4141N/A

4141N/A

4141N/A#include <sys/signalfd.h>

4141N/A#include <sys/ioctl.h>

4141N/A#include <linux/sockios.h>

4141N/A#include <sys/statvfs.h>

4141N/A#include <sys/mman.h>

4141N/A#include <sys/timerfd.h>

4141N/A

4141N/A#include <libudev.h>

4141N/A

4141N/A#include "sd-journal.h"

4141N/A#include "sd-messages.h"

4141N/A#include "sd-daemon.h"

4141N/A#include "fileio.h"

4141N/A#include "mkdir.h"

4141N/A#include "hashmap.h"

4141N/A#include "journal-file.h"

4141N/A#include "socket-util.h"

4141N/A#include "cgroup-util.h"

4141N/A#include "list.h"

4141N/A#include "missing.h"

4141N/A#include "conf-parser.h"

4141N/A#include "selinux-util.h"

4141N/A#include "journal-internal.h"

4141N/A#include "journal-vacuum.h"

4141N/A#include "journal-authenticate.h"

4141N/A#include "journald-rate-limit.h"

4141N/A#include "journald-kmsg.h"

4141N/A#include "journald-syslog.h"

4141N/A#include "journald-stream.h"

4141N/A#include "journald-console.h"

4141N/A#include "journald-native.h"

4141N/A#include "journald-server.h"

4141N/A

4141N/A#ifdef HAVE_ACL

4141N/A#include <sys/acl.h>

4141N/A#include <acl/libacl.h>

4141N/A#include "acl-util.h"

4141N/A#endif

4141N/A

4141N/A#ifdef HAVE_SELINUX

4141N/A#include <selinux/selinux.h>

4141N/A#endif

4141N/A

4141N/A#define USER_JOURNALS_MAX 1024

4141N/A

4141N/A#define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE)

4141N/A#define DEFAULT_RATE_LIMIT_INTERVAL (30*USEC_PER_SEC)

4141N/A#define DEFAULT_RATE_LIMIT_BURST 1000

4141N/A

4141N/A#define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)

4141N/A

4141N/Astatic const char* const storage_table[] = {

4141N/A [STORAGE_AUTO] = "auto",

4141N/A [STORAGE_VOLATILE] = "volatile",

4141N/A [STORAGE_PERSISTENT] = "persistent",

4141N/A [STORAGE_NONE] = "none"

4141N/A};

4141N/A

4141N/ADEFINE_STRING_TABLE_LOOKUP(storage, Storage);

4141N/ADEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting");

4141N/A

4141N/Astatic const char* const split_mode_table[] = {

4141N/A [SPLIT_NONE] = "none",

4141N/A [SPLIT_UID] = "uid",

4141N/A [SPLIT_LOGIN] = "login"

4141N/A};

4141N/A

4141N/ADEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode);

4141N/ADEFINE_CONFIG_PARSE_ENUM(config_parse_split_mode, split_mode, SplitMode, "Failed to parse split mode setting");

4141N/A

4141N/Astatic uint64_t available_space(Server *s, bool verbose) {

4141N/A char ids[33];

4141N/A _cleanup_free_ char *p = NULL;

4141N/A sd_id128_t machine;

4141N/A struct statvfs ss;

4141N/A uint64_t sum = 0, ss_avail = 0, avail = 0;

4141N/A int r;

4141N/A _cleanup_closedir_ DIR *d = NULL;

4141N/A usec_t ts;

4141N/A const char *f;

4141N/A JournalMetrics *m;

4141N/A

4141N/A ts = now(CLOCK_MONOTONIC);

4141N/A

4141N/A if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts

4141N/A && !verbose)

4141N/A return s->cached_available_space;

4141N/A

4141N/A r = sd_id128_get_machine(&machine);

4141N/A if (r < 0)

4141N/A return 0;

4141N/A

4141N/A if (s->system_journal) {

4141N/A f = "/var/log/journal/";

4141N/A m = &s->system_metrics;

4141N/A } else {

4141N/A f = "/run/log/journal/";

4141N/A m = &s->runtime_metrics;

4141N/A }

4141N/A

4141N/A assert(m);

4141N/A

4141N/A p = strappend(f, sd_id128_to_string(machine, ids));

4141N/A if (!p)

4141N/A return 0;

4141N/A

4141N/A d = opendir(p);

4141N/A if (!d)

4141N/A return 0;

4141N/A

4141N/A if (fstatvfs(dirfd(d), &ss) < 0)

4141N/A return 0;

4141N/A

4141N/A for (;;) {

4141N/A struct stat st;

4141N/A struct dirent *de;

4141N/A union dirent_storage buf;

4141N/A

4141N/A r = readdir_r(d, &buf.de, &de);

4141N/A if (r != 0)

4141N/A break;

4141N/A

4141N/A if (!de)

4141N/A break;

4141N/A

4141N/A if (!endswith(de->d_name, ".journal") &&

4141N/A !endswith(de->d_name, ".journal~"))

4141N/A continue;

4141N/A

4141N/A if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)

4141N/A continue;

4141N/A

4141N/A if (!S_ISREG(st.st_mode))

4141N/A continue;

4141N/A

4141N/A sum += (uint64_t) st.st_blocks * 512UL;

4141N/A }

4141N/A

4141N/A ss_avail = ss.f_bsize * ss.f_bavail;

4141N/A avail = ss_avail > m->keep_free ? ss_avail - m->keep_free : 0;

4141N/A

4141N/A s->cached_available_space = MIN(m->max_use, avail) > sum ? MIN(m->max_use, avail) - sum : 0;

4141N/A s->cached_available_space_timestamp = ts;

4141N/A

4141N/A if (verbose) {

4141N/A char fb1[FORMAT_BYTES_MAX], fb2[FORMAT_BYTES_MAX], fb3[FORMAT_BYTES_MAX],

4141N/A fb4[FORMAT_BYTES_MAX], fb5[FORMAT_BYTES_MAX];

4141N/A

4141N/A server_driver_message(s, SD_MESSAGE_JOURNAL_USAGE,

4141N/A "%s journal is using %s (max %s, leaving %s of free %s, current limit %s).",

4141N/A s->system_journal ? "Permanent" : "Runtime",

4141N/A format_bytes(fb1, sizeof(fb1), sum),

4141N/A format_bytes(fb2, sizeof(fb2), m->max_use),

4141N/A format_bytes(fb3, sizeof(fb3), m->keep_free),

4141N/A format_bytes(fb4, sizeof(fb4), ss_avail),

4141N/A format_bytes(fb5, sizeof(fb5), MIN(m->max_use, avail)));

4141N/A }

4141N/A

4141N/A return s->cached_available_space;

4141N/A}

4141N/A

4141N/Avoid server_fix_perms(Server *s, JournalFile *f, uid_t uid) {

4141N/A int r;

4141N/A#ifdef HAVE_ACL

4141N/A acl_t acl;

4141N/A acl_entry_t entry;

4141N/A acl_permset_t permset;

4141N/A#endif

4141N/A

4141N/A assert(f);

4141N/A

4141N/A r = fchmod(f->fd, 0640);

4141N/A if (r < 0)

4141N/A log_warning("Failed to fix access mode on %s, ignoring: %s", f->path, strerror(-r));

4141N/A

4141N/A#ifdef HAVE_ACL

4141N/A if (uid <= 0)

4141N/A return;

4141N/A

4141N/A acl = acl_get_fd(f->fd);

4141N/A if (!acl) {

4141N/A log_warning("Failed to read ACL on %s, ignoring: %m", f->path);

4141N/A return;

4141N/A }

4141N/A

4141N/A r = acl_find_uid(acl, uid, &entry);

4141N/A if (r <= 0) {

4141N/A

4141N/A if (acl_create_entry(&acl, &entry) < 0 ||

4141N/A acl_set_tag_type(entry, ACL_USER) < 0 ||

4141N/A acl_set_qualifier(entry, &uid) < 0) {

4141N/A log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);

4141N/A goto finish;

4141N/A }

4141N/A }

4141N/A

4141N/A /* We do not recalculate the mask unconditionally here,

4141N/A if (acl_get_permset(entry, &permset) < 0 ||

4141N/A acl_add_perm(permset, ACL_READ) < 0 ||

4141N/A calc_acl_mask_if_needed(&acl) < 0) {

4141N/A log_warning("Failed to patch ACL on %s, ignoring: %m", f->path);

4141N/A goto finish;

4141N/A }

4141N/A

4141N/A if (acl_set_fd(f->fd, acl) < 0)

4141N/A log_warning("Failed to set ACL on %s, ignoring: %m", f->path);

4141N/A

4141N/Afinish:

4141N/A acl_free(acl);

4141N/A#endif

4141N/A}

4141N/A

4141N/Astatic JournalFile* find_journal(Server *s, uid_t uid) {

4141N/A _cleanup_free_ char *p = NULL;

4141N/A int r;

4141N/A JournalFile *f;

4141N/A sd_id128_t machine;

4141N/A

4141N/A assert(s);

4141N/A

4141N/A /* We split up user logs only on /var, not on /run. If the

4141N/A

4141N/A if (s->runtime_journal)

4141N/A return s->runtime_journal;

4141N/A

4141N/A if (uid <= 0)

4141N/A return s->system_journal;

4141N/A

4141N/A r = sd_id128_get_machine(&machine);

4141N/A if (r < 0)

4141N/A return s->system_journal;

4141N/A

4141N/A f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid));

4141N/A if (f)

4141N/A return f;

4141N/A

4141N/A if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-%lu.journal",

4141N/A SD_ID128_FORMAT_VAL(machine), (unsigned long) uid) < 0)

4141N/A return s->system_journal;

4141N/A

4141N/A while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {

4141N/A /* Too many open? Then let's close one */

4141N/A f = hashmap_steal_first(s->user_journals);

4141N/A assert(f);

4141N/A journal_file_close(f);

4141N/A }

4141N/A

4141N/A r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &f);

4141N/A if (r < 0)

4141N/A return s->system_journal;

4141N/A

4141N/A server_fix_perms(s, f, uid);

4141N/A

4141N/A r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);

4141N/A if (r < 0) {

4141N/A journal_file_close(f);

4141N/A return s->system_journal;

4141N/A }

4141N/A

4141N/A return f;

4141N/A}

4141N/A

4141N/Avoid server_rotate(Server *s) {

4141N/A JournalFile *f;

4141N/A void *k;

4141N/A Iterator i;

4141N/A int r;

4141N/A

4141N/A log_debug("Rotating...");

4141N/A

4141N/A if (s->runtime_journal) {

4141N/A r = journal_file_rotate(&s->runtime_journal, s->compress, false);

4141N/A if (r < 0)

4141N/A if (s->runtime_journal)

4141N/A log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));

4141N/A else

4141N/A log_error("Failed to create new runtime journal: %s", strerror(-r));

4141N/A else

4141N/A server_fix_perms(s, s->runtime_journal, 0);

4141N/A }

4141N/A

4141N/A if (s->system_journal) {

4141N/A r = journal_file_rotate(&s->system_journal, s->compress, s->seal);

4141N/A if (r < 0)

4141N/A if (s->system_journal)

4141N/A log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));

4141N/A else

4141N/A log_error("Failed to create new system journal: %s", strerror(-r));

4141N/A

4141N/A else

4141N/A server_fix_perms(s, s->system_journal, 0);

4141N/A }

4141N/A

4141N/A HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {

4141N/A r = journal_file_rotate(&f, s->compress, s->seal);

4141N/A if (r < 0)

4141N/A if (f)

4141N/A log_error("Failed to rotate %s: %s", f->path, strerror(-r));

4141N/A else {

4141N/A log_error("Failed to create user journal: %s", strerror(-r));

4141N/A hashmap_remove(s->user_journals, k);

4141N/A }

4141N/A else {

4141N/A hashmap_replace(s->user_journals, k, f);

4141N/A server_fix_perms(s, f, PTR_TO_UINT32(k));

4141N/A }

4141N/A }

4141N/A}

4141N/A

4141N/Avoid server_sync(Server *s) {

4141N/A JournalFile *f;

4141N/A void *k;

4141N/A Iterator i;

4141N/A int r;

4141N/A

4141N/A if (s->system_journal) {

4141N/A r = journal_file_set_offline(s->system_journal);

4141N/A if (r < 0)

4141N/A log_error("Failed to sync system journal: %s", strerror(-r));

4141N/A }

4141N/A

4141N/A HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {

4141N/A r = journal_file_set_offline(f);

4141N/A if (r < 0)

4141N/A log_error("Failed to sync user journal: %s", strerror(-r));

4141N/A }

4141N/A

4141N/A if (s->sync_event_source) {

4141N/A r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_OFF);

4141N/A if (r < 0)

4141N/A log_error("Failed to disable sync timer source: %s", strerror(-r));

4141N/A }

4141N/A

4141N/A s->sync_scheduled = false;

4141N/A}

4141N/A

4141N/Avoid server_vacuum(Server *s) {

4141N/A char ids[33];

4141N/A sd_id128_t machine;

4141N/A int r;

4141N/A

4141N/A log_debug("Vacuuming...");

4141N/A

4141N/A s->oldest_file_usec = 0;

4141N/A

4141N/A r = sd_id128_get_machine(&machine);

4141N/A if (r < 0) {

4141N/A log_error("Failed to get machine ID: %s", strerror(-r));

4141N/A return;

4141N/A }

4141N/A

4141N/A sd_id128_to_string(machine, ids);

4141N/A

4141N/A if (s->system_journal) {

4141N/A char *p = strappenda("/var/log/journal/", ids);

4141N/A

4141N/A r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec);

4141N/A if (r < 0 && r != -ENOENT)

4141N/A log_error("Failed to vacuum %s: %s", p, strerror(-r));

4141N/A }

4141N/A

4141N/A if (s->runtime_journal) {

4141N/A char *p = strappenda("/run/log/journal/", ids);

4141N/A

4141N/A r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec);

4141N/A if (r < 0 && r != -ENOENT)

4141N/A log_error("Failed to vacuum %s: %s", p, strerror(-r));

4141N/A }

4141N/A

4141N/A s->cached_available_space_timestamp = 0;

4141N/A}

4141N/A

4141N/Astatic void server_cache_machine_id(Server *s) {

4141N/A sd_id128_t id;

int r;

assert(s);

r = sd_id128_get_machine(&id);

if (r < 0)

return;

sd_id128_to_string(id, stpcpy(s->machine_id_field, "_MACHINE_ID="));

}

static void server_cache_boot_id(Server *s) {

sd_id128_t id;

int r;

assert(s);

r = sd_id128_get_boot(&id);

if (r < 0)

return;

sd_id128_to_string(id, stpcpy(s->boot_id_field, "_BOOT_ID="));

}

static void server_cache_hostname(Server *s) {

_cleanup_free_ char *t = NULL;

char *x;

assert(s);

t = gethostname_malloc();

if (!t)

return;

x = strappend("_HOSTNAME=", t);

if (!x)

return;

free(s->hostname_field);

s->hostname_field = x;

}

bool shall_try_append_again(JournalFile *f, int r) {

/* -E2BIG Hit configured limit

if (r == -E2BIG || r == -EFBIG || r == -EDQUOT || r == -ENOSPC)

log_debug("%s: Allocation limit reached, rotating.", f->path);

else if (r == -EHOSTDOWN)

log_info("%s: Journal file from other machine, rotating.", f->path);

else if (r == -EBUSY)

log_info("%s: Unclean shutdown, rotating.", f->path);

else if (r == -EPROTONOSUPPORT)

log_info("%s: Unsupported feature, rotating.", f->path);

else if (r == -EBADMSG || r == -ENODATA || r == ESHUTDOWN)

log_warning("%s: Journal file corrupted, rotating.", f->path);

else

return false;

return true;

}

static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n, int priority) {

JournalFile *f;

bool vacuumed = false;

int r;

assert(s);

assert(iovec);

assert(n > 0);

f = find_journal(s, uid);

if (!f)

return;

if (journal_file_rotate_suggested(f, s->max_file_usec)) {

log_debug("%s: Journal header limits reached or header out-of-date, rotating.", f->path);

server_rotate(s);

server_vacuum(s);

vacuumed = true;

f = find_journal(s, uid);

if (!f)

return;

}

r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);

if (r >= 0) {

server_schedule_sync(s, priority);

return;

}

if (vacuumed || !shall_try_append_again(f, r)) {

size_t size = 0;

unsigned i;

for (i = 0; i < n; i++)

size += iovec[i].iov_len;

log_error("Failed to write entry (%d items, %zu bytes), ignoring: %s", n, size, strerror(-r));

return;

}

server_rotate(s);

server_vacuum(s);

f = find_journal(s, uid);

if (!f)

return;

log_debug("Retrying write.");

r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);

if (r < 0) {

size_t size = 0;

unsigned i;

for (i = 0; i < n; i++)

size += iovec[i].iov_len;

log_error("Failed to write entry (%d items, %zu bytes) despite vacuuming, ignoring: %s", n, size, strerror(-r));

} else

server_schedule_sync(s, priority);

}

static void dispatch_message_real(

Server *s,

struct iovec *iovec, unsigned n, unsigned m,

struct ucred *ucred,

struct timeval *tv,

const char *label, size_t label_len,

const char *unit_id,

int priority,

pid_t object_pid) {

char pid[sizeof("_PID=") + DECIMAL_STR_MAX(pid_t)],

uid[sizeof("_UID=") + DECIMAL_STR_MAX(uid_t)],

gid[sizeof("_GID=") + DECIMAL_STR_MAX(gid_t)],

owner_uid[sizeof("_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)],

source_time[sizeof("_SOURCE_REALTIME_TIMESTAMP=") + DECIMAL_STR_MAX(usec_t)],

o_uid[sizeof("OBJECT_UID=") + DECIMAL_STR_MAX(uid_t)],

o_gid[sizeof("OBJECT_GID=") + DECIMAL_STR_MAX(gid_t)],

o_owner_uid[sizeof("OBJECT_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)];

uid_t object_uid;

gid_t object_gid;

char *x;

int r;

char *t, *c;

uid_t realuid = 0, owner = 0, journal_uid;

bool owner_valid = false;

#ifdef HAVE_AUDIT

char audit_session[sizeof("_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)],

audit_loginuid[sizeof("_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)],

o_audit_session[sizeof("OBJECT_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)],

o_audit_loginuid[sizeof("OBJECT_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)];

uint32_t audit;

uid_t loginuid;

#endif

assert(s);

assert(iovec);

assert(n > 0);

assert(n + N_IOVEC_META_FIELDS + (object_pid ? N_IOVEC_OBJECT_FIELDS : 0) <= m);

if (ucred) {

realuid = ucred->uid;

sprintf(pid, "_PID=%lu", (unsigned long) ucred->pid);

IOVEC_SET_STRING(iovec[n++], pid);

sprintf(uid, "_UID=%lu", (unsigned long) ucred->uid);

IOVEC_SET_STRING(iovec[n++], uid);

sprintf(gid, "_GID=%lu", (unsigned long) ucred->gid);

IOVEC_SET_STRING(iovec[n++], gid);

r = get_process_comm(ucred->pid, &t);

if (r >= 0) {

x = strappenda("_COMM=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

r = get_process_exe(ucred->pid, &t);

if (r >= 0) {

x = strappenda("_EXE=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

r = get_process_cmdline(ucred->pid, 0, false, &t);

if (r >= 0) {

x = strappenda("_CMDLINE=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

r = get_process_capeff(ucred->pid, &t);

if (r >= 0) {

x = strappenda("_CAP_EFFECTIVE=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

#ifdef HAVE_AUDIT

r = audit_session_from_pid(ucred->pid, &audit);

if (r >= 0) {

sprintf(audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit);

IOVEC_SET_STRING(iovec[n++], audit_session);

}

r = audit_loginuid_from_pid(ucred->pid, &loginuid);

if (r >= 0) {

sprintf(audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid);

IOVEC_SET_STRING(iovec[n++], audit_loginuid);

}

#endif

r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &c);

if (r >= 0) {

char *session = NULL;

x = strappenda("_SYSTEMD_CGROUP=", c);

IOVEC_SET_STRING(iovec[n++], x);

r = cg_path_get_session(c, &t);

if (r >= 0) {

session = strappenda("_SYSTEMD_SESSION=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], session);

}

if (cg_path_get_owner_uid(c, &owner) >= 0) {

owner_valid = true;

sprintf(owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner);

IOVEC_SET_STRING(iovec[n++], owner_uid);

}

if (cg_path_get_unit(c, &t) >= 0) {

x = strappenda("_SYSTEMD_UNIT=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

} else if (unit_id && !session) {

x = strappenda("_SYSTEMD_UNIT=", unit_id);

IOVEC_SET_STRING(iovec[n++], x);

}

if (cg_path_get_user_unit(c, &t) >= 0) {

x = strappenda("_SYSTEMD_USER_UNIT=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

} else if (unit_id && session) {

x = strappenda("_SYSTEMD_USER_UNIT=", unit_id);

IOVEC_SET_STRING(iovec[n++], x);

}

if (cg_path_get_slice(c, &t) >= 0) {

x = strappenda("_SYSTEMD_SLICE=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

free(c);

} else if (unit_id) {

x = strappenda("_SYSTEMD_UNIT=", unit_id);

IOVEC_SET_STRING(iovec[n++], x);

}

#ifdef HAVE_SELINUX

if (use_selinux()) {

if (label) {

x = alloca(sizeof("_SELINUX_CONTEXT=") + label_len);

*((char*) mempcpy(stpcpy(x, "_SELINUX_CONTEXT="), label, label_len)) = 0;

IOVEC_SET_STRING(iovec[n++], x);

} else {

security_context_t con;

if (getpidcon(ucred->pid, &con) >= 0) {

x = strappenda("_SELINUX_CONTEXT=", con);

freecon(con);

IOVEC_SET_STRING(iovec[n++], x);

}

}

}

#endif

}

assert(n <= m);

if (object_pid) {

r = get_process_uid(object_pid, &object_uid);

if (r >= 0) {

sprintf(o_uid, "OBJECT_UID=%lu", (unsigned long) object_uid);

IOVEC_SET_STRING(iovec[n++], o_uid);

}

r = get_process_gid(object_pid, &object_gid);

if (r >= 0) {

sprintf(o_gid, "OBJECT_GID=%lu", (unsigned long) object_gid);

IOVEC_SET_STRING(iovec[n++], o_gid);

}

r = get_process_comm(object_pid, &t);

if (r >= 0) {

x = strappenda("OBJECT_COMM=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

r = get_process_exe(object_pid, &t);

if (r >= 0) {

x = strappenda("OBJECT_EXE=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

r = get_process_cmdline(object_pid, 0, false, &t);

if (r >= 0) {

x = strappenda("OBJECT_CMDLINE=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

#ifdef HAVE_AUDIT

r = audit_session_from_pid(object_pid, &audit);

if (r >= 0) {

sprintf(o_audit_session, "OBJECT_AUDIT_SESSION=%lu", (unsigned long) audit);

IOVEC_SET_STRING(iovec[n++], o_audit_session);

}

r = audit_loginuid_from_pid(object_pid, &loginuid);

if (r >= 0) {

sprintf(o_audit_loginuid, "OBJECT_AUDIT_LOGINUID=%lu", (unsigned long) loginuid);

IOVEC_SET_STRING(iovec[n++], o_audit_loginuid);

}

#endif

r = cg_pid_get_path_shifted(object_pid, s->cgroup_root, &c);

if (r >= 0) {

x = strappenda("OBJECT_SYSTEMD_CGROUP=", c);

IOVEC_SET_STRING(iovec[n++], x);

r = cg_path_get_session(c, &t);

if (r >= 0) {

x = strappenda("OBJECT_SYSTEMD_SESSION=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

if (cg_path_get_owner_uid(c, &owner) >= 0) {

sprintf(o_owner_uid, "OBJECT_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner);

IOVEC_SET_STRING(iovec[n++], o_owner_uid);

}

if (cg_path_get_unit(c, &t) >= 0) {

x = strappenda("OBJECT_SYSTEMD_UNIT=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

if (cg_path_get_user_unit(c, &t) >= 0) {

x = strappenda("OBJECT_SYSTEMD_USER_UNIT=", t);

free(t);

IOVEC_SET_STRING(iovec[n++], x);

}

free(c);

}

}

assert(n <= m);

if (tv) {

sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu", (unsigned long long) timeval_load(tv));

IOVEC_SET_STRING(iovec[n++], source_time);

}

/* Note that strictly speaking storing the boot id here is

if (!isempty(s->boot_id_field))

IOVEC_SET_STRING(iovec[n++], s->boot_id_field);

if (!isempty(s->machine_id_field))

IOVEC_SET_STRING(iovec[n++], s->machine_id_field);

if (!isempty(s->hostname_field))

IOVEC_SET_STRING(iovec[n++], s->hostname_field);

assert(n <= m);

if (s->split_mode == SPLIT_UID && realuid > 0)

/* Split up strictly by any UID */

journal_uid = realuid;

else if (s->split_mode == SPLIT_LOGIN && realuid > 0 && owner_valid && owner > 0)

/* Split up by login UIDs, this avoids creation of

journal_uid = owner;

else

journal_uid = 0;

write_to_journal(s, journal_uid, iovec, n, priority);

}

void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) {

char mid[11 + 32 + 1];

char buffer[16 + LINE_MAX + 1];

struct iovec iovec[N_IOVEC_META_FIELDS + 4];

int n = 0;

va_list ap;

struct ucred ucred = {};

assert(s);

assert(format);

IOVEC_SET_STRING(iovec[n++], "PRIORITY=6");

IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=driver");

memcpy(buffer, "MESSAGE=", 8);

va_start(ap, format);

vsnprintf(buffer + 8, sizeof(buffer) - 8, format, ap);

va_end(ap);

char_array_0(buffer);

IOVEC_SET_STRING(iovec[n++], buffer);

if (!sd_id128_equal(message_id, SD_ID128_NULL)) {

snprintf(mid, sizeof(mid), MESSAGE_ID(message_id));

char_array_0(mid);

IOVEC_SET_STRING(iovec[n++], mid);

}

ucred.pid = getpid();

ucred.uid = getuid();

ucred.gid = getgid();

dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0);

}

void server_dispatch_message(

Server *s,

struct iovec *iovec, unsigned n, unsigned m,

struct ucred *ucred,

struct timeval *tv,

const char *label, size_t label_len,

const char *unit_id,

int priority,

pid_t object_pid) {

int rl, r;

_cleanup_free_ char *path = NULL;

char *c;

assert(s);

assert(iovec || n == 0);

if (n == 0)

return;

if (LOG_PRI(priority) > s->max_level_store)

return;

/* Stop early in case the information will not be stored

if (s->storage == STORAGE_NONE)

return;

if (!ucred)

goto finish;

r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &path);

if (r < 0)

goto finish;

/* example: /user/lennart/3/foobar

c = strchr(path, '/');

if (c) {

c = strchr(c+1, '/');

if (c) {

c = strchr(c+1, '/');

if (c)

*c = 0;

}

}

rl = journal_rate_limit_test(s->rate_limit, path,

priority & LOG_PRIMASK, available_space(s, false));

if (rl == 0)

return;

/* Write a suppression message if we suppressed something */

if (rl > 1)

server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED,

"Suppressed %u messages from %s", rl - 1, path);

finish:

dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id, priority, object_pid);

}

static int system_journal_open(Server *s) {

int r;

char *fn;

sd_id128_t machine;

char ids[33];

r = sd_id128_get_machine(&machine);

if (r < 0) {

log_error("Failed to get machine id: %s", strerror(-r));

return r;

}

sd_id128_to_string(machine, ids);

if (!s->system_journal &&

(s->storage == STORAGE_PERSISTENT || s->storage == STORAGE_AUTO) &&

access("/run/systemd/journal/flushed", F_OK) >= 0) {

/* If in auto mode: first try to create the machine

if (s->storage == STORAGE_PERSISTENT)

(void) mkdir("/var/log/journal/", 0755);

fn = strappenda("/var/log/journal/", ids);

(void) mkdir(fn, 0755);

fn = strappenda(fn, "/system.journal");

r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &s->system_journal);

if (r >= 0)

server_fix_perms(s, s->system_journal, 0);

else if (r < 0) {

if (r != -ENOENT && r != -EROFS)

log_warning("Failed to open system journal: %s", strerror(-r));

r = 0;

}

}

if (!s->runtime_journal &&

(s->storage != STORAGE_NONE)) {

fn = strjoin("/run/log/journal/", ids, "/system.journal", NULL);

if (!fn)

return -ENOMEM;

if (s->system_journal) {

/* Try to open the runtime journal, but only

r = journal_file_open(fn, O_RDWR, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal);

free(fn);

if (r < 0) {

if (r != -ENOENT)

log_warning("Failed to open runtime journal: %s", strerror(-r));

r = 0;

}

} else {

/* OK, we really need the runtime journal, so create

(void) mkdir_parents(fn, 0755);

r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal);

free(fn);

if (r < 0) {

log_error("Failed to open runtime journal: %s", strerror(-r));

return r;

}

}

if (s->runtime_journal)

server_fix_perms(s, s->runtime_journal, 0);

}

available_space(s, true);

return r;

}

int server_flush_to_var(Server *s) {

sd_id128_t machine;

sd_journal *j = NULL;

char ts[FORMAT_TIMESPAN_MAX];

usec_t start;

unsigned n = 0;

int r;

assert(s);

if (s->storage != STORAGE_AUTO &&

s->storage != STORAGE_PERSISTENT)

return 0;

if (!s->runtime_journal)

return 0;

system_journal_open(s);

if (!s->system_journal)

return 0;

log_debug("Flushing to /var...");

start = now(CLOCK_MONOTONIC);

r = sd_id128_get_machine(&machine);

if (r < 0)

return r;

r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY);

if (r < 0) {

log_error("Failed to read runtime journal: %s", strerror(-r));

return r;

}

sd_journal_set_data_threshold(j, 0);

SD_JOURNAL_FOREACH(j) {

Object *o = NULL;

JournalFile *f;

f = j->current_file;

assert(f && f->current_offset > 0);

n++;

r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);

if (r < 0) {

log_error("Can't read entry: %s", strerror(-r));

goto finish;

}

r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);

if (r >= 0)

continue;

if (!shall_try_append_again(s->system_journal, r)) {

log_error("Can't write entry: %s", strerror(-r));

goto finish;

}

server_rotate(s);

server_vacuum(s);

if (!s->system_journal) {

log_notice("Didn't flush runtime journal since rotation of system journal wasn't successful.");

r = -EIO;

goto finish;

}

log_debug("Retrying write.");

r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL);

if (r < 0) {

log_error("Can't write entry: %s", strerror(-r));

goto finish;

}

}

finish:

journal_file_post_change(s->system_journal);

journal_file_close(s->runtime_journal);

s->runtime_journal = NULL;

if (r >= 0)

rm_rf("/run/log/journal", false, true, false);

sd_journal_close(j);

server_driver_message(s, SD_ID128_NULL, "Time spent on flushing to /var is %s for %u entries.", format_timespan(ts, sizeof(ts), now(CLOCK_MONOTONIC) - start, 0), n);

return r;

}

int process_datagram(sd_event_source *es, int fd, uint32_t revents, void *userdata) {

Server *s = userdata;

assert(s);

assert(fd == s->native_fd || fd == s->syslog_fd);

if (revents != EPOLLIN) {

log_error("Got invalid event from epoll for datagram fd: %"PRIx32, revents);

return -EIO;

}

for (;;) {

struct ucred *ucred = NULL;

struct timeval *tv = NULL;

struct cmsghdr *cmsg;

char *label = NULL;

size_t label_len = 0;

struct iovec iovec;

union {

struct cmsghdr cmsghdr;

/* We use NAME_MAX space for the

uint8_t buf[CMSG_SPACE(sizeof(struct ucred)) +

CMSG_SPACE(sizeof(struct timeval)) +

CMSG_SPACE(sizeof(int)) + /* fd */

CMSG_SPACE(NAME_MAX)]; /* selinux label */

} control = {};

struct msghdr msghdr = {

.msg_iov = &iovec,

.msg_iovlen = 1,

.msg_control = &control,

.msg_controllen = sizeof(control),

};

ssize_t n;

int v;

int *fds = NULL;

unsigned n_fds = 0;

if (ioctl(fd, SIOCINQ, &v) < 0) {

log_error("SIOCINQ failed: %m");

return -errno;

}

if (!GREEDY_REALLOC(s->buffer, s->buffer_size, LINE_MAX + (size_t) v))

return log_oom();

iovec.iov_base = s->buffer;

iovec.iov_len = s->buffer_size;

n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);

if (n < 0) {

if (errno == EINTR || errno == EAGAIN)

return 0;

log_error("recvmsg() failed: %m");

return -errno;

}

for (cmsg = CMSG_FIRSTHDR(&msghdr); cmsg; cmsg = CMSG_NXTHDR(&msghdr, cmsg)) {

if (cmsg->cmsg_level == SOL_SOCKET &&

cmsg->cmsg_type == SCM_CREDENTIALS &&

cmsg->cmsg_len == CMSG_LEN(sizeof(struct ucred)))

ucred = (struct ucred*) CMSG_DATA(cmsg);

else if (cmsg->cmsg_level == SOL_SOCKET &&

cmsg->cmsg_type == SCM_SECURITY) {

label = (char*) CMSG_DATA(cmsg);

label_len = cmsg->cmsg_len - CMSG_LEN(0);

} else if (cmsg->cmsg_level == SOL_SOCKET &&

cmsg->cmsg_type == SO_TIMESTAMP &&

cmsg->cmsg_len == CMSG_LEN(sizeof(struct timeval)))

tv = (struct timeval*) CMSG_DATA(cmsg);

else if (cmsg->cmsg_level == SOL_SOCKET &&

cmsg->cmsg_type == SCM_RIGHTS) {

fds = (int*) CMSG_DATA(cmsg);

n_fds = (cmsg->cmsg_len - CMSG_LEN(0)) / sizeof(int);

}

}

if (fd == s->syslog_fd) {

if (n > 0 && n_fds == 0) {

s->buffer[n] = 0;

server_process_syslog_message(s, strstrip(s->buffer), ucred, tv, label, label_len);

} else if (n_fds > 0)

log_warning("Got file descriptors via syslog socket. Ignoring.");

} else {

if (n > 0 && n_fds == 0)

server_process_native_message(s, s->buffer, n, ucred, tv, label, label_len);

else if (n == 0 && n_fds == 1)

server_process_native_file(s, fds[0], ucred, tv, label, label_len);

else if (n_fds > 0)

log_warning("Got too many file descriptors via native socket. Ignoring.");

}

close_many(fds, n_fds);

}

return 0;

}

static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {

Server *s = userdata;

assert(s);

log_info("Received request to flush runtime journal from PID %"PRIu32, si->ssi_pid);

touch("/run/systemd/journal/flushed");

server_flush_to_var(s);

server_sync(s);

return 0;

}

static int dispatch_sigusr2(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {

Server *s = userdata;

assert(s);

log_info("Received request to rotate journal from PID %"PRIu32, si->ssi_pid);

server_rotate(s);

server_vacuum(s);

return 0;

}

static int dispatch_sigterm(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {

Server *s = userdata;

assert(s);

log_info("Received SIG%s", signal_to_string(si->ssi_signo));

sd_event_exit(s->event, 0);

return 0;

}

static int setup_signals(Server *s) {

sigset_t mask;

int r;

assert(s);

assert_se(sigemptyset(&mask) == 0);

sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, -1);

assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);

r = sd_event_add_signal(s->event, SIGUSR1, dispatch_sigusr1, s, &s->sigusr1_event_source);

if (r < 0)

return r;

r = sd_event_add_signal(s->event, SIGUSR2, dispatch_sigusr2, s, &s->sigusr2_event_source);

if (r < 0)

return r;

r = sd_event_add_signal(s->event, SIGTERM, dispatch_sigterm, s, &s->sigterm_event_source);

if (r < 0)

return r;

r = sd_event_add_signal(s->event, SIGINT, dispatch_sigterm, s, &s->sigint_event_source);

if (r < 0)

return r;

return 0;

}

static int server_parse_proc_cmdline(Server *s) {

_cleanup_free_ char *line = NULL;

char *w, *state;

size_t l;

int r;

r = proc_cmdline(&line);

if (r < 0)

log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r));

if (r <= 0)

return 0;

FOREACH_WORD_QUOTED(w, l, line, state) {

_cleanup_free_ char *word;

word = strndup(w, l);

if (!word)

return -ENOMEM;

if (startswith(word, "systemd.journald.forward_to_syslog=")) {

r = parse_boolean(word + 35);

if (r < 0)

log_warning("Failed to parse forward to syslog switch %s. Ignoring.", word + 35);

else

s->forward_to_syslog = r;

} else if (startswith(word, "systemd.journald.forward_to_kmsg=")) {

r = parse_boolean(word + 33);

if (r < 0)

log_warning("Failed to parse forward to kmsg switch %s. Ignoring.", word + 33);

else

s->forward_to_kmsg = r;

} else if (startswith(word, "systemd.journald.forward_to_console=")) {

r = parse_boolean(word + 36);

if (r < 0)

log_warning("Failed to parse forward to console switch %s. Ignoring.", word + 36);

else

s->forward_to_console = r;

} else if (startswith(word, "systemd.journald"))

log_warning("Invalid systemd.journald parameter. Ignoring.");

}

return 0;

}

static int server_parse_config_file(Server *s) {

static const char fn[] = "/etc/systemd/journald.conf";

_cleanup_fclose_ FILE *f = NULL;

int r;

assert(s);

f = fopen(fn, "re");

if (!f) {

if (errno == ENOENT)

return 0;

log_warning("Failed to open configuration file %s: %m", fn);

return -errno;

}

r = config_parse(NULL, fn, f, "Journal\0", config_item_perf_lookup,

(void*) journald_gperf_lookup, false, false, s);

if (r < 0)

log_warning("Failed to parse configuration file: %s", strerror(-r));

return r;

}

static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) {

Server *s = userdata;

assert(s);

server_sync(s);

return 0;

}

int server_schedule_sync(Server *s, int priority) {

int r;

assert(s);

if (priority <= LOG_CRIT) {

/* Immediately sync to disk when this is of priority CRIT, ALERT, EMERG */

server_sync(s);

return 0;

}

if (s->sync_scheduled)

return 0;