_test(s->rate_limit, path, priority, available_space(s));
 579
 580         if (rl == 0) {
 581                 free(path);
 582                 return;
 583         }
 584
 585         if (rl > 1) {
 586                 int j = 0;
 587                 char suppress_message[LINE_MAX];
 588                 struct iovec suppress_iovec[15];
 589
 590                 /* Write a suppression message if we suppressed something */
 591
 592                 snprintf(suppress_message, sizeof(suppress_message), "MESSAGE=Suppressed %u messages from %s", rl - 1, path);
 593                 char_array_0(suppress_message);
 594
 595                 IOVEC_SET_STRING(suppress_iovec[j++], "PRIORITY=5");
 596                 IOVEC_SET_STRING(suppress_iovec[j++], suppress_message);
 597
 598                 dispatch_message_real(s, suppress_iovec, j, ELEMENTSOF(suppress_iovec), NULL, NULL);
 599         }
 600
 601         free(path);
 602
 603 finish:
 604         dispatch_message_real(s, iovec, n, m, ucred, tv);
 605 }
 606
 607 static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) {
 608         char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL;
 609         struct iovec iovec[16];
 610         unsigned n = 0;
 611         int priority = LOG_USER | LOG_INFO;
 612
 613         assert(s);
 614         assert(buf);
 615
 616         parse_syslog_priority((char**) &buf, &priority);
 617         skip_syslog_date((char**) &buf);
 618
 619         if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0)
 620                 IOVEC_SET_STRING(iovec[n++], syslog_priority);
 621
 622         if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0)
 623                 IOVEC_SET_STRING(iovec[n++], syslog_facility);
 624
 625         message = strappend("MESSAGE=", buf);
 626         if (message)
 627                 IOVEC_SET_STRING(iovec[n++], message);
 628
 629         dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, priority & LOG_PRIMASK);
 630
 631         free(message);
 632         free(syslog_facility);
 633         free(syslog_priority);
 634 }
 635
 636 static bool valid_user_field(const char *p, size_t l) {
 637         const char *a;
 638
 639         /* We kinda enforce POSIX syntax recommendations for
 640            environment variables here, but make a couple of additional
 641            requirements.
 642
 643            http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */
 644
 645         /* No empty field names */
 646         if (l <= 0)
 647                 return false;
 648
 649         /* Don't allow names longer than 64 chars */
 650         if (l > 64)
 651                 return false;
 652
 653         /* Variables starting with an underscore are protected */
 654         if (p[0] == '_')
 655                 return false;
 656
 657         /* Don't allow digits as first character */
 658         if (p[0] >= '0' && p[0] <= '9')
 659                 return false;
 660
 661         /* Only allow A-Z0-9 and '_' */
 662         for (a = p; a < p + l; a++)
 663                 if (!((*a >= 'A' && *a <= 'Z') ||
 664                       (*a >= '0' && *a <= '9') ||
 665                       *a == '_'))
 666                         return false;
 667
 668         return true;
 669 }
 670
 671 static void process_native_message(Server *s, const void *buffer, size_t buffer_size, struct ucred *ucred, struct timeval *tv) {
 672         struct iovec *iovec = NULL;
 673         unsigned n = 0, m = 0, j;
 674         const char *p;
 675         size_t remaining;
 676         int priority = LOG_INFO;
 677
 678         assert(s);
 679         assert(buffer || n == 0);
 680
 681         p = buffer;
 682         remaining = buffer_size;
 683
 684         while (remaining > 0) {
 685                 const char *e, *q;
 686
 687                 e = memchr(p, '\n', remaining);
 688
 689                 if (!e) {
 690                         /* Trailing noise, let's ignore it, and flush what we collected */
 691                         log_debug("Received message with trailing noise, ignoring.");
 692                         break;
 693                 }
 694
 695                 if (e == p) {
 696                         /* Entry separator */
 697                         dispatch_message(s, iovec, n, m, ucred, tv, priority);
 698                         n = 0;
 699                         priority = LOG_INFO;
 700
 701                         p++;
 702                         remaining--;
 703                         continue;
 704                 }
 705
 706                 if (*p == '.' || *p == '#') {
 707                         /* Ignore control commands for now, and
 708                          * comments too. */
 709                         remaining -= (e - p) + 1;
 710                         p = e + 1;
 711                         continue;
 712                 }
 713
 714                 /* A property follows */
 715
 716                 if (n+13 >= m) {
 717                         struct iovec *c;
 718                         unsigned u;
 719
 720                         u = MAX((n+13U) * 2U, 4U);
 721                         c = realloc(iovec, u * sizeof(struct iovec));
 722                         if (!c) {
 723                                 log_error("Out of memory");
 724                                 break;
 725                         }
 726
 727                         iovec = c;
 728                         m = u;
 729                 }
 730
 731                 q = memchr(p, '=', e - p);
 732                 if (q) {
 733                         if (valid_user_field(p, q - p)) {
 734                                 /* If the field name starts with an
 735                                  * underscore, skip the variable,
 736                                  * since that indidates a trusted
 737                                  * field */
 738                                 iovec[n].iov_base = (char*) p;
 739                                 iovec[n].iov_len = e - p;
 740                                 n++;
 741
 742                                 /* We need to determine the priority
 743                                  * of this entry for the rate limiting
 744                                  * logic */
 745                                 if (e - p == 10 &&
 746                                     memcmp(p, "PRIORITY=", 10) == 0 &&
 747                                     p[10] >= '0' &&
 748                                     p[10] <= '9')
 749                                         priority = p[10] - '0';
 750                         }
 751
 752                         remaining -= (e - p) + 1;
 753                         p = e + 1;
 754                         continue;
 755                 } else {
 756                         uint64_t l;
 757                         char *k;
 758
 759                         if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
 760                                 log_debug("Failed to parse message, ignoring.");
 761                                 break;
 762                         }
 763
 764                         memcpy(&l, e + 1, sizeof(uint64_t));
 765                         l = le64toh(l);
 766
 767                         if (remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
 768                             e[1+sizeof(uint64_t)+l] != '\n') {
 769                                 log_debug("Failed to parse message, ignoring.");
 770                                 break;
 771                         }
 772
 773                         k = malloc((e - p) + 1 + l);
 774                         if (!k) {
 775                                 log_error("Out of memory");
 776                                 break;
 777                         }
 778
 779                         memcpy(k, p, e - p);
 780                         k[e - p] = '=';
 781                         memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);
 782
 783                         if (valid_user_field(p, e - p)) {
 784                                 iovec[n].iov_base = k;
 785                                 iovec[n].iov_len = (e - p) + 1 + l;
 786                                 n++;
 787                         } else
 788                                 free(k);
 789
 790                         remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
 791                         p = e + 1 + sizeof(uint64_t) + l + 1;
 792                 }
 793         }
 794
 795         dispatch_message(s, iovec, n, m, ucred, tv, priority);
 796
 797         for (j = 0; j < n; j++)
 798                 if (iovec[j].iov_base < buffer ||
 799                     (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size)
 800                         free(iovec[j].iov_base);
 801 }
 802
 803 static int stdout_stream_log(StdoutStream *s, const char *p, size_t l) {
 804         struct iovec iovec[15];
 805         char *message = NULL, *syslog_priority = NULL;
 806         unsigned n = 0;
 807         size_t tag_len;
 808         int priority;
 809
 810         assert(s);
 811         assert(p);
 812
 813         priority = s->priority;
 814
 815         if (s->priority_prefix &&
 816             l > 3 &&
 817             p[0] == '<' &&
 818             p[1] >= '0' && p[1] <= '7' &&
 819             p[2] == '>') {
 820
 821                 priority = p[1] - '0';
 822                 p += 3;
 823                 l -= 3;
 824         }
 825
 826         if (l <= 0)
 827                 return 0;
 828
 829         if (asprintf(&syslog_priority, "PRIORITY=%i", priority) >= 0)
 830                 IOVEC_SET_STRING(iovec[n++], syslog_priority);
 831
 832         tag_len = s->tag ? strlen(s->tag) + 2: 0;
 833         message = malloc(8 + tag_len + l);
 834         if (message) {
 835                 memcpy(message, "MESSAGE=", 8);
 836
 837                 if (s->tag) {
 838                         memcpy(message+8, s->tag, tag_len-2);
 839                         memcpy(message+8+tag_len-2, ": ", 2);
 840                 }
 841
 842                 memcpy(message+8+tag_len, p, l);
 843                 iovec[n].iov_base = message;
 844                 iovec[n].iov_len = 8+tag_len+l;
 845                 n++;
 846         }
 847
 848         dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, priority);
 849
 850         if (s->tee_console) {
 851                 int console;
 852
 853                 console = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
 854                 if (console >= 0) {
 855                         n = 0;
 856                         if (s->tag) {
 857                                 IOVEC_SET_STRING(iovec[n++], s->tag);
 858                                 IOVEC_SET_STRING(iovec[n++], ": ");
 859                         }
 860
 861                         iovec[n].iov_base = (void*) p;
 862                         iovec[n].iov_len = l;
 863                         n++;
 864
 865                         IOVEC_SET_STRING(iovec[n++], (char*) "\n");
 866
 867                         writev(console, iovec, n);
 868                 }
 869         }
 870
 871         free(message);
 872         free(syslog_priority);
 873
 874         return 0;
 875 }
 876
 877 static int stdout_stream_line(StdoutStream *s, const char *p, size_t l) {
 878         assert(s);
 879         assert(p);
 880
 881         while (l > 0 && strchr(WHITESPACE, *p)) {
 882                 l--;
 883                 p++;
 884         }
 885
 886         while (l > 0 && strchr(WHITESPACE, *(p+l-1)))
 887                 l--;
 888
 889         switch (s->state) {
 890
 891         case STDOUT_STREAM_TAG:
 892
 893                 if (l > 0) {
 894                         s->tag = strndup(p, l);
 895                         if (!s->tag) {
 896                                 log_error("Out of memory");
 897                                 return -EINVAL;
 898                         }
 899                 }
 900
 901                 s->state = STDOUT_STREAM_PRIORITY;
 902                 return 0;
 903
 904         case STDOUT_STREAM_PRIORITY:
 905                 if (l != 1 || *p < '0' || *p > '7') {
 906                         log_warning("Failed to parse log priority line.");
 907                         return -EINVAL;
 908                 }
 909
 910                 s->priority = *p - '0';
 911                 s->state = STDOUT_STREAM_PRIORITY_PREFIX;
 912                 return 0;
 913
 914         case STDOUT_STREAM_PRIORITY_PREFIX:
 915                 if (l != 1 || *p < '0' || *p > '1') {
 916                         log_warning("Failed to parse priority prefix line.");
 917                         return -EINVAL;
 918                 }
 919
 920                 s->priority_prefix = *p - '0';
 921                 s->state = STDOUT_STREAM_TEE_CONSOLE;
 922                 return 0;
 923
 924         case STDOUT_STREAM_TEE_CONSOLE:
 925                 if (l != 1 || *p < '0' || *p > '1') {
 926                         log_warning("Failed to parse tee to console line.");
 927                         return -EINVAL;
 928                 }
 929
 930                 s->tee_console = *p - '0';
 931                 s->state = STDOUT_STREAM_RUNNING;
 932                 return 0;
 933
 934         case STDOUT_STREAM_RUNNING:
 935                 return stdout_stream_log(s, p, l);
 936         }
 937
 938         assert_not_reached("Unknown stream state");
 939 }
 940
 941 static int stdout_stream_scan(StdoutStream *s, bool force_flush) {
 942         char *p;
 943         size_t remaining;
 944         int r;
 945
 946         assert(s);
 947
 948         p = s->buffer;
 949         remaining = s->length;
 950         for (;;) {
 951                 char *end;
 952                 size_t skip;
 953
 954                 end = memchr(p, '\n', remaining);
 955                 if (!end) {
 956                         if (remaining >= LINE_MAX) {
 957                                 end = p + LINE_MAX;
 958                                 skip = LINE_MAX;
 959                         } else
 960                                 break;
 961                 } else
 962                         skip = end - p + 1;
 963
 964                 r = stdout_stream_line(s, p, end - p);
 965                 if (r < 0)
 966                         return r;
 967
 968                 remaining -= skip;
 969                 p += skip;
 970         }
 971
 972         if (force_flush && remaining > 0) {
 973                 r = stdout_stream_line(s, p, remaining);
 974                 if (r < 0)
 975                         return r;
 976
 977                 p += remaining;
 978                 remaining = 0;
 979         }
 980
 981         if (p > s->buffer) {
 982                 memmove(s->buffer, p, remaining);
 983                 s->length = remaining;
 984         }
 985
 986         return 0;
 987 }
 988
 989 static int stdout_stream_process(StdoutStream *s) {
 990         ssize_t l;
 991         int r;
 992
 993         assert(s);
 994
 995         l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length);
 996         if (l < 0) {
 997
 998                 if (errno == EAGAIN)
 999                         return 0;
1000
1001                 log_warning("Failed to read from stream: %m");
1002                 return -errno;
1003         }
1004
1005         if (l == 0) {
1006                 r = stdout_stream_scan(s, true);
1007                 if (r < 0)
1008                         return r;
1009
1010                 return 0;
1011         }
1012
1013         s->length += l;
1014         r = stdout_stream_scan(s, false);
1015         if (r < 0)
1016                 return r;
1017
1018         return 1;
1019
1020 }
1021
1022 static void stdout_stream_free(StdoutStream *s) {
1023         assert(s);
1024
1025         if (s->server) {
1026                 assert(s->server->n_stdout_streams > 0);
1027                 s->server->n_stdout_streams --;
1028                 LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s);
1029         }
1030
1031         if (s->fd >= 0) {
1032                 if (s->server)
1033                         epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL);
1034
1035                 close_nointr_nofail(s->fd);
1036         }
1037
1038         free(s->tag);
1039         free(s);
1040 }
1041
1042 static int stdout_stream_new(Server *s) {
1043         StdoutStream *stream;
1044         int fd, r;
1045         socklen_t len;
1046         struct epoll_event ev;
1047
1048         assert(s);
1049
1050         fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC);
1051         if (fd < 0) {
1052                 if (errno == EAGAIN)
1053                         return 0;
1054
1055                 log_error("Failed to accept stdout connection: %m");
1056                 return -errno;
1057         }
1058
1059         if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) {
1060                 log_warning("Too many stdout streams, refusing connection.");
1061                 close_nointr_nofail(fd);
1062                 return 0;
1063         }
1064
1065         stream = new0(StdoutStream, 1);
1066         if (!stream) {
1067                 log_error("Out of memory.");
1068                 close_nointr_nofail(fd);
1069                 return -ENOMEM;
1070         }
1071
1072         stream->fd = fd;
1073
1074         len = sizeof(stream->ucred);
1075         if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) {
1076                 log_error("Failed to determine peer credentials: %m");
1077                 r = -errno;
1078                 goto fail;
1079         }
1080
1081         if (shutdown(fd, SHUT_WR) < 0) {
1082                 log_error("Failed to shutdown writing side of socket: %m");
1083                 r = -errno;
1084                 goto fail;
1085         }
1086
1087         zero(ev);
1088         ev.data.ptr = stream;
1089         ev.events = EPOLLIN;
1090         if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) {
1091                 log_error("Failed to add stream to event loop: %m");
1092                 r = -errno;
1093                 goto fail;
1094         }
1095
1096         stream->server = s;
1097         LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream);
1098         s->n_stdout_streams ++;
1099
1100         return 0;
1101
1102 fail:
1103         stdout_stream_free(stream);
1104         return r;
1105 }
1106
1107 static int system_journal_open(Server *s) {
1108         int r;
1109         char *fn;
1110         sd_id128_t machine;
1111         char ids[33];
1112
1113         r = sd_id128_get_machine(&machine);
1114         if (r < 0)
1115                 return r;
1116
1117         sd_id128_to_string(machine, ids);
1118
1119         if (!s->system_journal) {
1120
1121                 /* First try to create the machine path, but not the prefix */
1122                 fn = strappend("/var/log/journal/", ids);
1123                 if (!fn)
1124                         return -ENOMEM;
1125                 (void) mkdir(fn, 0755);
1126                 free(fn);
1127
1128                 /* The create the system journal file */
1129                 fn = join("/var/log/journal/", ids, "/system.journal", NULL);
1130                 if (!fn)
1131                         return -ENOMEM;
1132
1133                 r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
1134                 free(fn);
1135
1136                 if (r >= 0) {
1137                         s->system_journal->metrics = s->metrics;
1138                         s->system_journal->compress = s->compress;
1139
1140                         fix_perms(s->system_journal, 0);
1141                 } else if (r < 0) {
1142
1143                         if (r == -ENOENT)
1144                                 r = 0;
1145                         else {
1146                                 log_error("Failed to open system journal: %s", strerror(-r));
1147                                 return r;
1148                         }
1149                 }
1150         }
1151
1152         if (!s->runtime_journal) {
1153
1154                 fn = join("/run/log/journal/", ids, "/system.journal", NULL);
1155                 if (!fn)
1156                         return -ENOMEM;
1157
1158                 if (s->system_journal) {
1159
1160                         /* Try to open the runtime journal, but only
1161                          * if it already exists, so that we can flush
1162                          * it into the system journal */
1163
1164                         r = journal_file_open(fn, O_RDWR, 0640, NULL, &s->runtime_journal);
1165                         free(fn);
1166
1167                         if (r < 0) {
1168
1169                                 if (r == -ENOENT)
1170                                         r = 0;
1171                                 else {
1172                                         log_error("Failed to open runtime journal: %s", strerror(-r));
1173                                         return r;
1174                                 }
1175                         }
1176
1177                 } else {
1178
1179                         /* OK, we really need the runtime journal, so create
1180                          * it if necessary. */
1181
1182                         (void) mkdir_parents(fn, 0755);
1183                         r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
1184                         free(fn);
1185
1186                         if (r < 0) {
1187                                 log_error("Failed to open runtime journal: %s", strerror(-r));
1188                                 return r;
1189                         }
1190                 }
1191
1192                 if (s->runtime_journal) {
1193                         s->runtime_journal->metrics = s->metrics;
1194                         s->runtime_journal->compress = s->compress;
1195
1196                         fix_perms(s->runtime_journal, 0);
1197                 }
1198         }
1199
1200         return r;
1201 }
1202
1203 static int server_flush_to_var(Server *s) {
1204         char path[] = "/run/log/journal/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
1205         Object *o = NULL;
1206         int r;
1207         sd_id128_t machine;
1208         sd_journal *j;
1209         usec_t ts;
1210
1211         assert(s);
1212
1213         if (!s->runtime_journal)
1214                 return 0;
1215
1216         ts = now(CLOCK_MONOTONIC);
1217         if (s->var_available_timestamp + RECHECK_VAR_AVAILABLE_USEC > ts)
1218                 return 0;
1219
1220         s->var_available_timestamp = ts;
1221
1222         system_journal_open(s);
1223
1224         if (!s->system_journal)
1225                 return 0;
1226
1227         r = sd_id128_get_machine(&machine);
1228         if (r < 0) {
1229                 log_error("Failed to get machine id: %s", strerror(-r));
1230                 return r;
1231         }
1232
1233         r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY);
1234         if (r < 0) {
1235                 log_error("Failed to read runtime journal: %s", strerror(-r));
1236                 return r;
1237         }
1238
1239         SD_JOURNAL_FOREACH(j) {
1240                 JournalFile *f;
1241
1242                 f = j->current_file;
1243                 assert(f && f->current_offset > 0);
1244
1245                 r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
1246                 if (r < 0) {
1247                         log_error("Can't read entry: %s", strerror(-r));
1248                         goto finish;
1249                 }
1250
1251                 r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
1252                 if (r == -E2BIG) {
1253                         log_info("Allocation limit reached.");
1254
1255                         journal_file_post_change(s->system_journal);
1256                         server_vacuum(s);
1257
1258                         r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);
1259                 }
1260
1261                 if (r < 0) {
1262                         log_error("Can't write entry: %s", strerror(-r));
1263                         goto finish;
1264                 }
1265         }
1266
1267 finish:
1268         journal_file_post_change(s->system_journal);
1269
1270         journal_file_close(s->runtime_journal);
1271         s->runtime_journal = NULL;
1272
1273         if (r >= 0) {
1274                 sd_id128_to_string(machine, path + 17);
1275                 rm_rf(path, false, true, false);
1276         }
1277
1278         return r;
1279 }
1280
1281 static void forward_syslog(Server *s, const void *buffer, size_t length, struct ucred *ucred, struct timeval *tv) {
1282         struct msghdr msghdr;
1283         struct iovec iovec;
1284         struct cmsghdr *cmsg;
1285         union {
1286                 struct cmsghdr cmsghdr;
1287                 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
1288                             CMSG_SPACE(sizeof(struct timeval))];
1289         } control;
1290         union sockaddr_union sa;
1291
1292         assert(s);
1293
1294         zero(msghdr);
1295
1296         zero(iovec);
1297         iovec.iov_base = (void*) buffer;
1298         iovec.iov_len = length;
1299         msghdr.msg_iov = &iovec;
1300         msghdr.msg_iovlen = 1;
1301
1302         zero(sa);
1303         sa.un.sun_family = AF_UNIX;
1304         strncpy(sa.un.sun_path, "/run/systemd/syslog", sizeof(sa.un.sun_path));
1305         msghdr.msg_name = &sa;
1306         msghdr.msg_namelen = offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path);
1307
1308         zero(control);
1309         msghdr.msg_control = &control;
1310         msghdr.msg_controllen = sizeof(control);
1311
1312         cmsg = CMSG_FIRSTHDR(&msghdr);
1313         cmsg->cmsg_level = SOL_SOCKET;
1314         cmsg->cmsg_type = SCM_CREDENTIALS;
1315         cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
1316         memcpy(CMSG_DATA(cmsg), ucred, sizeof(struct ucred));
1317         msghdr.msg_controllen = cmsg->cmsg_len;
1318
1319         /* Forward the syslog message we received via /dev/log to
1320          * /run/systemd/syslog. Unfortunately we currently can't set
1321          * the SO_TIMESTAMP auxiliary data, and hence we don't. */
1322
1323         if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
1324                 return;
1325
1326         if (errno == ESRCH) {
1327                 struct ucred u;
1328
1329                 /* Hmm, presumably the sender process vanished
1330                  * by now, so let's fix it as good as we
1331                  * can, and retry */
1332
1333                 u = *ucred;
1334                 u.pid = getpid();
1335                 memcpy(CMSG_DATA(cmsg), &u, sizeof(struct ucred));
1336
1337                 if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
1338                         return;
1339         }
1340
1341         log_debug("Failed to forward syslog message: %m");
1342 }
1343
1344 static int process_event(Server *s, struct epoll_event *ev) {
1345         assert(s);
1346
1347         if (ev->data.fd == s->signal_fd) {
1348                 struct signalfd_siginfo sfsi;
1349                 ssize_t n;
1350
1351                 if (ev->events != EPOLLIN) {
1352                         log_info("Got invalid event from epoll.");
1353                         return -EIO;
1354                 }
1355
1356                 n = read(s->signal_fd, &sfsi, sizeof(sfsi));
1357                 if (n != sizeof(sfsi)) {
1358
1359                         if (n >= 0)
1360                                 return -EIO;
1361
1362                         if (errno == EINTR || errno == EAGAIN)
1363                                 return 0;
1364
1365                         return -errno;
1366                 }
1367
1368                 if (sfsi.ssi_signo == SIGUSR1) {
1369                         server_flush_to_var(s);
1370                         return 0;
1371                 }
1372
1373                 log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo));
1374                 return 0;
1375
1376         } else if (ev->data.fd == s->native_fd ||
1377                    ev->data.fd == s->syslog_fd) {
1378
1379                 if (ev->events != EPOLLIN) {
1380                         log_info("Got invalid event from epoll.");
1381                         return -EIO;
1382                 }
1383
1384                 for (;;) {
1385                         struct msghdr msghdr;
1386                         struct iovec iovec;
1387                         struct ucred *ucred = NULL;
1388                         struct timeval *tv = NULL;
1389                         struct cmsghdr *cmsg;
1390                         union {
1391                                 struct cmsghdr cmsghdr;
1392                                 uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +
1393                                             CMSG_SPACE(sizeof(struct timeval))];
1394                         } control;
1395                         ssize_t n;
1396                         int v;
1397
1398                         if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
1399                                 log_error("SIOCINQ failed: %m");
1400                                 return -errno;
1401                         }
1402
1403                         if (v <= 0)
1404                                 return 1;
1405
1406                         if (s->buffer_size < (size_t) v) {
1407                                 void *b;
1408                                 size_t l;
1409
1410                                 l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
1411                                 b = realloc(s->buffer, l+1);
1412
1413                                 if (!b) {
1414                                         log_error("Couldn't increase buffer.");
1415                                         return -ENOMEM;
1416                                 }
1417
1418                                 s->buffer_size = l;
1419                                 s->buffer = b;
1420                         }
1421
1422                         zero(iovec);
1423                         iovec.iov_base = s->buffer;
1424                         iovec.iov_len = s->buffer_size;
1425
1426                         zero(control);
1427                         zero(msghdr);
1428                         msghdr.msg_iov = &iovec;
1429                         msghdr.msg_iovlen = 1;
1430                         msghdr.msg_control = &control;
1431                         msghdr.msg_controllen = sizeof(control);
1432
1433                         n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT);
1434                         if (n < 0) {
1435
1436                                 if (errno == EINTR || errno == EAGAIN)
1437                                         return 1;
1438
1439                                 log_error("recvmsg() failed: %m");
1440                                 return -errno;
1441                         }
1442
1443                         for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) {
1444
1445                                 if (cmsg->cmsg_level == SOL_SOCKET &&
1446                                     cmsg->cmsg_type == SCM_CREDENTIALS &&
1447                                     cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)))
1448                                         ucred = (struct ucred*) CMSG_DATA(cmsg);
1449                                 else if (cmsg->cmsg_level == SOL_SOCKET &&
1450                                          cmsg->cmsg_type == SO_TIMESTAMP &&
1451                                          cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval)))
1452                                         tv = (struct timeval*) CMSG_DATA(cmsg);
1453                         }
1454
1455                         if (ev->data.fd == s->syslog_fd) {
1456                                 char *e;
1457
1458                                 e = memchr(s->buffer, '\n', n);
1459                                 if (e)
1460                                         *e = 0;
1461                                 else
1462                                         s->buffer[n] = 0;
1463
1464                                 forward_syslog(s, s->buffer, n, ucred, tv);
1465                                 process_syslog_message(s, strstrip(s->buffer), ucred, tv);
1466                         } else
1467                                 process_native_message(s, s->buffer, n, ucred, tv);
1468                 }
1469
1470                 return 1;
1471
1472         } else if (ev->data.fd == s->stdout_fd) {
1473
1474                 if (ev->events != EPOLLIN) {
1475                         log_info("Got invalid event from epoll.");
1476                         return -EIO;
1477                 }
1478
1479                 stdout_stream_new(s);
1480                 return 1;
1481
1482         } else {
1483                 StdoutStream *stream;
1484
1485                 if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
1486                         log_info("Got invalid event from epoll.");
1487                         return -EIO;
1488                 }
1489
1490                 /* If it is none of the well-known fds, it must be an
1491                  * stdout stream fd. Note that this is a bit ugly here
1492                  * (since we rely that none of the well-known fds
1493                  * could be interpreted as pointer), but nonetheless
1494                  * safe, since the well-known fds would never get an
1495                  * fd > 4096, i.e. beyond the first memory page */
1496
1497                 stream = ev->data.ptr;
1498
1499                 if (stdout_stream_process(stream) <= 0)
1500                         stdout_stream_free(stream);
1501
1502                 return 1;
1503         }
1504
1505         log_error("Unknown event.");
1506         return 0;
1507 }
1508
1509 static int open_syslog_socket(Server *s) {
1510         union sockaddr_union sa;
1511         int one, r;
1512         struct epoll_event ev;
1513         struct timeval tv;
1514
1515         assert(s);
1516
1517         if (s->syslog_fd < 0) {
1518
1519                 s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
1520                 if (s->syslog_fd < 0) {
1521                         log_error("socket() failed: %m");
1522                         return -errno;
1523                 }
1524
1525                 zero(sa);
1526                 sa.un.sun_family = AF_UNIX;
1527                 strncpy(sa.un.sun_path, "/dev/log", sizeof(sa.un.sun_path));
1528
1529                 unlink(sa.un.sun_path);
1530
1531                 r = bind(s->syslog_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1532                 if (r < 0) {
1533                         log_error("bind() failed: %m");
1534                         return -errno;
1535                 }
1536
1537                 chmod(sa.un.sun_path, 0666);
1538         }
1539
1540         one = 1;
1541         r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
1542         if (r < 0) {
1543                 log_error("SO_PASSCRED failed: %m");
1544                 return -errno;
1545         }
1546
1547         one = 1;
1548         r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
1549         if (r < 0) {
1550                 log_error("SO_TIMESTAMP failed: %m");
1551                 return -errno;
1552         }
1553
1554         /* Since we use the same socket for forwarding this to some
1555          * other syslog implementation, make sure we don't hang
1556          * forever */
1557         timeval_store(&tv, SYSLOG_TIMEOUT_USEC);
1558         if (setsockopt(s->syslog_fd, SOL_SOCKET, SO_SNDTIMEO, &tv, sizeof(tv)) < 0) {
1559                 log_error("SO_SNDTIMEO failed: %m");
1560                 return -errno;
1561         }
1562
1563         zero(ev);
1564         ev.events = EPOLLIN;
1565         ev.data.fd = s->syslog_fd;
1566         if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->syslog_fd, &ev) < 0) {
1567                 log_error("Failed to add syslog server fd to epoll object: %m");
1568                 return -errno;
1569         }
1570
1571         return 0;
1572 }
1573
1574 static int open_native_socket(Server*s) {
1575         union sockaddr_union sa;
1576         int one, r;
1577         struct epoll_event ev;
1578
1579         assert(s);
1580
1581         if (s->native_fd < 0) {
1582
1583                 s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0);
1584                 if (s->native_fd < 0) {
1585                         log_error("socket() failed: %m");
1586                         return -errno;
1587                 }
1588
1589                 zero(sa);
1590                 sa.un.sun_family = AF_UNIX;
1591                 strncpy(sa.un.sun_path, "/run/systemd/journal", sizeof(sa.un.sun_path));
1592
1593                 unlink(sa.un.sun_path);
1594
1595                 r = bind(s->native_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1596                 if (r < 0) {
1597                         log_error("bind() failed: %m");
1598                         return -errno;
1599                 }
1600
1601                 chmod(sa.un.sun_path, 0666);
1602         }
1603
1604         one = 1;
1605         r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
1606         if (r < 0) {
1607                 log_error("SO_PASSCRED failed: %m");
1608                 return -errno;
1609         }
1610
1611         one = 1;
1612         r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
1613         if (r < 0) {
1614                 log_error("SO_TIMESTAMP failed: %m");
1615                 return -errno;
1616         }
1617
1618         zero(ev);
1619         ev.events = EPOLLIN;
1620         ev.data.fd = s->native_fd;
1621         if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->native_fd, &ev) < 0) {
1622                 log_error("Failed to add native server fd to epoll object: %m");
1623                 return -errno;
1624         }
1625
1626         return 0;
1627 }
1628
1629 static int open_stdout_socket(Server *s) {
1630         union sockaddr_union sa;
1631         int r;
1632         struct epoll_event ev;
1633
1634         assert(s);
1635
1636         if (s->stdout_fd < 0) {
1637
1638                 s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0);
1639                 if (s->stdout_fd < 0) {
1640                         log_error("socket() failed: %m");
1641                         return -errno;
1642                 }
1643
1644                 zero(sa);
1645                 sa.un.sun_family = AF_UNIX;
1646                 strncpy(sa.un.sun_path, "/run/systemd/stdout", sizeof(sa.un.sun_path));
1647
1648                 unlink(sa.un.sun_path);
1649
1650                 r = bind(s->stdout_fd, &sa.sa, offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path));
1651                 if (r < 0) {
1652                         log_error("bind() failed: %m");
1653                         return -errno;
1654                 }
1655
1656                 chmod(sa.un.sun_path, 0666);
1657
1658                 if (listen(s->stdout_fd, SOMAXCONN) < 0) {
1659                         log_error("liste() failed: %m");
1660                         return -errno;
1661                 }
1662         }
1663
1664         zero(ev);
1665         ev.events = EPOLLIN;
1666         ev.data.fd = s->stdout_fd;
1667         if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->stdout_fd, &ev) < 0) {
1668                 log_error("Failed to add stdout server fd to epoll object: %m");
1669                 return -errno;
1670         }
1671
1672         return 0;
1673 }
1674
1675 static int open_signalfd(Server *s) {
1676         sigset_t mask;
1677         struct epoll_event ev;
1678
1679         assert(s);
1680
1681         assert_se(sigemptyset(&mask) == 0);
1682         sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, -1);
1683         assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
1684
1685         s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
1686         if (s->signal_fd < 0) {
1687                 log_error("signalfd(): %m");
1688                 return -errno;
1689         }
1690
1691         zero(ev);
1692         ev.events = EPOLLIN;
1693         ev.data.fd = s->signal_fd;
1694
1695         if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) {
1696                 log_error("epoll_ctl(): %m");
1697                 return -errno;
1698         }
1699
1700         return 0;
1701 }
1702
1703 static int server_init(Server *s) {
1704         int n, r, fd;
1705
1706         assert(s);
1707
1708         zero(*s);
1709         s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = -1;
1710         s->metrics.max_size = DEFAULT_MAX_SIZE;
1711         s->metrics.min_size = DEFAULT_MIN_SIZE;
1712         s->metrics.keep_free = DEFAULT_KEEP_FREE;
1713         s->max_use = DEFAULT_MAX_USE;
1714         s->compress = true;
1715
1716         s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
1717         if (!s->user_journals) {
1718                 log_error("Out of memory.");
1719                 return -ENOMEM;
1720         }
1721
1722         s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
1723         if (s->epoll_fd < 0) {
1724                 log_error("Failed to create epoll object: %m");
1725                 return -errno;
1726         }
1727
1728         n = sd_listen_fds(true);
1729         if (n < 0) {
1730                 log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
1731                 return n;
1732         }
1733
1734         for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) {
1735
1736                 if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/native", 0) > 0) {
1737
1738                         if (s->native_fd >= 0) {
1739                                 log_error("Too many native sockets passed.");
1740                                 return -EINVAL;
1741                         }
1742
1743                         s->native_fd = fd;
1744
1745                 } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/stdout", 0) > 0) {
1746
1747                         if (s->stdout_fd >= 0) {
1748                                 log_error("Too many stdout sockets passed.");
1749                                 return -EINVAL;
1750                         }
1751
1752                         s->stdout_fd = fd;
1753
1754                 } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
1755
1756                         if (s->syslog_fd >= 0) {
1757                                 log_error("Too many /dev/log sockets passed.");
1758                                 return -EINVAL;
1759                         }
1760
1761                         s->syslog_fd = fd;
1762
1763                 } else {
1764                         log_error("Unknown socket passed.");
1765                         return -EINVAL;
1766                 }
1767         }
1768
1769         r = open_syslog_socket(s);
1770         if (r < 0)
1771                 return r;
1772
1773         r = open_native_socket(s);
1774         if (r < 0)
1775                 return r;
1776
1777         r = open_stdout_socket(s);
1778         if (r < 0)
1779                 return r;
1780
1781         r = system_journal_open(s);
1782         if (r < 0)
1783                 return r;
1784
1785         r = open_signalfd(s);
1786         if (r < 0)
1787                 return r;
1788
1789         s->rate_limit = journal_rate_limit_new(DEFAULT_RATE_LIMIT_INTERVAL, DEFAULT_RATE_LIMIT_BURST);
1790         if (!s->rate_limit)
1791                 return -ENOMEM;
1792
1793         return 0;
1794 }
1795
1796 static void server_done(Server *s) {
1797         JournalFile *f;
1798         assert(s);
1799
1800         while (s->stdout_streams)
1801                 stdout_stream_free(s->stdout_streams);
1802
1803         if (s->system_journal)
1804                 journal_file_close(s->system_journal);
1805
1806         if (s->runtime_journal)
1807                 journal_file_close(s->runtime_journal);
1808
1809         while ((f = hashmap_steal_first(s->user_journals)))
1810                 journal_file_close(f);
1811
1812         hashmap_free(s->user_journals);
1813
1814         if (s->epoll_fd >= 0)
1815                 close_nointr_nofail(s->epoll_fd);
1816
1817         if (s->signal_fd >= 0)
1818                 close_nointr_nofail(s->signal_fd);
1819
1820         if (s->syslog_fd >= 0)
1821                 close_nointr_nofail(s->syslog_fd);
1822
1823         if (s->native_fd >= 0)
1824                 close_nointr_nofail(s->native_fd);
1825
1826         if (s->stdout_fd >= 0)
1827                 close_nointr_nofail(s->stdout_fd);
1828
1829         if (s->rate_limit)
1830                 journal_rate_limit_free(s->rate_limit);
1831 }
1832
1833 int main(int argc, char *argv[]) {
1834         Server server;
1835         int r;
1836
1837         /* if (getppid() != 1) { */
1838         /*         log_error("This program should be invoked by init only."); */
1839         /*         return EXIT_FAILURE; */
1840         /* } */
1841
1842         if (argc > 1) {
1843                 log_error("This program does not take arguments.");
1844                 return EXIT_FAILURE;
1845         }
1846
1847         log_set_target(LOG_TARGET_CONSOLE);
1848         log_set_max_level(LOG_DEBUG);
1849         log_parse_environment();
1850         log_open();
1851
1852         umask(0022);
1853
1854         r = server_init(&server);
1855         if (r < 0)
1856                 goto finish;
1857
1858         log_debug("systemd-journald running as pid %lu", (unsigned long) getpid());
1859
1860         sd_notify(false,
1861                   "READY=1\n"
1862                   "STATUS=Processing requests...");
1863
1864         server_vacuum(&server);
1865         server_flush_to_var(&server);
1866
1867         for (;;) {
1868                 struct epoll_event event;
1869
1870                 r = epoll_wait(server.epoll_fd, &event, 1, -1);
1871                 if (r < 0) {
1872
1873                         if (errno == EINTR)
1874                                 continue;
1875
1876                         log_error("epoll_wait() failed: %m");
1877                         r = -errno;
1878                         goto finish;
1879                 } else if (r == 0)
1880                         break;
1881
1882                 r = process_event(&server, &event);
1883                 if (r < 0)
1884                         goto finish;
1885                 else if (r == 0)
1886                         break;
1887         }
1888
1889         log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid());
1890
1891 finish:
1892         sd_notify(false,
1893                   "STATUS=Shutting down...");
1894
1895         server_done(&server);
1896
1897         return r < 0 ? EXIT_FAILURE : EXIT_SUCCESS;
1898 }