#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 "journald-server.h"
#ifdef HAVE_ACL
#include <sys/acl.h>
return s->cached_available_space;
}
-static void server_read_file_gid(Server *s) {
- const char *g = "systemd-journal";
- int r;
-
- assert(s);
-
- if (s->file_gid_valid)
- return;
-
- r = get_group_creds(&g, &s->file_gid);
- if (r < 0)
- log_warning("Failed to resolve '%s' group: %s", g, strerror(-r));
-
- /* if we couldn't read the gid, then it will be 0, but that's
- * fine and we shouldn't try to resolve the group again, so
- * let's just pretend it worked right-away. */
- s->file_gid_valid = true;
-}
-
void server_fix_perms(Server *s, JournalFile *f, uid_t uid) {
int r;
#ifdef HAVE_ACL
assert(f);
- server_read_file_gid(s);
-
- r = fchmod_and_fchown(f->fd, 0640, 0, s->file_gid);
+ r = fchmod(f->fd, 0640);
if (r < 0)
- log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r));
+ log_warning("Failed to fix access mode on %s, ignoring: %s", f->path, strerror(-r));
#ifdef HAVE_ACL
if (uid <= 0)
if (r < 0)
if (f)
log_error("Failed to rotate %s: %s", f->path, strerror(-r));
- else
+ else {
log_error("Failed to create user journal: %s", strerror(-r));
+ hashmap_remove(s->user_journals, k);
+ }
else {
hashmap_replace(s->user_journals, k, f);
server_fix_perms(s, f, PTR_TO_UINT32(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;
}
}
free(c);
+ } else if (unit_id) {
+ x = strappenda("_SYSTEMD_UNIT=", unit_id);
+ IOVEC_SET_STRING(iovec[n++], x);
}
#ifdef HAVE_SELINUX
- if (label) {
- x = alloca(sizeof("_SELINUX_CONTEXT=") + label_len);
+ 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;
+ *((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);
+ if (getpidcon(ucred->pid, &con) >= 0) {
+ x = strappenda("_SELINUX_CONTEXT=", con);
- freecon(con);
- IOVEC_SET_STRING(iovec[n++], x);
+ freecon(con);
+ IOVEC_SET_STRING(iovec[n++], x);
+ }
}
}
#endif
}
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.");
- 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 != EPOLLIN) {
- log_error("Got invalid event from epoll.");
- 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.");
- 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.");
+ return 0;
+}
- } 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.");
- }
+static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) {
+ Server *s = userdata;
- 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.");
- 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.");
- 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_info("Received SIG%s", signal_to_string(si->ssi_signo));
- log_error("Unknown event.");
+ sd_event_request_quit(s->event);
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, SIGUSR1, dispatch_sigusr1, s, &s->sigusr1_event_source);
+ if (r < 0)
+ return r;
- zero(ev);
- ev.events = EPOLLIN;
- ev.data.fd = s->signal_fd;
+ r = sd_event_add_signal(s->event, SIGUSR2, dispatch_sigusr2, s, &s->sigusr2_event_source);
+ 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, 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;
- int r;
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;
return r;
}
-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_get_now_monotonic(s->event, &when);
+ if (r < 0)
+ return r;
+
+ when += s->sync_interval_usec;
+
+ if (!s->sync_event_source) {
+ r = sd_event_add_monotonic(s->event, when, 0, server_dispatch_sync, s, &s->sync_event_source);
+ if (r < 0)
+ return r;
- timespec_store(&sync_timer_enable.it_value, s->sync_interval_usec);
+ 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 = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_enable, NULL);
+ 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;
}
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 = -1;
s->compress = true;
s->seal = true;
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));
if (r < 0)
return r;
- r = server_open_sync_timer(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;
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);
+ 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_unref(s->event);
if (s->syslog_fd >= 0)
close_nointr_nofail(s->syslog_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);
-
if (s->rate_limit)
journal_rate_limit_free(s->rate_limit);
~ianmdlvl / elogind.git / blobdiff