Copyright 2011 Lennart Poettering
systemd is free software; you can redistribute it and/or modify it
- under the terms of the GNU General Public License as published by
- the Free Software Foundation; either version 2 of the License, or
+ under the terms of the GNU Lesser General Public License as published by
+ the Free Software Foundation; either version 2.1 of the License, or
(at your option) any later version.
systemd is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- General Public License for more details.
+ Lesser General Public License for more details.
- You should have received a copy of the GNU General Public License
+ You should have received a copy of the GNU Lesser General Public License
along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/
#include <systemd/sd-messages.h>
#include <systemd/sd-daemon.h>
+#include "mkdir.h"
#include "hashmap.h"
#include "journal-file.h"
#include "socket-util.h"
#include "conf-parser.h"
#include "journald.h"
#include "virt.h"
+#include "missing.h"
#ifdef HAVE_ACL
#include <sys/acl.h>
#include "acl-util.h"
#endif
+#ifdef HAVE_SELINUX
+#include <selinux/selinux.h>
+#endif
+
#define USER_JOURNALS_MAX 1024
#define STDOUT_STREAMS_MAX 4096
#define RECHECK_VAR_AVAILABLE_USEC (30*USEC_PER_SEC)
-#define SYSLOG_TIMEOUT_USEC (250*USEC_PER_MSEC)
+#define N_IOVEC_META_FIELDS 17
-#define N_IOVEC_META_FIELDS 16
+#define ENTRY_SIZE_MAX (1024*1024*32)
typedef enum StdoutStreamState {
STDOUT_STREAM_IDENTIFIER,
int fd;
struct ucred ucred;
+#ifdef HAVE_SELINUX
+ security_context_t security_context;
+#endif
char *identifier;
int priority;
for (;;) {
struct stat st;
struct dirent buf, *de;
- int k;
- k = readdir_r(d, &buf, &de);
- if (k != 0) {
- r = -k;
- goto finish;
- }
+ r = readdir_r(d, &buf, &de);
+ if (r != 0)
+ break;
if (!de)
break;
- if (!dirent_is_file_with_suffix(de, ".journal"))
+ if (!endswith(de->d_name, ".journal") &&
+ !endswith(de->d_name, ".journal~"))
continue;
if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
continue;
- sum += (uint64_t) st.st_blocks * (uint64_t) st.st_blksize;
+ if (!S_ISREG(st.st_mode))
+ continue;
+
+ sum += (uint64_t) st.st_blocks * 512UL;
}
avail = sum >= m->max_use ? 0 : m->max_use - sum;
journal_file_close(f);
}
- r = journal_file_open(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
+ r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->system_journal, &f);
free(p);
if (r < 0)
return s->system_journal;
server_fix_perms(s, f, uid);
- f->metrics = s->system_metrics;
- f->compress = s->compress;
r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f);
if (r < 0) {
if (s->runtime_journal) {
r = journal_file_rotate(&s->runtime_journal);
if (r < 0)
- log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r));
+ 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);
if (r < 0)
- log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r));
+ 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);
}
HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) {
r = journal_file_rotate(&f);
if (r < 0)
- log_error("Failed to rotate %s: %s", f->path, strerror(-r));
- else
+ if (f->path)
+ log_error("Failed to rotate %s: %s", f->path, strerror(-r));
+ else
+ log_error("Failed to create user journal: %s", strerror(-r));
+ else {
hashmap_replace(s->user_journals, k, f);
+ server_fix_perms(s, s->system_journal, PTR_TO_UINT32(k));
+ }
}
}
return path;
}
-static void dispatch_message_real(Server *s,
- struct iovec *iovec, unsigned n, unsigned m,
- struct ucred *ucred,
- struct timeval *tv) {
+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) {
char *pid = NULL, *uid = NULL, *gid = NULL,
*source_time = NULL, *boot_id = NULL, *machine_id = NULL,
*comm = NULL, *cmdline = NULL, *hostname = NULL,
*audit_session = NULL, *audit_loginuid = NULL,
*exe = NULL, *cgroup = NULL, *session = NULL,
- *owner_uid = NULL, *unit = NULL;
+ *owner_uid = NULL, *unit = NULL, *selinux_context = NULL;
char idbuf[33];
sd_id128_t id;
exe = strappend("_EXE=", t);
free(t);
- if (comm)
+ if (exe)
IOVEC_SET_STRING(iovec[n++], exe);
}
if (sd_pid_get_owner_uid(ucred->uid, &owner) >= 0)
if (asprintf(&owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner) >= 0)
IOVEC_SET_STRING(iovec[n++], owner_uid);
+
+#ifdef HAVE_SELINUX
+ if (label) {
+ selinux_context = malloc(sizeof("_SELINUX_CONTEXT=") + label_len);
+ if (selinux_context) {
+ memcpy(selinux_context, "_SELINUX_CONTEXT=", sizeof("_SELINUX_CONTEXT=")-1);
+ memcpy(selinux_context+sizeof("_SELINUX_CONTEXT=")-1, label, label_len);
+ selinux_context[sizeof("_SELINUX_CONTEXT=")-1+label_len] = 0;
+ IOVEC_SET_STRING(iovec[n++], selinux_context);
+ }
+ } else {
+ security_context_t con;
+
+ if (getpidcon(ucred->pid, &con) >= 0) {
+ selinux_context = strappend("_SELINUX_CONTEXT=", con);
+ if (selinux_context)
+ IOVEC_SET_STRING(iovec[n++], selinux_context);
+
+ freecon(con);
+ }
+ }
+#endif
}
if (tv) {
else {
r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL);
- if (r == -E2BIG && !vacuumed) {
- log_info("Allocation limit reached.");
+ if ((r == -E2BIG || /* hit limit */
+ r == -EFBIG || /* hit fs limit */
+ r == -EDQUOT || /* quota hit */
+ r == -ENOSPC || /* disk full */
+ r == -EBADMSG || /* corrupted */
+ r == -ENODATA || /* truncated */
+ r == -EHOSTDOWN || /* other machine */
+ r == -EPROTONOSUPPORT) && /* unsupported feature */
+ !vacuumed) {
+
+ if (r == -E2BIG)
+ log_info("Allocation limit reached, rotating.");
+ else
+ log_warning("Journal file corrupted, rotating.");
server_rotate(s);
server_vacuum(s);
free(session);
free(owner_uid);
free(unit);
+ free(selinux_context);
}
static void driver_message(Server *s, sd_id128_t message_id, const char *format, ...) {
ucred.uid = getuid();
ucred.gid = getgid();
- dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL);
+ dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0);
}
static void dispatch_message(Server *s,
struct iovec *iovec, unsigned n, unsigned m,
struct ucred *ucred,
struct timeval *tv,
+ const char *label, size_t label_len,
int priority) {
int rl;
char *path = NULL, *c;
free(path);
finish:
- dispatch_message_real(s, iovec, n, m, ucred, tv);
+ dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len);
}
static void forward_syslog_iovec(Server *s, const struct iovec *iovec, unsigned n_iovec, struct ucred *ucred, struct timeval *tv) {
if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
return;
+ /* The socket is full? I guess the syslog implementation is
+ * too slow, and we shouldn't wait for that... */
+ if (errno == EAGAIN)
+ return;
+
if (ucred && errno == ESRCH) {
struct ucred u;
if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0)
return;
+
+ if (errno == EAGAIN)
+ return;
}
log_debug("Failed to forward syslog message: %m");
*buf += strspn(*buf, WHITESPACE);
}
-static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) {
+static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv, const char *label, size_t label_len) {
char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL, *syslog_identifier = NULL, *syslog_pid = NULL;
struct iovec iovec[N_IOVEC_META_FIELDS + 6];
unsigned n = 0;
if (message)
IOVEC_SET_STRING(iovec[n++], message);
- dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, priority);
+ dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, label, label_len, priority);
free(message);
free(identifier);
return true;
}
-static void process_native_message(Server *s, const void *buffer, size_t buffer_size, struct ucred *ucred, struct timeval *tv) {
+static void process_native_message(
+ Server *s,
+ const void *buffer, size_t buffer_size,
+ struct ucred *ucred,
+ struct timeval *tv,
+ const char *label, size_t label_len) {
+
struct iovec *iovec = NULL;
unsigned n = 0, m = 0, j, tn = (unsigned) -1;
const char *p;
char *identifier = NULL, *message = NULL;
assert(s);
- assert(buffer || n == 0);
+ assert(buffer || buffer_size == 0);
p = buffer;
remaining = buffer_size;
if (e == p) {
/* Entry separator */
- dispatch_message(s, iovec, n, m, ucred, tv, priority);
+ dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, priority);
n = 0;
priority = LOG_INFO;
p[17] >= '0' && p[17] <= '9')
priority = (priority & LOG_PRIMASK) | (((p[16] - '0')*10 + (p[17] - '0')) << 3);
- else if (l >= 12 &&
- memcmp(p, "SYSLOG_IDENTIFIER=", 11) == 0) {
+ else if (l >= 19 &&
+ memcmp(p, "SYSLOG_IDENTIFIER=", 18) == 0) {
char *t;
- t = strndup(p + 11, l - 11);
+ t = strndup(p + 18, l - 18);
if (t) {
free(identifier);
identifier = t;
p = e + 1;
continue;
} else {
+ le64_t l_le;
uint64_t l;
char *k;
break;
}
- memcpy(&l, e + 1, sizeof(uint64_t));
- l = le64toh(l);
+ memcpy(&l_le, e + 1, sizeof(uint64_t));
+ l = le64toh(l_le);
if (remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
e[1+sizeof(uint64_t)+l] != '\n') {
forward_console(s, identifier, message, ucred);
}
- dispatch_message(s, iovec, n, m, ucred, tv, priority);
+ dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, priority);
finish:
for (j = 0; j < n; j++) {
free(iovec[j].iov_base);
}
+ free(iovec);
free(identifier);
free(message);
}
+static void process_native_file(
+ Server *s,
+ int fd,
+ struct ucred *ucred,
+ struct timeval *tv,
+ const char *label, size_t label_len) {
+
+ struct stat st;
+ void *p;
+ ssize_t n;
+
+ assert(s);
+ assert(fd >= 0);
+
+ /* Data is in the passed file, since it didn't fit in a
+ * datagram. We can't map the file here, since clients might
+ * then truncate it and trigger a SIGBUS for us. So let's
+ * stupidly read it */
+
+ if (fstat(fd, &st) < 0) {
+ log_error("Failed to stat passed file, ignoring: %m");
+ return;
+ }
+
+ if (!S_ISREG(st.st_mode)) {
+ log_error("File passed is not regular. Ignoring.");
+ return;
+ }
+
+ if (st.st_size <= 0)
+ return;
+
+ if (st.st_size > ENTRY_SIZE_MAX) {
+ log_error("File passed too large. Ignoring.");
+ return;
+ }
+
+ p = malloc(st.st_size);
+ if (!p) {
+ log_error("Out of memory");
+ return;
+ }
+
+ n = pread(fd, p, st.st_size, 0);
+ if (n < 0)
+ log_error("Failed to read file, ignoring: %s", strerror(-n));
+ else if (n > 0)
+ process_native_message(s, p, n, ucred, tv, label, label_len);
+
+ free(p);
+}
+
static int stdout_stream_log(StdoutStream *s, const char *p) {
struct iovec iovec[N_IOVEC_META_FIELDS + 5];
char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL, *syslog_identifier = NULL;
unsigned n = 0;
int priority;
+ char *label = NULL;
+ size_t label_len = 0;
assert(s);
assert(p);
+ if (isempty(p))
+ return 0;
+
priority = s->priority;
if (s->level_prefix)
if (message)
IOVEC_SET_STRING(iovec[n++], message);
- dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, priority);
+#ifdef HAVE_SELINUX
+ if (s->security_context) {
+ label = (char*) s->security_context;
+ label_len = strlen((char*) s->security_context);
+ }
+#endif
+
+ dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, label, label_len, priority);
free(message);
free(syslog_priority);
switch (s->state) {
case STDOUT_STREAM_IDENTIFIER:
- s->identifier = strdup(p);
- if (!s->identifier) {
- log_error("Out of memory");
- return -ENOMEM;
+ if (isempty(p))
+ s->identifier = NULL;
+ else {
+ s->identifier = strdup(p);
+ if (!s->identifier) {
+ log_error("Out of memory");
+ return -ENOMEM;
+ }
}
s->state = STDOUT_STREAM_PRIORITY;
close_nointr_nofail(s->fd);
}
+#ifdef HAVE_SELINUX
+ if (s->security_context)
+ freecon(s->security_context);
+#endif
+
free(s->identifier);
free(s);
}
goto fail;
}
+#ifdef HAVE_SELINUX
+ if (getpeercon(fd, &stream->security_context) < 0)
+ log_error("Failed to determine peer security context.");
+#endif
+
if (shutdown(fd, SHUT_WR) < 0) {
log_error("Failed to shutdown writing side of socket: %m");
r = -errno;
assert(s);
assert(p);
+ if (isempty(p))
+ return;
+
parse_syslog_priority((char **) &p, &priority);
if (s->forward_to_kmsg && (priority & LOG_FACMASK) != LOG_KERN)
if (message)
IOVEC_SET_STRING(iovec[n++], message);
-
- dispatch_message(s, iovec, n, ELEMENTSOF(iovec), NULL, NULL, priority);
+ dispatch_message(s, iovec, n, ELEMENTSOF(iovec), NULL, NULL, NULL, 0, priority);
free(message);
free(syslog_priority);
if (!fn)
return -ENOMEM;
- r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
+ r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal);
free(fn);
if (r >= 0) {
* it if necessary. */
(void) mkdir_parents(fn, 0755);
- r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
+ r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal);
free(fn);
if (r < 0) {
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(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 (v <= 0)
- return 1;
-
if (s->buffer_size < (size_t) v) {
void *b;
size_t l;
msghdr.msg_control = &control;
msghdr.msg_controllen = sizeof(control);
- n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT);
+ n = recvmsg(ev->data.fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC);
if (n < 0) {
if (errno == EINTR || errno == EAGAIN)
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 (ev->data.fd == s->syslog_fd) {
char *e;
- e = memchr(s->buffer, '\n', n);
- if (e)
- *e = 0;
- else
- s->buffer[n] = 0;
+ if (n > 0 && n_fds == 0) {
+ e = memchr(s->buffer, '\n', n);
+ if (e)
+ *e = 0;
+ else
+ s->buffer[n] = 0;
+
+ 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)
+ process_native_message(s, s->buffer, n, ucred, tv, label, label_len);
+ else if (n == 0 && n_fds == 1)
+ 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.");
+ }
- process_syslog_message(s, strstrip(s->buffer), ucred, tv);
- } else
- process_native_message(s, s->buffer, n, ucred, tv);
+ close_many(fds, n_fds);
}
return 1;
union sockaddr_union sa;
int one, r;
struct epoll_event ev;
- struct timeval tv;
assert(s);
}
chmod(sa.un.sun_path, 0666);
- }
+ } else
+ fd_nonblock(s->syslog_fd, 1);
one = 1;
r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
return -errno;
}
+#ifdef HAVE_SELINUX
+ one = 1;
+ r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
+ if (r < 0)
+ log_warning("SO_PASSSEC failed: %m");
+#endif
+
one = 1;
r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
if (r < 0) {
return -errno;
}
- /* Since we use the same socket for forwarding this to some
- * other syslog implementation, make sure we don't hang
- * forever */
- timeval_store(&tv, SYSLOG_TIMEOUT_USEC);
- if (setsockopt(s->syslog_fd, SOL_SOCKET, SO_SNDTIMEO, &tv, sizeof(tv)) < 0) {
- log_error("SO_SNDTIMEO failed: %m");
- return -errno;
- }
-
zero(ev);
ev.events = EPOLLIN;
ev.data.fd = s->syslog_fd;
}
chmod(sa.un.sun_path, 0666);
- }
+ } else
+ fd_nonblock(s->native_fd, 1);
one = 1;
r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
return -errno;
}
+#ifdef HAVE_SELINUX
+ one = 1;
+ r = setsockopt(s->syslog_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
+ if (r < 0)
+ log_warning("SO_PASSSEC failed: %m");
+#endif
+
one = 1;
r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
if (r < 0) {
log_error("liste() failed: %m");
return -errno;
}
- }
+ } else
+ fd_nonblock(s->stdout_fd, 1);
zero(ev);
ev.events = EPOLLIN;
if (!s->import_proc_kmsg)
return 0;
-
- s->proc_kmsg_fd = open("/proc/kmsg", O_CLOEXEC|O_NONBLOCK|O_NOCTTY);
+ s->proc_kmsg_fd = open("/proc/kmsg", O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_NOCTTY);
if (s->proc_kmsg_fd < 0) {
log_warning("Failed to open /proc/kmsg, ignoring: %m");
return 0;
assert(s);
- fn = "/etc/systemd/systemd-journald.conf";
+ fn = "/etc/systemd/journald.conf";
f = fopen(fn, "re");
if (!f) {
if (errno == ENOENT)
if (r < 0)
return r;
- r = system_journal_open(s);
- if (r < 0)
- return r;
-
r = open_signalfd(s);
if (r < 0)
return r;
if (!s->rate_limit)
return -ENOMEM;
+ r = system_journal_open(s);
+ if (r < 0)
+ return r;
+
return 0;
}
return EXIT_FAILURE;
}
- log_set_target(LOG_TARGET_CONSOLE);
+ log_set_target(LOG_TARGET_SAFE);
+ log_set_facility(LOG_SYSLOG);
log_parse_environment();
log_open();
~ianmdlvl / elogind.git / blobdiff