#include <sys/timerfd.h>
#include <libudev.h>
-#include <systemd/sd-journal.h>
-#include <systemd/sd-messages.h>
-#include <systemd/sd-daemon.h>
+#include "sd-journal.h"
+#include "sd-messages.h"
+#include "sd-daemon.h"
#include "fileio.h"
#include "mkdir.h"
#include "hashmap.h"
#include "socket-util.h"
#include "cgroup-util.h"
#include "list.h"
-#include "virt.h"
#include "missing.h"
#include "conf-parser.h"
+#include "selinux-util.h"
#include "journal-internal.h"
#include "journal-vacuum.h"
#include "journal-authenticate.h"
-#include "journald-server.h"
#include "journald-rate-limit.h"
#include "journald-kmsg.h"
#include "journald-syslog.h"
#include "journald-stream.h"
#include "journald-console.h"
#include "journald-native.h"
-#include "selinux-util.h"
+#include "journald-server.h"
#ifdef HAVE_ACL
#include <sys/acl.h>
#define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE)
#define DEFAULT_RATE_LIMIT_INTERVAL (30*USEC_PER_SEC)
#define DEFAULT_RATE_LIMIT_BURST 1000
+#define DEFAULT_MAX_FILE_USEC USEC_PER_MONTH
#define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC)
-static const char* const storage_table[] = {
+static const char* const storage_table[_STORAGE_MAX] = {
[STORAGE_AUTO] = "auto",
[STORAGE_VOLATILE] = "volatile",
[STORAGE_PERSISTENT] = "persistent",
DEFINE_STRING_TABLE_LOOKUP(storage, Storage);
DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting");
-static const char* const split_mode_table[] = {
- [SPLIT_NONE] = "none",
+static const char* const split_mode_table[_SPLIT_MAX] = {
+ [SPLIT_LOGIN] = "login",
[SPLIT_UID] = "uid",
- [SPLIT_LOGIN] = "login"
+ [SPLIT_NONE] = "none",
};
DEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode);
for (;;) {
struct stat st;
struct dirent *de;
- union dirent_storage buf;
- r = readdir_r(d, &buf.de, &de);
- if (r != 0)
- break;
+ errno = 0;
+ de = readdir(d);
+ if (!de && errno != 0)
+ return 0;
if (!de)
break;
}
ss_avail = ss.f_bsize * ss.f_bavail;
- avail = ss_avail > m->keep_free ? ss_avail - m->keep_free : 0;
- s->cached_available_space = MIN(m->max_use, avail) > sum ? MIN(m->max_use, avail) - sum : 0;
+ /* If we reached a high mark, we will always allow this much
+ * again, unless usage goes above max_use. This watermark
+ * value is cached so that we don't give up space on pressure,
+ * but hover below the maximum usage. */
+
+ if (m->use < sum)
+ m->use = sum;
+
+ avail = LESS_BY(ss_avail, m->keep_free);
+
+ s->cached_available_space = LESS_BY(MIN(m->max_use, avail), sum);
s->cached_available_space_timestamp = ts;
if (verbose) {
fb4[FORMAT_BYTES_MAX], fb5[FORMAT_BYTES_MAX];
server_driver_message(s, SD_MESSAGE_JOURNAL_USAGE,
- "%s journal is using %s (max %s, leaving %s of free %s, current limit %s).",
+ "%s journal is using %s (max allowed %s, "
+ "trying to leave %s free of %s available → current limit %s).",
s->system_journal ? "Permanent" : "Runtime",
format_bytes(fb1, sizeof(fb1), sum),
format_bytes(fb2, sizeof(fb2), m->max_use),
format_bytes(fb3, sizeof(fb3), m->keep_free),
format_bytes(fb4, sizeof(fb4), ss_avail),
- format_bytes(fb5, sizeof(fb5), MIN(m->max_use, avail)));
+ format_bytes(fb5, sizeof(fb5), s->cached_available_space + sum));
}
return s->cached_available_space;
log_warning("Failed to fix access mode on %s, ignoring: %s", f->path, strerror(-r));
#ifdef HAVE_ACL
- if (uid <= 0)
+ if (uid <= SYSTEM_UID_MAX)
return;
acl = acl_get_fd(f->fd);
if (s->runtime_journal)
return s->runtime_journal;
- if (uid <= 0)
+ if (uid <= SYSTEM_UID_MAX)
return s->system_journal;
r = sd_id128_get_machine(&machine);
if (f)
return f;
- if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-%lu.journal",
- SD_ID128_FORMAT_VAL(machine), (unsigned long) uid) < 0)
+ if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-"UID_FMT".journal",
+ SD_ID128_FORMAT_VAL(machine), uid) < 0)
return s->system_journal;
while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) {
return f;
}
+static int do_rotate(Server *s, JournalFile **f, const char* name,
+ bool seal, uint32_t uid) {
+ int r;
+ assert(s);
+
+ if (!*f)
+ return -EINVAL;
+
+ r = journal_file_rotate(f, s->compress, seal);
+ if (r < 0)
+ if (*f)
+ log_error("Failed to rotate %s: %s",
+ (*f)->path, strerror(-r));
+ else
+ log_error("Failed to create new %s journal: %s",
+ name, strerror(-r));
+ else
+ server_fix_perms(s, *f, uid);
+ return r;
+}
+
void server_rotate(Server *s) {
JournalFile *f;
void *k;
log_debug("Rotating...");
- if (s->runtime_journal) {
- r = journal_file_rotate(&s->runtime_journal, s->compress, false);
- if (r < 0)
- if (s->runtime_journal)
- log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
- else
- log_error("Failed to create new runtime journal: %s", strerror(-r));
- else
- server_fix_perms(s, s->runtime_journal, 0);
- }
-
- if (s->system_journal) {
- r = journal_file_rotate(&s->system_journal, s->compress, s->seal);
- if (r < 0)
- if (s->system_journal)
- log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
- else
- log_error("Failed to create new system journal: %s", strerror(-r));
-
- else
- server_fix_perms(s, s->system_journal, 0);
- }
+ do_rotate(s, &s->runtime_journal, "runtime", false, 0);
+ do_rotate(s, &s->system_journal, "system", s->seal, 0);
HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
- r = journal_file_rotate(&f, s->compress, s->seal);
- if (r < 0)
- if (f)
- log_error("Failed to rotate %s: %s", f->path, strerror(-r));
- else {
- log_error("Failed to create user journal: %s", strerror(-r));
- hashmap_remove(s->user_journals, k);
- }
- else {
+ r = do_rotate(s, &f, "user", s->seal, PTR_TO_UINT32(k));
+ if (r >= 0)
hashmap_replace(s->user_journals, k, f);
- server_fix_perms(s, f, PTR_TO_UINT32(k));
- }
+ else if (!f)
+ /* Old file has been closed and deallocated */
+ hashmap_remove(s->user_journals, k);
}
}
void server_sync(Server *s) {
- static const struct itimerspec sync_timer_disable = {};
JournalFile *f;
void *k;
Iterator i;
log_error("Failed to sync user journal: %s", strerror(-r));
}
- r = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_disable, NULL);
- if (r < 0)
- log_error("Failed to disable max timer: %m");
+ if (s->sync_event_source) {
+ r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_OFF);
+ if (r < 0)
+ log_error("Failed to disable sync timer source: %s", strerror(-r));
+ }
s->sync_scheduled = false;
}
+static void do_vacuum(Server *s, char *ids, JournalFile *f, const char* path,
+ JournalMetrics *metrics) {
+ char *p;
+ int r;
+
+ if (!f)
+ return;
+
+ p = strappenda(path, ids);
+ r = journal_directory_vacuum(p, metrics->max_use, s->max_retention_usec, &s->oldest_file_usec);
+ if (r < 0 && r != -ENOENT)
+ log_error("Failed to vacuum %s: %s", p, strerror(-r));
+}
+
void server_vacuum(Server *s) {
char ids[33];
sd_id128_t machine;
log_error("Failed to get machine ID: %s", strerror(-r));
return;
}
-
sd_id128_to_string(machine, ids);
- if (s->system_journal) {
- char *p = strappenda("/var/log/journal/", ids);
+ do_vacuum(s, ids, s->system_journal, "/var/log/journal/", &s->system_metrics);
+ do_vacuum(s, ids, s->runtime_journal, "/run/log/journal/", &s->runtime_metrics);
- r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec);
- if (r < 0 && r != -ENOENT)
- log_error("Failed to vacuum %s: %s", p, strerror(-r));
- }
+ s->cached_available_space_timestamp = 0;
+}
- if (s->runtime_journal) {
- char *p = strappenda("/run/log/journal/", ids);
+static void server_cache_machine_id(Server *s) {
+ sd_id128_t id;
+ int r;
- r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec);
- if (r < 0 && r != -ENOENT)
- log_error("Failed to vacuum %s: %s", p, strerror(-r));
- }
+ assert(s);
- s->cached_available_space_timestamp = 0;
+ 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) {
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)],
- boot_id[sizeof("_BOOT_ID=") + 32] = "_BOOT_ID=",
- machine_id[sizeof("_MACHINE_ID=") + 32] = "_MACHINE_ID=",
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;
- sd_id128_t id;
int r;
char *t, *c;
uid_t realuid = 0, owner = 0, journal_uid;
if (ucred) {
realuid = ucred->uid;
- sprintf(pid, "_PID=%lu", (unsigned long) ucred->pid);
+ sprintf(pid, "_PID="PID_FMT, ucred->pid);
IOVEC_SET_STRING(iovec[n++], pid);
- sprintf(uid, "_UID=%lu", (unsigned long) ucred->uid);
+ sprintf(uid, "_UID="UID_FMT, ucred->uid);
IOVEC_SET_STRING(iovec[n++], uid);
- sprintf(gid, "_GID=%lu", (unsigned long) ucred->gid);
+ sprintf(gid, "_GID="GID_FMT, ucred->gid);
IOVEC_SET_STRING(iovec[n++], gid);
r = get_process_comm(ucred->pid, &t);
#ifdef HAVE_AUDIT
r = audit_session_from_pid(ucred->pid, &audit);
if (r >= 0) {
- sprintf(audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit);
+ sprintf(audit_session, "_AUDIT_SESSION=%"PRIu32, 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);
+ sprintf(audit_loginuid, "_AUDIT_LOGINUID="UID_FMT, loginuid);
IOVEC_SET_STRING(iovec[n++], audit_loginuid);
}
#endif
- r = cg_pid_get_path_shifted(ucred->pid, NULL, &c);
+ r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &c);
if (r >= 0) {
char *session = NULL;
if (cg_path_get_owner_uid(c, &owner) >= 0) {
owner_valid = true;
- sprintf(owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner);
+ sprintf(owner_uid, "_SYSTEMD_OWNER_UID="UID_FMT, owner);
IOVEC_SET_STRING(iovec[n++], owner_uid);
}
}
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);
+ x = alloca(strlen("_SELINUX_CONTEXT=") + label_len + 1);
*((char*) mempcpy(stpcpy(x, "_SELINUX_CONTEXT="), label, label_len)) = 0;
IOVEC_SET_STRING(iovec[n++], x);
if (object_pid) {
r = get_process_uid(object_pid, &object_uid);
if (r >= 0) {
- sprintf(o_uid, "OBJECT_UID=%lu", (unsigned long) object_uid);
+ sprintf(o_uid, "OBJECT_UID="UID_FMT, 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);
+ sprintf(o_gid, "OBJECT_GID="GID_FMT, object_gid);
IOVEC_SET_STRING(iovec[n++], o_gid);
}
#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);
+ sprintf(o_audit_session, "OBJECT_AUDIT_SESSION=%"PRIu32, 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);
+ sprintf(o_audit_loginuid, "OBJECT_AUDIT_LOGINUID="UID_FMT, loginuid);
IOVEC_SET_STRING(iovec[n++], o_audit_loginuid);
}
#endif
- r = cg_pid_get_path_shifted(object_pid, NULL, &c);
+ 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);
}
if (cg_path_get_owner_uid(c, &owner) >= 0) {
- sprintf(o_owner_uid, "OBJECT_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner);
+ sprintf(o_owner_uid, "OBJECT_SYSTEMD_OWNER_UID="UID_FMT, owner);
IOVEC_SET_STRING(iovec[n++], o_owner_uid);
}
/* Note that strictly speaking storing the boot id here is
* redundant since the entry includes this in-line
* anyway. However, we need this indexed, too. */
- r = sd_id128_get_boot(&id);
- if (r >= 0) {
- sd_id128_to_string(id, boot_id + strlen("_BOOT_ID="));
- IOVEC_SET_STRING(iovec[n++], boot_id);
- }
+ if (!isempty(s->boot_id_field))
+ IOVEC_SET_STRING(iovec[n++], s->boot_id_field);
- r = sd_id128_get_machine(&id);
- if (r >= 0) {
- sd_id128_to_string(id, machine_id + strlen("_MACHINE_ID="));
- IOVEC_SET_STRING(iovec[n++], machine_id);
- }
+ if (!isempty(s->machine_id_field))
+ IOVEC_SET_STRING(iovec[n++], s->machine_id_field);
- t = gethostname_malloc();
- if (t) {
- x = strappenda("_HOSTNAME=", t);
- free(t);
- IOVEC_SET_STRING(iovec[n++], x);
- }
+ if (!isempty(s->hostname_field))
+ IOVEC_SET_STRING(iovec[n++], s->hostname_field);
assert(n <= m);
/* 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
- * individual journals for system UIDs. We do this
- * only if the realuid is not root, in order not to
- * accidentally leak privileged information to the
- * user that is logged by a privileged process that is
- * part of an unprivileged session.*/
+ /* Split up by login UIDs. We do this only if the
+ * realuid is not root, in order not to accidentally
+ * leak privileged information to the user that is
+ * logged by a privileged process that is part of an
+ * unprivileged session.*/
journal_uid = owner;
else
journal_uid = 0;
if (!ucred)
goto finish;
- r = cg_pid_get_path_shifted(ucred->pid, NULL, &path);
+ r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &path);
if (r < 0)
goto finish;
/* OK, we really need the runtime journal, so create
* it if necessary. */
- (void) mkdir_parents(fn, 0755);
+ (void) mkdir("/run/log", 0755);
+ (void) mkdir("/run/log/journal", 0755);
+ (void) mkdir_parents(fn, 0750);
+
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);
}
int server_flush_to_var(Server *s) {
- int r;
sd_id128_t machine;
sd_journal *j = NULL;
+ char ts[FORMAT_TIMESPAN_MAX];
+ usec_t start;
+ unsigned n = 0;
+ int r;
assert(s);
log_debug("Flushing to /var...");
+ start = now(CLOCK_MONOTONIC);
+
r = sd_id128_get_machine(&machine);
if (r < 0)
return r;
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));
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_event(Server *s, struct epoll_event *ev) {
+int process_datagram(sd_event_source *es, int fd, uint32_t revents, void *userdata) {
+ Server *s = userdata;
+
assert(s);
- assert(ev);
+ 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 SELinux label
+ * here. The kernel currently enforces no
+ * limit, but according to suggestions from
+ * the SELinux people this will change and it
+ * will probably be identical to NAME_MAX. For
+ * now we use that, but this should be updated
+ * one day when the final limit is known.*/
+ 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),
+ };
- if (ev->data.fd == s->signal_fd) {
- struct signalfd_siginfo sfsi;
ssize_t n;
+ int v;
+ int *fds = NULL;
+ unsigned n_fds = 0;
- if (ev->events != EPOLLIN) {
- log_error("Got invalid event from epoll for %s: %"PRIx32,
- "signal fd", ev->events);
- return -EIO;
+ if (ioctl(fd, SIOCINQ, &v) < 0) {
+ log_error("SIOCINQ failed: %m");
+ return -errno;
}
- n = read(s->signal_fd, &sfsi, sizeof(sfsi));
- if (n != sizeof(sfsi)) {
+ if (!GREEDY_REALLOC(s->buffer, s->buffer_size, LINE_MAX + (size_t) v))
+ return log_oom();
- if (n >= 0)
- return -EIO;
+ 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 1;
+ return 0;
+ log_error("recvmsg() failed: %m");
return -errno;
}
- if (sfsi.ssi_signo == SIGUSR1) {
- log_info("Received request to flush runtime journal from PID %"PRIu32,
- sfsi.ssi_pid);
- touch("/run/systemd/journal/flushed");
- server_flush_to_var(s);
- server_sync(s);
- return 1;
- }
-
- if (sfsi.ssi_signo == SIGUSR2) {
- log_info("Received request to rotate journal from PID %"PRIu32,
- sfsi.ssi_pid);
- server_rotate(s);
- server_vacuum(s);
- return 1;
- }
-
- log_info("Received SIG%s", signal_to_string(sfsi.ssi_signo));
-
- return 0;
-
- } else if (ev->data.fd == s->sync_timer_fd) {
- int r;
- uint64_t t;
-
- log_debug("Got sync request from epoll.");
-
- r = read(ev->data.fd, (void *)&t, sizeof(t));
- if (r < 0)
- return 0;
-
- server_sync(s);
- return 1;
-
- } else if (ev->data.fd == s->dev_kmsg_fd) {
- int r;
-
- if (ev->events & EPOLLERR)
- log_warning("/dev/kmsg buffer overrun, some messages lost.");
-
- if (!(ev->events & EPOLLIN)) {
- log_error("Got invalid event from epoll for %s: %"PRIx32,
- "/dev/kmsg", ev->events);
- return -EIO;
+ 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);
+ }
}
- r = server_read_dev_kmsg(s);
- if (r < 0)
- return r;
-
- return 1;
+ 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 (ev->data.fd == s->native_fd ||
- ev->data.fd == s->syslog_fd) {
-
- if (ev->events != EPOLLIN) {
- log_error("Got invalid event from epoll for %s: %"PRIx32,
- ev->data.fd == s->native_fd ? "native fd" : "syslog fd",
- ev->events);
- return -EIO;
+ } 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.");
}
- for (;;) {
- struct msghdr msghdr;
- struct iovec iovec;
- struct ucred *ucred = NULL;
- struct timeval *tv = NULL;
- struct cmsghdr *cmsg;
- char *label = NULL;
- size_t label_len = 0;
- union {
- struct cmsghdr cmsghdr;
-
- /* We use NAME_MAX space for the
- * SELinux label here. The kernel
- * currently enforces no limit, but
- * according to suggestions from the
- * SELinux people this will change and
- * it will probably be identical to
- * NAME_MAX. For now we use that, but
- * this should be updated one day when
- * the final limit is known.*/
- 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;
- ssize_t n;
- int v;
- int *fds = NULL;
- unsigned n_fds = 0;
-
- if (ioctl(ev->data.fd, SIOCINQ, &v) < 0) {
- log_error("SIOCINQ failed: %m");
- return -errno;
- }
-
- if (s->buffer_size < (size_t) v) {
- void *b;
- size_t l;
-
- l = MAX(LINE_MAX + (size_t) v, s->buffer_size * 2);
- b = realloc(s->buffer, l+1);
-
- if (!b) {
- log_error("Couldn't increase buffer.");
- return -ENOMEM;
- }
-
- s->buffer_size = l;
- s->buffer = b;
- }
-
- zero(iovec);
- iovec.iov_base = s->buffer;
- iovec.iov_len = s->buffer_size;
-
- zero(control);
- zero(msghdr);
- msghdr.msg_iov = &iovec;
- msghdr.msg_iovlen = 1;
- msghdr.msg_control = &control;
- msghdr.msg_controllen = sizeof(control);
-
- n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
- if (n < 0) {
-
- if (errno == EINTR || errno == EAGAIN)
- return 1;
-
- 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);
- }
- }
+ close_many(fds, n_fds);
+ }
+}
- if (ev->data.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.");
+static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {
+ Server *s = userdata;
- } 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);
- }
+ assert(s);
- return 1;
+ log_info("Received request to flush runtime journal from PID %"PRIu32, si->ssi_pid);
- } else if (ev->data.fd == s->stdout_fd) {
+ touch("/run/systemd/journal/flushed");
+ server_flush_to_var(s);
+ server_sync(s);
- if (ev->events != EPOLLIN) {
- log_error("Got invalid event from epoll for %s: %"PRIx32,
- "stdout fd", ev->events);
- return -EIO;
- }
+ return 0;
+}
- stdout_stream_new(s);
- return 1;
+static int dispatch_sigusr2(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {
+ Server *s = userdata;
- } else {
- StdoutStream *stream;
+ assert(s);
- if ((ev->events|EPOLLIN|EPOLLHUP) != (EPOLLIN|EPOLLHUP)) {
- log_error("Got invalid event from epoll for %s: %"PRIx32,
- "stdout stream", ev->events);
- return -EIO;
- }
+ log_info("Received request to rotate journal from PID %"PRIu32, si->ssi_pid);
+ server_rotate(s);
+ server_vacuum(s);
- /* If it is none of the well-known fds, it must be an
- * stdout stream fd. Note that this is a bit ugly here
- * (since we rely that none of the well-known fds
- * could be interpreted as pointer), but nonetheless
- * safe, since the well-known fds would never get an
- * fd > 4096, i.e. beyond the first memory page */
+ return 0;
+}
- stream = ev->data.ptr;
+static int dispatch_sigterm(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {
+ Server *s = userdata;
- if (stdout_stream_process(stream) <= 0)
- stdout_stream_free(stream);
+ assert(s);
- return 1;
- }
+ log_received_signal(LOG_INFO, si);
- log_error("Unknown event.");
+ sd_event_exit(s->event, 0);
return 0;
}
-static int open_signalfd(Server *s) {
+static int setup_signals(Server *s) {
sigset_t mask;
- struct epoll_event ev;
+ int r;
assert(s);
sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, -1);
assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0);
- s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC);
- if (s->signal_fd < 0) {
- log_error("signalfd(): %m");
- return -errno;
- }
+ r = sd_event_add_signal(s->event, &s->sigusr1_event_source, SIGUSR1, dispatch_sigusr1, s);
+ if (r < 0)
+ return r;
- zero(ev);
- ev.events = EPOLLIN;
- ev.data.fd = s->signal_fd;
+ r = sd_event_add_signal(s->event, &s->sigusr2_event_source, SIGUSR2, dispatch_sigusr2, s);
+ if (r < 0)
+ return r;
- if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) {
- log_error("epoll_ctl(): %m");
- return -errno;
- }
+ r = sd_event_add_signal(s->event, &s->sigterm_event_source, SIGTERM, dispatch_sigterm, s);
+ if (r < 0)
+ return r;
+
+ r = sd_event_add_signal(s->event, &s->sigint_event_source, SIGINT, dispatch_sigterm, s);
+ if (r < 0)
+ return r;
return 0;
}
static int server_parse_proc_cmdline(Server *s) {
_cleanup_free_ char *line = NULL;
- char *w, *state;
- int r;
+ const char *w, *state;
size_t l;
+ int r;
- if (detect_container(NULL) > 0)
- return 0;
-
- r = read_one_line_file("/proc/cmdline", &line);
- if (r < 0) {
+ 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;
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.forward_to_wall=")) {
+ r = parse_boolean(word + 33);
+ if (r < 0)
+ log_warning("Failed to parse forward to wall switch %s. Ignoring.", word + 33);
+ else
+ s->forward_to_wall = r;
} else if (startswith(word, "systemd.journald"))
log_warning("Invalid systemd.journald parameter. Ignoring.");
}
+ /* do not warn about state here, since probably systemd already did */
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;
+ return config_parse(NULL, "/etc/systemd/journald.conf", NULL,
+ "Journal\0",
+ config_item_perf_lookup, journald_gperf_lookup,
+ false, false, true, s);
}
-static int server_open_sync_timer(Server *s) {
- int r;
- struct epoll_event ev;
+static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) {
+ Server *s = userdata;
assert(s);
- s->sync_timer_fd = timerfd_create(CLOCK_MONOTONIC, TFD_CLOEXEC);
- if (s->sync_timer_fd < 0)
- return -errno;
-
- zero(ev);
- ev.events = EPOLLIN;
- ev.data.fd = s->sync_timer_fd;
-
- r = epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->sync_timer_fd, &ev);
- if (r < 0) {
- log_error("Failed to add idle timer fd to epoll object: %m");
- return -errno;
- }
-
+ server_sync(s);
return 0;
}
if (s->sync_scheduled)
return 0;
- if (s->sync_interval_usec) {
- struct itimerspec sync_timer_enable = {};
+ if (s->sync_interval_usec > 0) {
+ usec_t when;
+
+ r = sd_event_now(s->event, CLOCK_MONOTONIC, &when);
+ if (r < 0)
+ return r;
+
+ when += s->sync_interval_usec;
- timespec_store(&sync_timer_enable.it_value, s->sync_interval_usec);
+ if (!s->sync_event_source) {
+ r = sd_event_add_time(
+ s->event,
+ &s->sync_event_source,
+ CLOCK_MONOTONIC,
+ when, 0,
+ server_dispatch_sync, s);
+ if (r < 0)
+ return r;
- r = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_enable, NULL);
+ r = sd_event_source_set_priority(s->sync_event_source, SD_EVENT_PRIORITY_IMPORTANT);
+ } else {
+ r = sd_event_source_set_time(s->sync_event_source, when);
+ if (r < 0)
+ return r;
+
+ r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_ONESHOT);
+ }
if (r < 0)
- return -errno;
+ return r;
+
+ s->sync_scheduled = true;
}
- s->sync_scheduled = true;
+ return 0;
+}
+
+static int dispatch_hostname_change(sd_event_source *es, int fd, uint32_t revents, void *userdata) {
+ Server *s = userdata;
+
+ assert(s);
+
+ server_cache_hostname(s);
+ return 0;
+}
+
+static int server_open_hostname(Server *s) {
+ int r;
+
+ assert(s);
+
+ s->hostname_fd = open("/proc/sys/kernel/hostname", O_RDONLY|O_CLOEXEC|O_NDELAY|O_NOCTTY);
+ if (s->hostname_fd < 0) {
+ log_error("Failed to open /proc/sys/kernel/hostname: %m");
+ return -errno;
+ }
+
+ r = sd_event_add_io(s->event, &s->hostname_event_source, s->hostname_fd, 0, dispatch_hostname_change, s);
+ if (r < 0) {
+ /* kernels prior to 3.2 don't support polling this file. Ignore
+ * the failure. */
+ if (r == -EPERM) {
+ log_warning("Failed to register hostname fd in event loop: %s. Ignoring.",
+ strerror(-r));
+ s->hostname_fd = safe_close(s->hostname_fd);
+ return 0;
+ }
+
+ log_error("Failed to register hostname fd in event loop: %s", strerror(-r));
+ return r;
+ }
+
+ r = sd_event_source_set_priority(s->hostname_event_source, SD_EVENT_PRIORITY_IMPORTANT-10);
+ if (r < 0) {
+ log_error("Failed to adjust priority of host name event source: %s", strerror(-r));
+ return r;
+ }
return 0;
}
assert(s);
zero(*s);
- s->sync_timer_fd = s->syslog_fd = s->native_fd = s->stdout_fd =
- s->signal_fd = s->epoll_fd = s->dev_kmsg_fd = -1;
+ s->syslog_fd = s->native_fd = s->stdout_fd = s->dev_kmsg_fd = s->hostname_fd = -1;
s->compress = true;
s->seal = true;
s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL;
s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST;
- s->forward_to_syslog = true;
+ s->forward_to_wall = true;
+
+ s->max_file_usec = DEFAULT_MAX_FILE_USEC;
s->max_level_store = LOG_DEBUG;
s->max_level_syslog = LOG_DEBUG;
s->max_level_kmsg = LOG_NOTICE;
s->max_level_console = LOG_INFO;
+ s->max_level_wall = LOG_EMERG;
memset(&s->system_metrics, 0xFF, sizeof(s->system_metrics));
memset(&s->runtime_metrics, 0xFF, sizeof(s->runtime_metrics));
server_parse_config_file(s);
server_parse_proc_cmdline(s);
if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) {
- log_debug("Setting both rate limit interval and burst from %llu,%u to 0,0",
- (long long unsigned) s->rate_limit_interval,
- s->rate_limit_burst);
+ log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0",
+ s->rate_limit_interval, s->rate_limit_burst);
s->rate_limit_interval = s->rate_limit_burst = 0;
}
mkdir_p("/run/systemd/journal", 0755);
- s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func);
+ s->user_journals = hashmap_new(NULL);
if (!s->user_journals)
return log_oom();
if (!s->mmap)
return log_oom();
- s->epoll_fd = epoll_create1(EPOLL_CLOEXEC);
- if (s->epoll_fd < 0) {
- log_error("Failed to create epoll object: %m");
- return -errno;
+ r = sd_event_default(&s->event);
+ if (r < 0) {
+ log_error("Failed to create event loop: %s", strerror(-r));
+ return r;
}
+ sd_event_set_watchdog(s->event, true);
+
n = sd_listen_fds(true);
if (n < 0) {
log_error("Failed to read listening file descriptors from environment: %s", strerror(-n));
s->stdout_fd = fd;
- } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) {
+ } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0 ||
+ sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/dev-log", 0) > 0) {
if (s->syslog_fd >= 0) {
log_error("Too many /dev/log sockets passed.");
if (r < 0)
return r;
- r = server_open_sync_timer(s);
+ r = server_open_hostname(s);
if (r < 0)
return r;
- r = open_signalfd(s);
+ r = setup_signals(s);
if (r < 0)
return r;
if (!s->udev)
return -ENOMEM;
- s->rate_limit = journal_rate_limit_new(s->rate_limit_interval,
- s->rate_limit_burst);
+ s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst);
if (!s->rate_limit)
return -ENOMEM;
+ r = cg_get_root_path(&s->cgroup_root);
+ if (r < 0)
+ return r;
+
+ server_cache_hostname(s);
+ server_cache_boot_id(s);
+ server_cache_machine_id(s);
+
r = system_journal_open(s);
if (r < 0)
return r;
hashmap_free(s->user_journals);
- if (s->epoll_fd >= 0)
- close_nointr_nofail(s->epoll_fd);
-
- if (s->signal_fd >= 0)
- close_nointr_nofail(s->signal_fd);
-
- if (s->syslog_fd >= 0)
- close_nointr_nofail(s->syslog_fd);
-
- if (s->native_fd >= 0)
- close_nointr_nofail(s->native_fd);
-
- if (s->stdout_fd >= 0)
- close_nointr_nofail(s->stdout_fd);
-
- if (s->dev_kmsg_fd >= 0)
- close_nointr_nofail(s->dev_kmsg_fd);
-
- if (s->sync_timer_fd >= 0)
- close_nointr_nofail(s->sync_timer_fd);
+ sd_event_source_unref(s->syslog_event_source);
+ sd_event_source_unref(s->native_event_source);
+ sd_event_source_unref(s->stdout_event_source);
+ sd_event_source_unref(s->dev_kmsg_event_source);
+ sd_event_source_unref(s->sync_event_source);
+ sd_event_source_unref(s->sigusr1_event_source);
+ sd_event_source_unref(s->sigusr2_event_source);
+ sd_event_source_unref(s->sigterm_event_source);
+ sd_event_source_unref(s->sigint_event_source);
+ sd_event_source_unref(s->hostname_event_source);
+ sd_event_unref(s->event);
+
+ safe_close(s->syslog_fd);
+ safe_close(s->native_fd);
+ safe_close(s->stdout_fd);
+ safe_close(s->dev_kmsg_fd);
+ safe_close(s->hostname_fd);
if (s->rate_limit)
journal_rate_limit_free(s->rate_limit);
free(s->buffer);
free(s->tty_path);
+ free(s->cgroup_root);
if (s->mmap)
mmap_cache_unref(s->mmap);
~ianmdlvl / elogind.git / blobdiff