X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournald.c;h=c29d75c9c4045a22af163f3fbe5a689d87eb30f2;hp=e7231d96f78c4dfb6404a7e011179b7b243b2bc1;hb=32fca2d5b746b86bbbbd4f1b57064076cfc2ae4b;hpb=de97b26ac5e29063632312ec1a20eb6318ca924c diff --git a/src/journal/journald.c b/src/journal/journald.c index e7231d96f..c29d75c9c 100644 --- a/src/journal/journald.c +++ b/src/journal/journald.c @@ -6,16 +6,16 @@ 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 . ***/ @@ -25,21 +25,41 @@ #include #include #include -#include -#include #include #include #include #include +#include +#include +#include + +#ifdef HAVE_LOGIND +#include +#endif + +#include "mkdir.h" #include "hashmap.h" #include "journal-file.h" -#include "sd-daemon.h" #include "socket-util.h" -#include "acl-util.h" #include "cgroup-util.h" #include "list.h" #include "journal-rate-limit.h" +#include "journal-internal.h" +#include "conf-parser.h" +#include "journald.h" +#include "virt.h" +#include "missing.h" + +#ifdef HAVE_ACL +#include +#include +#include "acl-util.h" +#endif + +#ifdef HAVE_SELINUX +#include +#endif #define USER_JOURNALS_MAX 1024 #define STDOUT_STREAMS_MAX 4096 @@ -49,42 +69,18 @@ #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC) -typedef struct StdoutStream StdoutStream; - -typedef struct Server { - int epoll_fd; - int signal_fd; - int syslog_fd; - int native_fd; - int stdout_fd; - - JournalFile *runtime_journal; - JournalFile *system_journal; - Hashmap *user_journals; - - uint64_t seqnum; - - char *buffer; - size_t buffer_size; - - JournalRateLimit *rate_limit; - - JournalMetrics metrics; - uint64_t max_use; - bool compress; +#define N_IOVEC_META_FIELDS 17 - uint64_t cached_available_space; - usec_t cached_available_space_timestamp; - - LIST_HEAD(StdoutStream, stdout_streams); - unsigned n_stdout_streams; -} Server; +#define ENTRY_SIZE_MAX (1024*1024*32) typedef enum StdoutStreamState { - STDOUT_STREAM_TAG, + STDOUT_STREAM_IDENTIFIER, + STDOUT_STREAM_UNIT_ID, STDOUT_STREAM_PRIORITY, - STDOUT_STREAM_PRIORITY_PREFIX, - STDOUT_STREAM_TEE_CONSOLE, + STDOUT_STREAM_LEVEL_PREFIX, + STDOUT_STREAM_FORWARD_TO_SYSLOG, + STDOUT_STREAM_FORWARD_TO_KMSG, + STDOUT_STREAM_FORWARD_TO_CONSOLE, STDOUT_STREAM_RUNNING } StdoutStreamState; @@ -95,11 +91,17 @@ struct StdoutStream { int fd; struct ucred ucred; +#ifdef HAVE_SELINUX + security_context_t security_context; +#endif - char *tag; + char *identifier; + char *unit_id; int priority; - bool priority_prefix:1; - bool tee_console:1; + bool level_prefix:1; + bool forward_to_syslog:1; + bool forward_to_kmsg:1; + bool forward_to_console:1; char buffer[LINE_MAX+1]; size_t length; @@ -107,16 +109,28 @@ struct StdoutStream { LIST_FIELDS(StdoutStream, stdout_stream); }; +static const char* const storage_table[] = { + [STORAGE_AUTO] = "auto", + [STORAGE_VOLATILE] = "volatile", + [STORAGE_PERSISTENT] = "persistent", + [STORAGE_NONE] = "none" +}; + +DEFINE_STRING_TABLE_LOOKUP(storage, Storage); +DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting"); + static uint64_t available_space(Server *s) { - char ids[33]; - sd_id128_t machine; - char *p; + char ids[33], *p; const char *f; + sd_id128_t machine; struct statvfs ss; uint64_t sum = 0, avail = 0, ss_avail = 0; int r; DIR *d; - usec_t ts = now(CLOCK_MONOTONIC); + usec_t ts; + JournalMetrics *m; + + ts = now(CLOCK_MONOTONIC); if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts) return s->cached_available_space; @@ -125,10 +139,15 @@ static uint64_t available_space(Server *s) { if (r < 0) return 0; - if (s->system_journal) + if (s->system_journal) { f = "/var/log/journal/"; - else + m = &s->system_metrics; + } else { f = "/run/log/journal/"; + m = &s->runtime_metrics; + } + + assert(m); p = strappend(f, sd_id128_to_string(machine, ids)); if (!p) @@ -146,31 +165,32 @@ static uint64_t available_space(Server *s) { 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 >= s->max_use ? 0 : s->max_use - sum; + avail = sum >= m->max_use ? 0 : m->max_use - sum; ss_avail = ss.f_bsize * ss.f_bavail; - ss_avail = ss_avail < s->metrics.keep_free ? 0 : ss_avail - s->metrics.keep_free; + ss_avail = ss_avail < m->keep_free ? 0 : ss_avail - m->keep_free; if (ss_avail < avail) avail = ss_avail; @@ -184,18 +204,42 @@ finish: return avail; } -static void fix_perms(JournalFile *f, uid_t uid) { +static void server_read_file_gid(Server *s) { + const char *adm = "adm"; + int r; + + assert(s); + + if (s->file_gid_valid) + return; + + r = get_group_creds(&adm, &s->file_gid); + if (r < 0) + log_warning("Failed to resolve 'adm' group: %s", 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; +} + +static void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { + int r; +#ifdef HAVE_ACL acl_t acl; acl_entry_t entry; acl_permset_t permset; - int r; +#endif assert(f); - r = fchmod_and_fchown(f->fd, 0640, 0, 0); + server_read_file_gid(s); + + r = fchmod_and_fchown(f->fd, 0640, 0, s->file_gid); if (r < 0) log_warning("Failed to fix access mode/rights on %s, ignoring: %s", f->path, strerror(-r)); +#ifdef HAVE_ACL if (uid <= 0) return; @@ -228,6 +272,7 @@ static void fix_perms(JournalFile *f, uid_t uid) { finish: acl_free(acl); +#endif } static JournalFile* find_journal(Server *s, uid_t uid) { @@ -239,8 +284,12 @@ static JournalFile* find_journal(Server *s, uid_t uid) { assert(s); - /* We split up user logs only on /var, not on /run */ - if (!s->system_journal) + /* We split up user logs only on /var, not on /run. If the + * runtime file is open, we write to it exclusively, in order + * to guarantee proper order as soon as we flush /run to + * /var and close the runtime file. */ + + if (s->runtime_journal) return s->runtime_journal; if (uid <= 0) @@ -264,15 +313,13 @@ static JournalFile* find_journal(Server *s, uid_t uid) { 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_metrics, s->system_journal, &f); free(p); if (r < 0) return s->system_journal; - fix_perms(f, uid); - f->metrics = s->metrics; - f->compress = s->compress; + server_fix_perms(s, f, uid); r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f); if (r < 0) { @@ -283,36 +330,56 @@ static JournalFile* find_journal(Server *s, uid_t uid) { return f; } -static void server_vacuum(Server *s) { - Iterator i; +static void server_rotate(Server *s) { + JournalFile *f; void *k; - char *p; - char ids[33]; - sd_id128_t machine; + Iterator i; int r; - JournalFile *f; log_info("Rotating..."); 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)); + } } +} + +static void server_vacuum(Server *s) { + char *p; + char ids[33]; + sd_id128_t machine; + int r; log_info("Vacuuming..."); @@ -322,25 +389,31 @@ static void server_vacuum(Server *s) { return; } - if (asprintf(&p, "/var/log/journal/%s", sd_id128_to_string(machine, ids)) < 0) { - log_error("Out of memory."); - return; - } + sd_id128_to_string(machine, ids); - r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free); - if (r < 0 && r != -ENOENT) - log_error("Failed to vacuum %s: %s", p, strerror(-r)); - free(p); + if (s->system_journal) { + if (asprintf(&p, "/var/log/journal/%s", ids) < 0) { + log_error("Out of memory."); + return; + } - if (asprintf(&p, "/run/log/journal/%s", ids) < 0) { - log_error("Out of memory."); - return; + r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free); + if (r < 0 && r != -ENOENT) + log_error("Failed to vacuum %s: %s", p, strerror(-r)); + free(p); } - r = journal_directory_vacuum(p, s->max_use, s->metrics.keep_free); - if (r < 0 && r != -ENOENT) - log_error("Failed to vacuum %s: %s", p, strerror(-r)); - free(p); + if (s->runtime_journal) { + if (asprintf(&p, "/run/log/journal/%s", ids) < 0) { + log_error("Out of memory."); + return; + } + + r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free); + if (r < 0 && r != -ENOENT) + log_error("Failed to vacuum %s: %s", p, strerror(-r)); + free(p); + } s->cached_available_space_timestamp = 0; } @@ -361,46 +434,124 @@ static char *shortened_cgroup_path(pid_t pid) { return NULL; } - if (streq(init_path, "/")) + if (endswith(init_path, "/system")) + init_path[strlen(init_path) - 7] = 0; + else if (streq(init_path, "/")) init_path[0] = 0; - if (startswith(process_path, init_path)) - path = process_path + strlen(init_path); - else + if (startswith(process_path, init_path)) { + char *p; + + p = strdup(process_path + strlen(init_path)); + if (!p) { + free(process_path); + free(init_path); + return NULL; + } + path = p; + } else { path = process_path; + process_path = NULL; + } + free(process_path); free(init_path); return path; } -static void dispatch_message_real(Server *s, - struct iovec *iovec, unsigned n, unsigned m, - struct ucred *ucred, - struct timeval *tv) { +static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n) { + 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)) { + log_info("Journal header limits reached or header out-of-date, rotating."); + server_rotate(s); + server_vacuum(s); + vacuumed = true; + } + + for (;;) { + r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL); + if (r >= 0) + return; + + if (vacuumed || + (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 */ + r != -EBUSY && /* unclean shutdown */ + r != -ESHUTDOWN /* already archived */)) { + log_error("Failed to write entry, ignoring: %s", strerror(-r)); + return; + } + + if (r == -E2BIG || r == -EFBIG || r == EDQUOT || r == ENOSPC) + log_info("Allocation limit reached, rotating."); + else if (r == -EHOSTDOWN) + log_info("Journal file from other machine, rotating."); + else if (r == -EBUSY) + log_info("Unlcean shutdown, rotating."); + else + log_warning("Journal file corrupted, rotating."); + + server_rotate(s); + server_vacuum(s); + vacuumed = true; + + f = find_journal(s, uid); + if (!f) + return; + + log_info("Retrying write."); + } +} + +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) { 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; + *exe = NULL, *cgroup = NULL, *session = NULL, + *owner_uid = NULL, *unit = NULL, *selinux_context = NULL; char idbuf[33]; sd_id128_t id; int r; char *t; uid_t loginuid = 0, realuid = 0; - JournalFile *f; - bool vacuumed = false; assert(s); assert(iovec); assert(n > 0); - assert(n + 13 <= m); + assert(n + N_IOVEC_META_FIELDS <= m); if (ucred) { - uint32_t session; - char *path; + uint32_t audit; +#ifdef HAVE_LOGIND + uid_t owner; +#endif realuid = ucred->uid; @@ -416,30 +567,33 @@ static void dispatch_message_real(Server *s, r = get_process_comm(ucred->pid, &t); if (r >= 0) { comm = strappend("_COMM=", t); + free(t); + if (comm) IOVEC_SET_STRING(iovec[n++], comm); - free(t); } r = get_process_exe(ucred->pid, &t); if (r >= 0) { exe = strappend("_EXE=", t); - if (comm) - IOVEC_SET_STRING(iovec[n++], exe); free(t); + + if (exe) + IOVEC_SET_STRING(iovec[n++], exe); } r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t); if (r >= 0) { cmdline = strappend("_CMDLINE=", t); + free(t); + if (cmdline) IOVEC_SET_STRING(iovec[n++], cmdline); - free(t); } - r = audit_session_from_pid(ucred->pid, &session); + r = audit_session_from_pid(ucred->pid, &audit); if (r >= 0) - if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) session) >= 0) + if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit) >= 0) IOVEC_SET_STRING(iovec[n++], audit_session); r = audit_loginuid_from_pid(ucred->pid, &loginuid); @@ -447,14 +601,59 @@ static void dispatch_message_real(Server *s, if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0) IOVEC_SET_STRING(iovec[n++], audit_loginuid); - path = shortened_cgroup_path(ucred->pid); - if (path) { - cgroup = strappend("_SYSTEMD_CGROUP=", path); + t = shortened_cgroup_path(ucred->pid); + if (t) { + cgroup = strappend("_SYSTEMD_CGROUP=", t); + free(t); + if (cgroup) IOVEC_SET_STRING(iovec[n++], cgroup); + } + +#ifdef HAVE_LOGIND + if (sd_pid_get_session(ucred->pid, &t) >= 0) { + session = strappend("_SYSTEMD_SESSION=", t); + free(t); + + if (session) + IOVEC_SET_STRING(iovec[n++], session); + } + + 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); +#endif - free(path); + if (cg_pid_get_unit(ucred->pid, &t) >= 0) { + unit = strappend("_SYSTEMD_UNIT=", t); + free(t); + } else if (unit_id) + unit = strappend("_SYSTEMD_UNIT=", unit_id); + + if (unit) + IOVEC_SET_STRING(iovec[n++], unit); + +#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) { @@ -479,33 +678,14 @@ static void dispatch_message_real(Server *s, t = gethostname_malloc(); if (t) { hostname = strappend("_HOSTNAME=", t); + free(t); if (hostname) IOVEC_SET_STRING(iovec[n++], hostname); - free(t); } assert(n <= m); -retry: - f = find_journal(s, realuid == 0 ? 0 : loginuid); - if (!f) - log_warning("Dropping message, as we can't find a place to store the data."); - else { - r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL); - - if (r == -E2BIG && !vacuumed) { - log_info("Allocation limit reached."); - - server_vacuum(s); - vacuumed = true; - - log_info("Retrying write."); - goto retry; - } - - if (r < 0) - log_error("Failed to write entry, ignoring: %s", strerror(-r)); - } + write_to_journal(s, realuid == 0 ? 0 : loginuid, iovec, n); free(pid); free(uid); @@ -520,15 +700,54 @@ retry: free(audit_session); free(audit_loginuid); free(cgroup); + free(session); + free(owner_uid); + free(unit); + free(selinux_context); +} + +static void 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=5"); + 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); + + snprintf(mid, sizeof(mid), "MESSAGE_ID=" SD_ID128_FORMAT_STR, SD_ID128_FORMAT_VAL(message_id)); + char_array_0(mid); + IOVEC_SET_STRING(iovec[n++], mid); + + zero(ucred); + ucred.pid = getpid(); + ucred.uid = getuid(); + ucred.gid = getgid(); + + dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL); } 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, + const char *unit_id, int priority) { int rl; - char *path, *c; + char *path = NULL, *c; assert(s); assert(iovec || n == 0); @@ -536,6 +755,9 @@ static void dispatch_message(Server *s, if (n == 0) return; + if (LOG_PRI(priority) > s->max_level_store) + return; + if (!ucred) goto finish; @@ -559,149 +781,488 @@ static void dispatch_message(Server *s, } } - rl = journal_rate_limit_test(s->rate_limit, path, priority, available_space(s)); + rl = journal_rate_limit_test(s->rate_limit, path, priority & LOG_PRIMASK, available_space(s)); if (rl == 0) { free(path); return; } - if (rl > 1) { - int j = 0; - char suppress_message[LINE_MAX]; - struct iovec suppress_iovec[15]; - - /* Write a suppression message if we suppressed something */ - - snprintf(suppress_message, sizeof(suppress_message), "MESSAGE=Suppressed %u messages from %s", rl - 1, path); - char_array_0(suppress_message); - - IOVEC_SET_STRING(suppress_iovec[j++], "PRIORITY=5"); - IOVEC_SET_STRING(suppress_iovec[j++], suppress_message); - - dispatch_message_real(s, suppress_iovec, j, ELEMENTSOF(suppress_iovec), NULL, NULL); - } + /* Write a suppression message if we suppressed something */ + if (rl > 1) + driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, "Suppressed %u messages from %s", rl - 1, path); free(path); finish: - dispatch_message_real(s, iovec, n, m, ucred, tv); + dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id); } -static void process_syslog_message(Server *s, const char *buf, struct ucred *ucred, struct timeval *tv) { - char *message = NULL, *syslog_priority = NULL, *syslog_facility = NULL; - struct iovec iovec[16]; - unsigned n = 0; - int priority = LOG_USER | LOG_INFO; +static void forward_syslog_iovec(Server *s, const struct iovec *iovec, unsigned n_iovec, struct ucred *ucred, struct timeval *tv) { + struct msghdr msghdr; + struct cmsghdr *cmsg; + union { + struct cmsghdr cmsghdr; + uint8_t buf[CMSG_SPACE(sizeof(struct ucred))]; + } control; + union sockaddr_union sa; assert(s); - assert(buf); + assert(iovec); + assert(n_iovec > 0); - parse_syslog_priority((char**) &buf, &priority); - skip_syslog_date((char**) &buf); + zero(msghdr); + msghdr.msg_iov = (struct iovec*) iovec; + msghdr.msg_iovlen = n_iovec; - if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0) - IOVEC_SET_STRING(iovec[n++], syslog_priority); + zero(sa); + sa.un.sun_family = AF_UNIX; + strncpy(sa.un.sun_path, "/run/systemd/journal/syslog", sizeof(sa.un.sun_path)); + msghdr.msg_name = &sa; + msghdr.msg_namelen = offsetof(union sockaddr_union, un.sun_path) + strlen(sa.un.sun_path); - if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0) - IOVEC_SET_STRING(iovec[n++], syslog_facility); + if (ucred) { + zero(control); + msghdr.msg_control = &control; + msghdr.msg_controllen = sizeof(control); + + cmsg = CMSG_FIRSTHDR(&msghdr); + cmsg->cmsg_level = SOL_SOCKET; + cmsg->cmsg_type = SCM_CREDENTIALS; + cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred)); + memcpy(CMSG_DATA(cmsg), ucred, sizeof(struct ucred)); + msghdr.msg_controllen = cmsg->cmsg_len; + } - message = strappend("MESSAGE=", buf); - if (message) - IOVEC_SET_STRING(iovec[n++], message); + /* Forward the syslog message we received via /dev/log to + * /run/systemd/syslog. Unfortunately we currently can't set + * the SO_TIMESTAMP auxiliary data, and hence we don't. */ - dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, priority & LOG_PRIMASK); + if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0) + return; - free(message); - free(syslog_facility); - free(syslog_priority); -} + /* The socket is full? I guess the syslog implementation is + * too slow, and we shouldn't wait for that... */ + if (errno == EAGAIN) + return; -static bool valid_user_field(const char *p, size_t l) { - const char *a; + if (ucred && errno == ESRCH) { + struct ucred u; - /* We kinda enforce POSIX syntax recommendations for - environment variables here, but make a couple of additional - requirements. + /* Hmm, presumably the sender process vanished + * by now, so let's fix it as good as we + * can, and retry */ - http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */ + u = *ucred; + u.pid = getpid(); + memcpy(CMSG_DATA(cmsg), &u, sizeof(struct ucred)); - /* No empty field names */ - if (l <= 0) - return false; + if (sendmsg(s->syslog_fd, &msghdr, MSG_NOSIGNAL) >= 0) + return; - /* Don't allow names longer than 64 chars */ - if (l > 64) - return false; + if (errno == EAGAIN) + return; + } - /* Variables starting with an underscore are protected */ - if (p[0] == '_') - return false; + log_debug("Failed to forward syslog message: %m"); +} - /* Don't allow digits as first character */ - if (p[0] >= '0' && p[0] <= '9') - return false; +static void forward_syslog_raw(Server *s, int priority, const char *buffer, struct ucred *ucred, struct timeval *tv) { + struct iovec iovec; - /* Only allow A-Z0-9 and '_' */ - for (a = p; a < p + l; a++) - if (!((*a >= 'A' && *a <= 'Z') || - (*a >= '0' && *a <= '9') || - *a == '_')) - return false; + assert(s); + assert(buffer); - return true; + if (LOG_PRI(priority) > s->max_level_syslog) + return; + + IOVEC_SET_STRING(iovec, buffer); + forward_syslog_iovec(s, &iovec, 1, ucred, tv); } -static void process_native_message(Server *s, const void *buffer, size_t buffer_size, struct ucred *ucred, struct timeval *tv) { - struct iovec *iovec = NULL; - unsigned n = 0, m = 0, j; - const char *p; - size_t remaining; - int priority = LOG_INFO; +static void forward_syslog(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred, struct timeval *tv) { + struct iovec iovec[5]; + char header_priority[6], header_time[64], header_pid[16]; + int n = 0; + time_t t; + struct tm *tm; + char *ident_buf = NULL; assert(s); - assert(buffer || n == 0); + assert(priority >= 0); + assert(priority <= 999); + assert(message); - p = buffer; - remaining = buffer_size; + if (LOG_PRI(priority) > s->max_level_syslog) + return; - while (remaining > 0) { - const char *e, *q; + /* First: priority field */ + snprintf(header_priority, sizeof(header_priority), "<%i>", priority); + char_array_0(header_priority); + IOVEC_SET_STRING(iovec[n++], header_priority); - e = memchr(p, '\n', remaining); + /* Second: timestamp */ + t = tv ? tv->tv_sec : ((time_t) (now(CLOCK_REALTIME) / USEC_PER_SEC)); + tm = localtime(&t); + if (!tm) + return; + if (strftime(header_time, sizeof(header_time), "%h %e %T ", tm) <= 0) + return; + IOVEC_SET_STRING(iovec[n++], header_time); - if (!e) { - /* Trailing noise, let's ignore it, and flush what we collected */ - log_debug("Received message with trailing noise, ignoring."); - break; + /* Third: identifier and PID */ + if (ucred) { + if (!identifier) { + get_process_comm(ucred->pid, &ident_buf); + identifier = ident_buf; } - if (e == p) { - /* Entry separator */ - dispatch_message(s, iovec, n, m, ucred, tv, priority); - n = 0; - priority = LOG_INFO; + snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid); + char_array_0(header_pid); - p++; - remaining--; - continue; - } + if (identifier) + IOVEC_SET_STRING(iovec[n++], identifier); - if (*p == '.' || *p == '#') { - /* Ignore control commands for now, and - * comments too. */ - remaining -= (e - p) + 1; - p = e + 1; - continue; - } + IOVEC_SET_STRING(iovec[n++], header_pid); + } else if (identifier) { + IOVEC_SET_STRING(iovec[n++], identifier); + IOVEC_SET_STRING(iovec[n++], ": "); + } + + /* Fourth: message */ + IOVEC_SET_STRING(iovec[n++], message); + + forward_syslog_iovec(s, iovec, n, ucred, tv); + + free(ident_buf); +} + +static int fixup_priority(int priority) { + + if ((priority & LOG_FACMASK) == 0) + return (priority & LOG_PRIMASK) | LOG_USER; + + return priority; +} + +static void forward_kmsg(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred) { + struct iovec iovec[5]; + char header_priority[6], header_pid[16]; + int n = 0; + char *ident_buf = NULL; + int fd; + + assert(s); + assert(priority >= 0); + assert(priority <= 999); + assert(message); + + if (LOG_PRI(priority) > s->max_level_kmsg) + return; + + /* Never allow messages with kernel facility to be written to + * kmsg, regardless where the data comes from. */ + priority = fixup_priority(priority); + + /* First: priority field */ + snprintf(header_priority, sizeof(header_priority), "<%i>", priority); + char_array_0(header_priority); + IOVEC_SET_STRING(iovec[n++], header_priority); + + /* Second: identifier and PID */ + if (ucred) { + if (!identifier) { + get_process_comm(ucred->pid, &ident_buf); + identifier = ident_buf; + } + + snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid); + char_array_0(header_pid); + + if (identifier) + IOVEC_SET_STRING(iovec[n++], identifier); + + IOVEC_SET_STRING(iovec[n++], header_pid); + } else if (identifier) { + IOVEC_SET_STRING(iovec[n++], identifier); + IOVEC_SET_STRING(iovec[n++], ": "); + } + + /* Fourth: message */ + IOVEC_SET_STRING(iovec[n++], message); + IOVEC_SET_STRING(iovec[n++], "\n"); + + fd = open("/dev/kmsg", O_WRONLY|O_NOCTTY|O_CLOEXEC); + if (fd < 0) { + log_debug("Failed to open /dev/kmsg for logging: %s", strerror(errno)); + goto finish; + } + + if (writev(fd, iovec, n) < 0) + log_debug("Failed to write to /dev/kmsg for logging: %s", strerror(errno)); + + close_nointr_nofail(fd); + +finish: + free(ident_buf); +} + +static void forward_console(Server *s, int priority, const char *identifier, const char *message, struct ucred *ucred) { + struct iovec iovec[4]; + char header_pid[16]; + int n = 0, fd; + char *ident_buf = NULL; + const char *tty; + + assert(s); + assert(message); + + if (LOG_PRI(priority) > s->max_level_console) + return; + + /* First: identifier and PID */ + if (ucred) { + if (!identifier) { + get_process_comm(ucred->pid, &ident_buf); + identifier = ident_buf; + } + + snprintf(header_pid, sizeof(header_pid), "[%lu]: ", (unsigned long) ucred->pid); + char_array_0(header_pid); + + if (identifier) + IOVEC_SET_STRING(iovec[n++], identifier); + + IOVEC_SET_STRING(iovec[n++], header_pid); + } else if (identifier) { + IOVEC_SET_STRING(iovec[n++], identifier); + IOVEC_SET_STRING(iovec[n++], ": "); + } + + /* Third: message */ + IOVEC_SET_STRING(iovec[n++], message); + IOVEC_SET_STRING(iovec[n++], "\n"); + + tty = s->tty_path ? s->tty_path : "/dev/console"; + + fd = open_terminal(tty, O_WRONLY|O_NOCTTY|O_CLOEXEC); + if (fd < 0) { + log_debug("Failed to open %s for logging: %s", tty, strerror(errno)); + goto finish; + } + + if (writev(fd, iovec, n) < 0) + log_debug("Failed to write to %s for logging: %s", tty, strerror(errno)); + + close_nointr_nofail(fd); + +finish: + free(ident_buf); +} + +static void read_identifier(const char **buf, char **identifier, char **pid) { + const char *p; + char *t; + size_t l, e; + + assert(buf); + assert(identifier); + assert(pid); + + p = *buf; + + p += strspn(p, WHITESPACE); + l = strcspn(p, WHITESPACE); + + if (l <= 0 || + p[l-1] != ':') + return; + + e = l; + l--; + + if (p[l-1] == ']') { + size_t k = l-1; + + for (;;) { + + if (p[k] == '[') { + t = strndup(p+k+1, l-k-2); + if (t) + *pid = t; + + l = k; + break; + } + + if (k == 0) + break; + + k--; + } + } + + t = strndup(p, l); + if (t) + *identifier = t; + + *buf = p + e; + *buf += strspn(*buf, WHITESPACE); +} + +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; + int priority = LOG_USER | LOG_INFO; + char *identifier = NULL, *pid = NULL; + const char *orig; + + assert(s); + assert(buf); + + orig = buf; + parse_syslog_priority((char**) &buf, &priority); + + if (s->forward_to_syslog) + forward_syslog_raw(s, priority, orig, ucred, tv); + + skip_syslog_date((char**) &buf); + read_identifier(&buf, &identifier, &pid); + + if (s->forward_to_kmsg) + forward_kmsg(s, priority, identifier, buf, ucred); + + if (s->forward_to_console) + forward_console(s, priority, identifier, buf, ucred); + + IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=syslog"); + + if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0) + IOVEC_SET_STRING(iovec[n++], syslog_priority); + + if (priority & LOG_FACMASK) + if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0) + IOVEC_SET_STRING(iovec[n++], syslog_facility); + + if (identifier) { + syslog_identifier = strappend("SYSLOG_IDENTIFIER=", identifier); + if (syslog_identifier) + IOVEC_SET_STRING(iovec[n++], syslog_identifier); + } + + if (pid) { + syslog_pid = strappend("SYSLOG_PID=", pid); + if (syslog_pid) + IOVEC_SET_STRING(iovec[n++], syslog_pid); + } + + message = strappend("MESSAGE=", buf); + if (message) + IOVEC_SET_STRING(iovec[n++], message); + + dispatch_message(s, iovec, n, ELEMENTSOF(iovec), ucred, tv, label, label_len, NULL, priority); + + free(message); + free(identifier); + free(pid); + free(syslog_priority); + free(syslog_facility); + free(syslog_identifier); +} + +static bool valid_user_field(const char *p, size_t l) { + const char *a; + + /* We kinda enforce POSIX syntax recommendations for + environment variables here, but make a couple of additional + requirements. + + http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */ + + /* No empty field names */ + if (l <= 0) + return false; + + /* Don't allow names longer than 64 chars */ + if (l > 64) + return false; + + /* Variables starting with an underscore are protected */ + if (p[0] == '_') + return false; + + /* Don't allow digits as first character */ + if (p[0] >= '0' && p[0] <= '9') + return false; + + /* Only allow A-Z0-9 and '_' */ + for (a = p; a < p + l; a++) + if (!((*a >= 'A' && *a <= 'Z') || + (*a >= '0' && *a <= '9') || + *a == '_')) + return false; + + return true; +} + +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; + size_t remaining; + int priority = LOG_INFO; + char *identifier = NULL, *message = NULL; + + assert(s); + assert(buffer || buffer_size == 0); + + p = buffer; + remaining = buffer_size; + + while (remaining > 0) { + const char *e, *q; + + e = memchr(p, '\n', remaining); + + if (!e) { + /* Trailing noise, let's ignore it, and flush what we collected */ + log_debug("Received message with trailing noise, ignoring."); + break; + } + + if (e == p) { + /* Entry separator */ + dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority); + n = 0; + priority = LOG_INFO; + + p++; + remaining--; + continue; + } + + if (*p == '.' || *p == '#') { + /* Ignore control commands for now, and + * comments too. */ + remaining -= (e - p) + 1; + p = e + 1; + continue; + } /* A property follows */ - if (n+13 >= m) { + if (n+N_IOVEC_META_FIELDS >= m) { struct iovec *c; unsigned u; - u = MAX((n+13U) * 2U, 4U); + u = MAX((n+N_IOVEC_META_FIELDS+1) * 2U, 4U); c = realloc(iovec, u * sizeof(struct iovec)); if (!c) { log_error("Out of memory"); @@ -715,28 +1276,63 @@ static void process_native_message(Server *s, const void *buffer, size_t buffer_ q = memchr(p, '=', e - p); if (q) { if (valid_user_field(p, q - p)) { + size_t l; + + l = e - p; + /* If the field name starts with an * underscore, skip the variable, * since that indidates a trusted * field */ iovec[n].iov_base = (char*) p; - iovec[n].iov_len = e - p; + iovec[n].iov_len = l; n++; /* We need to determine the priority * of this entry for the rate limiting * logic */ - if (e - p == 10 && - memcmp(p, "PRIORITY=", 10) == 0 && - p[10] >= '0' && - p[10] <= '9') - priority = p[10] - '0'; + if (l == 10 && + memcmp(p, "PRIORITY=", 9) == 0 && + p[9] >= '0' && p[9] <= '9') + priority = (priority & LOG_FACMASK) | (p[9] - '0'); + + else if (l == 17 && + memcmp(p, "SYSLOG_FACILITY=", 16) == 0 && + p[16] >= '0' && p[16] <= '9') + priority = (priority & LOG_PRIMASK) | ((p[16] - '0') << 3); + + else if (l == 18 && + memcmp(p, "SYSLOG_FACILITY=", 16) == 0 && + p[16] >= '0' && p[16] <= '9' && + p[17] >= '0' && p[17] <= '9') + priority = (priority & LOG_PRIMASK) | (((p[16] - '0')*10 + (p[17] - '0')) << 3); + + else if (l >= 19 && + memcmp(p, "SYSLOG_IDENTIFIER=", 18) == 0) { + char *t; + + t = strndup(p + 18, l - 18); + if (t) { + free(identifier); + identifier = t; + } + } else if (l >= 8 && + memcmp(p, "MESSAGE=", 8) == 0) { + char *t; + + t = strndup(p + 8, l - 8); + if (t) { + free(message); + message = t; + } + } } remaining -= (e - p) + 1; p = e + 1; continue; } else { + le64_t l_le; uint64_t l; char *k; @@ -745,8 +1341,8 @@ static void process_native_message(Server *s, const void *buffer, size_t buffer_ 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') { @@ -776,109 +1372,190 @@ static void process_native_message(Server *s, const void *buffer, size_t buffer_ } } - dispatch_message(s, iovec, n, m, ucred, tv, priority); + if (n <= 0) + goto finish; + + tn = n++; + IOVEC_SET_STRING(iovec[tn], "_TRANSPORT=journal"); + + if (message) { + if (s->forward_to_syslog) + forward_syslog(s, priority, identifier, message, ucred, tv); + + if (s->forward_to_kmsg) + forward_kmsg(s, priority, identifier, message, ucred); + + if (s->forward_to_console) + forward_console(s, priority, identifier, message, ucred); + } + + dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority); + +finish: + for (j = 0; j < n; j++) { + if (j == tn) + continue; - for (j = 0; j < n; j++) if (iovec[j].iov_base < buffer || (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size) free(iovec[j].iov_base); + } + + free(iovec); + free(identifier); + free(message); } -static int stdout_stream_log(StdoutStream *s, const char *p, size_t l) { - struct iovec iovec[15]; - char *message = NULL, *syslog_priority = NULL; - unsigned n = 0; - size_t tag_len; - int priority; +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(p); + assert(fd >= 0); - priority = s->priority; + /* 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->priority_prefix && - l > 3 && - p[0] == '<' && - p[1] >= '0' && p[1] <= '7' && - p[2] == '>') { + if (!S_ISREG(st.st_mode)) { + log_error("File passed is not regular. Ignoring."); + return; + } + + if (st.st_size <= 0) + return; - priority = p[1] - '0'; - p += 3; - l -= 3; + if (st.st_size > ENTRY_SIZE_MAX) { + log_error("File passed too large. Ignoring."); + return; } - if (l <= 0) + 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; - if (asprintf(&syslog_priority, "PRIORITY=%i", priority) >= 0) - IOVEC_SET_STRING(iovec[n++], syslog_priority); + priority = s->priority; - tag_len = s->tag ? strlen(s->tag) + 2: 0; - message = malloc(8 + tag_len + l); - if (message) { - memcpy(message, "MESSAGE=", 8); + if (s->level_prefix) + parse_syslog_priority((char**) &p, &priority); - if (s->tag) { - memcpy(message+8, s->tag, tag_len-2); - memcpy(message+8+tag_len-2, ": ", 2); - } + if (s->forward_to_syslog || s->server->forward_to_syslog) + forward_syslog(s->server, fixup_priority(priority), s->identifier, p, &s->ucred, NULL); - memcpy(message+8+tag_len, p, l); - iovec[n].iov_base = message; - iovec[n].iov_len = 8+tag_len+l; - n++; - } + if (s->forward_to_kmsg || s->server->forward_to_kmsg) + forward_kmsg(s->server, priority, s->identifier, p, &s->ucred); - dispatch_message(s->server, iovec, n, ELEMENTSOF(iovec), &s->ucred, NULL, priority); + if (s->forward_to_console || s->server->forward_to_console) + forward_console(s->server, priority, s->identifier, p, &s->ucred); - if (s->tee_console) { - int console; + IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=stdout"); - console = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC); - if (console >= 0) { - n = 0; - if (s->tag) { - IOVEC_SET_STRING(iovec[n++], s->tag); - IOVEC_SET_STRING(iovec[n++], ": "); - } + if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0) + IOVEC_SET_STRING(iovec[n++], syslog_priority); - iovec[n].iov_base = (void*) p; - iovec[n].iov_len = l; - n++; + if (priority & LOG_FACMASK) + if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0) + IOVEC_SET_STRING(iovec[n++], syslog_facility); - IOVEC_SET_STRING(iovec[n++], (char*) "\n"); + if (s->identifier) { + syslog_identifier = strappend("SYSLOG_IDENTIFIER=", s->identifier); + if (syslog_identifier) + IOVEC_SET_STRING(iovec[n++], syslog_identifier); + } - writev(console, iovec, n); - } + message = strappend("MESSAGE=", p); + if (message) + IOVEC_SET_STRING(iovec[n++], message); + +#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, s->unit_id, priority); free(message); free(syslog_priority); + free(syslog_facility); + free(syslog_identifier); return 0; } -static int stdout_stream_line(StdoutStream *s, const char *p, size_t l) { +static int stdout_stream_line(StdoutStream *s, char *p) { + int r; + assert(s); assert(p); - while (l > 0 && strchr(WHITESPACE, *p)) { - l--; - p++; - } - - while (l > 0 && strchr(WHITESPACE, *(p+l-1))) - l--; + p = strstrip(p); switch (s->state) { - case STDOUT_STREAM_TAG: - - if (l > 0) { - s->tag = strndup(p, l); - if (!s->tag) { + case STDOUT_STREAM_IDENTIFIER: + if (isempty(p)) + s->identifier = NULL; + else { + s->identifier = strdup(p); + if (!s->identifier) { log_error("Out of memory"); - return -EINVAL; + return -ENOMEM; + } + } + + s->state = STDOUT_STREAM_UNIT_ID; + return 0; + + case STDOUT_STREAM_UNIT_ID: + if (s->ucred.uid == 0) { + if (isempty(p)) + s->unit_id = NULL; + else { + s->unit_id = strdup(p); + if (!s->unit_id) { + log_error("Out of memory"); + return -ENOMEM; + } } } @@ -886,37 +1563,61 @@ static int stdout_stream_line(StdoutStream *s, const char *p, size_t l) { return 0; case STDOUT_STREAM_PRIORITY: - if (l != 1 || *p < '0' || *p > '7') { + r = safe_atoi(p, &s->priority); + if (r < 0 || s->priority <= 0 || s->priority >= 999) { log_warning("Failed to parse log priority line."); return -EINVAL; } - s->priority = *p - '0'; - s->state = STDOUT_STREAM_PRIORITY_PREFIX; + s->state = STDOUT_STREAM_LEVEL_PREFIX; + return 0; + + case STDOUT_STREAM_LEVEL_PREFIX: + r = parse_boolean(p); + if (r < 0) { + log_warning("Failed to parse level prefix line."); + return -EINVAL; + } + + s->level_prefix = !!r; + s->state = STDOUT_STREAM_FORWARD_TO_SYSLOG; + return 0; + + case STDOUT_STREAM_FORWARD_TO_SYSLOG: + r = parse_boolean(p); + if (r < 0) { + log_warning("Failed to parse forward to syslog line."); + return -EINVAL; + } + + s->forward_to_syslog = !!r; + s->state = STDOUT_STREAM_FORWARD_TO_KMSG; return 0; - case STDOUT_STREAM_PRIORITY_PREFIX: - if (l != 1 || *p < '0' || *p > '1') { - log_warning("Failed to parse priority prefix line."); + case STDOUT_STREAM_FORWARD_TO_KMSG: + r = parse_boolean(p); + if (r < 0) { + log_warning("Failed to parse copy to kmsg line."); return -EINVAL; } - s->priority_prefix = *p - '0'; - s->state = STDOUT_STREAM_TEE_CONSOLE; + s->forward_to_kmsg = !!r; + s->state = STDOUT_STREAM_FORWARD_TO_CONSOLE; return 0; - case STDOUT_STREAM_TEE_CONSOLE: - if (l != 1 || *p < '0' || *p > '1') { - log_warning("Failed to parse tee to console line."); + case STDOUT_STREAM_FORWARD_TO_CONSOLE: + r = parse_boolean(p); + if (r < 0) { + log_warning("Failed to parse copy to console line."); return -EINVAL; } - s->tee_console = *p - '0'; + s->forward_to_console = !!r; s->state = STDOUT_STREAM_RUNNING; return 0; case STDOUT_STREAM_RUNNING: - return stdout_stream_log(s, p, l); + return stdout_stream_log(s, p); } assert_not_reached("Unknown stream state"); @@ -936,16 +1637,17 @@ static int stdout_stream_scan(StdoutStream *s, bool force_flush) { size_t skip; end = memchr(p, '\n', remaining); - if (!end) { - if (remaining >= LINE_MAX) { - end = p + LINE_MAX; - skip = LINE_MAX; - } else - break; - } else + if (end) skip = end - p + 1; + else if (remaining >= sizeof(s->buffer) - 1) { + end = p + sizeof(s->buffer) - 1; + skip = remaining; + } else + break; - r = stdout_stream_line(s, p, end - p); + *end = 0; + + r = stdout_stream_line(s, p); if (r < 0) return r; @@ -954,7 +1656,8 @@ static int stdout_stream_scan(StdoutStream *s, bool force_flush) { } if (force_flush && remaining > 0) { - r = stdout_stream_line(s, p, remaining); + p[remaining] = 0; + r = stdout_stream_line(s, p); if (r < 0) return r; @@ -967,129 +1670,537 @@ static int stdout_stream_scan(StdoutStream *s, bool force_flush) { s->length = remaining; } - return 0; + return 0; +} + +static int stdout_stream_process(StdoutStream *s) { + ssize_t l; + int r; + + assert(s); + + l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length); + if (l < 0) { + + if (errno == EAGAIN) + return 0; + + log_warning("Failed to read from stream: %m"); + return -errno; + } + + if (l == 0) { + r = stdout_stream_scan(s, true); + if (r < 0) + return r; + + return 0; + } + + s->length += l; + r = stdout_stream_scan(s, false); + if (r < 0) + return r; + + return 1; + +} + +static void stdout_stream_free(StdoutStream *s) { + assert(s); + + if (s->server) { + assert(s->server->n_stdout_streams > 0); + s->server->n_stdout_streams --; + LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s); + } + + if (s->fd >= 0) { + if (s->server) + epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL); + + close_nointr_nofail(s->fd); + } + +#ifdef HAVE_SELINUX + if (s->security_context) + freecon(s->security_context); +#endif + + free(s->identifier); + free(s); +} + +static int stdout_stream_new(Server *s) { + StdoutStream *stream; + int fd, r; + socklen_t len; + struct epoll_event ev; + + assert(s); + + fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC); + if (fd < 0) { + if (errno == EAGAIN) + return 0; + + log_error("Failed to accept stdout connection: %m"); + return -errno; + } + + if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) { + log_warning("Too many stdout streams, refusing connection."); + close_nointr_nofail(fd); + return 0; + } + + stream = new0(StdoutStream, 1); + if (!stream) { + log_error("Out of memory."); + close_nointr_nofail(fd); + return -ENOMEM; + } + + stream->fd = fd; + + len = sizeof(stream->ucred); + if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) { + log_error("Failed to determine peer credentials: %m"); + r = -errno; + goto fail; + } + +#ifdef HAVE_SELINUX + if (getpeercon(fd, &stream->security_context) < 0 && errno != ENOPROTOOPT) + log_error("Failed to determine peer security context: %m"); +#endif + + if (shutdown(fd, SHUT_WR) < 0) { + log_error("Failed to shutdown writing side of socket: %m"); + r = -errno; + goto fail; + } + + zero(ev); + ev.data.ptr = stream; + ev.events = EPOLLIN; + if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) { + log_error("Failed to add stream to event loop: %m"); + r = -errno; + goto fail; + } + + stream->server = s; + LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream); + s->n_stdout_streams ++; + + return 0; + +fail: + stdout_stream_free(stream); + return r; +} + +static int parse_kernel_timestamp(char **_p, usec_t *t) { + usec_t r; + int k, i; + char *p; + + assert(_p); + assert(*_p); + assert(t); + + p = *_p; + + if (strlen(p) < 14 || p[0] != '[' || p[13] != ']' || p[6] != '.') + return 0; + + r = 0; + + for (i = 1; i <= 5; i++) { + r *= 10; + + if (p[i] == ' ') + continue; + + k = undecchar(p[i]); + if (k < 0) + return 0; + + r += k; + } + + for (i = 7; i <= 12; i++) { + r *= 10; + + k = undecchar(p[i]); + if (k < 0) + return 0; + + r += k; + } + + *t = r; + *_p += 14; + *_p += strspn(*_p, WHITESPACE); + + return 1; +} + +static bool is_us(const char *pid) { + pid_t t; + + assert(pid); + + if (parse_pid(pid, &t) < 0) + return false; + + return t == getpid(); +} + +static void proc_kmsg_line(Server *s, const char *p) { + struct iovec iovec[N_IOVEC_META_FIELDS + 7]; + char *message = NULL, *syslog_priority = NULL, *syslog_pid = NULL, *syslog_facility = NULL, *syslog_identifier = NULL, *source_time = NULL; + int priority = LOG_KERN | LOG_INFO; + unsigned n = 0; + usec_t usec; + char *identifier = NULL, *pid = NULL; + + assert(s); + assert(p); + + if (isempty(p)) + return; + + parse_syslog_priority((char **) &p, &priority); + + if (s->forward_to_kmsg && (priority & LOG_FACMASK) != LOG_KERN) + return; + + if (parse_kernel_timestamp((char **) &p, &usec) > 0) { + if (asprintf(&source_time, "_SOURCE_MONOTONIC_TIMESTAMP=%llu", + (unsigned long long) usec) >= 0) + IOVEC_SET_STRING(iovec[n++], source_time); + } + + IOVEC_SET_STRING(iovec[n++], "_TRANSPORT=kernel"); + + if (asprintf(&syslog_priority, "PRIORITY=%i", priority & LOG_PRIMASK) >= 0) + IOVEC_SET_STRING(iovec[n++], syslog_priority); + + if ((priority & LOG_FACMASK) == LOG_KERN) { + + if (s->forward_to_syslog) + forward_syslog(s, priority, "kernel", p, NULL, NULL); + + IOVEC_SET_STRING(iovec[n++], "SYSLOG_IDENTIFIER=kernel"); + } else { + read_identifier(&p, &identifier, &pid); + + /* Avoid any messages we generated ourselves via + * log_info() and friends. */ + if (pid && is_us(pid)) + goto finish; + + if (s->forward_to_syslog) + forward_syslog(s, priority, identifier, p, NULL, NULL); + + if (identifier) { + syslog_identifier = strappend("SYSLOG_IDENTIFIER=", identifier); + if (syslog_identifier) + IOVEC_SET_STRING(iovec[n++], syslog_identifier); + } + + if (pid) { + syslog_pid = strappend("SYSLOG_PID=", pid); + if (syslog_pid) + IOVEC_SET_STRING(iovec[n++], syslog_pid); + } + + if (asprintf(&syslog_facility, "SYSLOG_FACILITY=%i", LOG_FAC(priority)) >= 0) + IOVEC_SET_STRING(iovec[n++], syslog_facility); + } + + message = strappend("MESSAGE=", p); + if (message) + IOVEC_SET_STRING(iovec[n++], message); + + dispatch_message(s, iovec, n, ELEMENTSOF(iovec), NULL, NULL, NULL, 0, NULL, priority); + +finish: + free(message); + free(syslog_priority); + free(syslog_identifier); + free(syslog_pid); + free(syslog_facility); + free(source_time); + free(identifier); + free(pid); +} + +static void proc_kmsg_scan(Server *s) { + char *p; + size_t remaining; + + assert(s); + + p = s->proc_kmsg_buffer; + remaining = s->proc_kmsg_length; + for (;;) { + char *end; + size_t skip; + + end = memchr(p, '\n', remaining); + if (end) + skip = end - p + 1; + else if (remaining >= sizeof(s->proc_kmsg_buffer) - 1) { + end = p + sizeof(s->proc_kmsg_buffer) - 1; + skip = remaining; + } else + break; + + *end = 0; + + proc_kmsg_line(s, p); + + remaining -= skip; + p += skip; + } + + if (p > s->proc_kmsg_buffer) { + memmove(s->proc_kmsg_buffer, p, remaining); + s->proc_kmsg_length = remaining; + } +} + +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) + 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 + * path, but not the prefix. + * + * If in persistent mode: create /var/log/journal and + * the machine path */ + + if (s->storage == STORAGE_PERSISTENT) + (void) mkdir("/var/log/journal/", 0755); + + fn = strappend("/var/log/journal/", ids); + if (!fn) + return -ENOMEM; + + (void) mkdir(fn, 0755); + free(fn); + + fn = strjoin("/var/log/journal/", ids, "/system.journal", NULL); + if (!fn) + return -ENOMEM; + + r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, &s->system_metrics, NULL, &s->system_journal); + free(fn); + + if (r >= 0) { + s->system_journal->compress = s->compress; + + 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 + * if it already exists, so that we can flush + * it into the system journal */ + + r = journal_file_open(fn, O_RDWR, 0640, &s->runtime_metrics, 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 + * it if necessary. */ + + (void) mkdir_parents(fn, 0755); + r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, &s->runtime_metrics, 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) { + s->runtime_journal->compress = s->compress; + + server_fix_perms(s, s->runtime_journal, 0); + } + } + + return r; } -static int stdout_stream_process(StdoutStream *s) { - ssize_t l; +static int server_flush_to_var(Server *s) { + Object *o = NULL; int r; + sd_id128_t machine; + sd_journal *j; assert(s); - l = read(s->fd, s->buffer+s->length, sizeof(s->buffer)-1-s->length); - if (l < 0) { - - if (errno == EAGAIN) - return 0; + if (s->storage != STORAGE_AUTO && + s->storage != STORAGE_PERSISTENT) + return 0; - log_warning("Failed to read from stream: %m"); - return -errno; - } + if (!s->runtime_journal) + return 0; - if (l == 0) { - r = stdout_stream_scan(s, true); - if (r < 0) - return r; + system_journal_open(s); + if (!s->system_journal) return 0; + + log_info("Flushing to /var..."); + + r = sd_id128_get_machine(&machine); + if (r < 0) { + log_error("Failed to get machine id: %s", strerror(-r)); + return r; } - s->length += l; - r = stdout_stream_scan(s, false); - if (r < 0) + r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY); + if (r < 0) { + log_error("Failed to read runtime journal: %s", strerror(-r)); return r; + } - return 1; + SD_JOURNAL_FOREACH(j) { + JournalFile *f; -} + f = j->current_file; + assert(f && f->current_offset > 0); -static void stdout_stream_free(StdoutStream *s) { - assert(s); + 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; + } - if (s->server) { - assert(s->server->n_stdout_streams > 0); - s->server->n_stdout_streams --; - LIST_REMOVE(StdoutStream, stdout_stream, s->server->stdout_streams, s); + r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); + if (r == -E2BIG) { + log_info("Allocation limit reached."); + + journal_file_post_change(s->system_journal); + server_rotate(s); + server_vacuum(s); + + 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; + } } - if (s->fd >= 0) { - if (s->server) - epoll_ctl(s->server->epoll_fd, EPOLL_CTL_DEL, s->fd, NULL); +finish: + journal_file_post_change(s->system_journal); - close_nointr_nofail(s->fd); + journal_file_close(s->runtime_journal); + s->runtime_journal = NULL; + + if (r >= 0) { + char path[] = "/run/log/journal/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"; + sd_id128_to_string(machine, path + 17); + rm_rf(path, false, true, false); } - free(s->tag); - free(s); + return r; } -static int stdout_stream_new(Server *s) { - StdoutStream *stream; - int fd, r; - socklen_t len; - struct epoll_event ev; - +static int server_read_proc_kmsg(Server *s) { + ssize_t l; assert(s); + assert(s->proc_kmsg_fd >= 0); - fd = accept4(s->stdout_fd, NULL, NULL, SOCK_NONBLOCK|SOCK_CLOEXEC); - if (fd < 0) { - if (errno == EAGAIN) + l = read(s->proc_kmsg_fd, s->proc_kmsg_buffer + s->proc_kmsg_length, sizeof(s->proc_kmsg_buffer) - 1 - s->proc_kmsg_length); + if (l == 0) /* the kernel is stupid and in some race + * conditions returns 0 in the middle of the + * stream. */ + return 0; + if (l < 0) { + + if (errno == EAGAIN || errno == EINTR) return 0; - log_error("Failed to accept stdout connection: %m"); + log_error("Failed to read from kernel: %m"); return -errno; } - if (s->n_stdout_streams >= STDOUT_STREAMS_MAX) { - log_warning("Too many stdout streams, refusing connection."); - close_nointr_nofail(fd); - return 0; - } + s->proc_kmsg_length += l; - stream = new0(StdoutStream, 1); - if (!stream) { - log_error("Out of memory."); - close_nointr_nofail(fd); - return -ENOMEM; - } + proc_kmsg_scan(s); + return 1; +} - stream->fd = fd; +static int server_flush_proc_kmsg(Server *s) { + int r; - len = sizeof(stream->ucred); - if (getsockopt(fd, SOL_SOCKET, SO_PEERCRED, &stream->ucred, &len) < 0) { - log_error("Failed to determine peer credentials: %m"); - r = -errno; - goto fail; - } + assert(s); - if (shutdown(fd, SHUT_WR) < 0) { - log_error("Failed to shutdown writing side of socket: %m"); - r = -errno; - goto fail; - } + if (s->proc_kmsg_fd < 0) + return 0; - zero(ev); - ev.data.ptr = stream; - ev.events = EPOLLIN; - if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, fd, &ev) < 0) { - log_error("Failed to add stream to event loop: %m"); - r = -errno; - goto fail; - } + log_info("Flushing /proc/kmsg..."); - stream->server = s; - LIST_PREPEND(StdoutStream, stdout_stream, s->stdout_streams, stream); - s->n_stdout_streams ++; + for (;;) { + r = server_read_proc_kmsg(s); + if (r < 0) + return r; - return 0; + if (r == 0) + break; + } -fail: - stdout_stream_free(stream); - return r; + return 0; } static int process_event(Server *s, struct epoll_event *ev) { assert(s); + assert(ev); if (ev->data.fd == s->signal_fd) { struct signalfd_siginfo sfsi; @@ -1107,14 +2218,40 @@ static int process_event(Server *s, struct epoll_event *ev) { return -EIO; if (errno == EINTR || errno == EAGAIN) - return 0; + return 1; return -errno; } + if (sfsi.ssi_signo == SIGUSR1) { + touch("/run/systemd/journal/flushed"); + server_flush_to_var(s); + return 1; + } + + if (sfsi.ssi_signo == SIGUSR2) { + server_rotate(s); + server_vacuum(s); + return 1; + } + log_debug("Received SIG%s", signal_to_string(sfsi.ssi_signo)); return 0; + } else if (ev->data.fd == s->proc_kmsg_fd) { + int r; + + if (ev->events != EPOLLIN) { + log_info("Got invalid event from epoll."); + return -EIO; + } + + r = server_read_proc_kmsg(s); + if (r < 0) + return r; + + return 1; + } else if (ev->data.fd == s->native_fd || ev->data.fd == s->syslog_fd) { @@ -1129,22 +2266,35 @@ static int process_event(Server *s, struct epoll_event *ev) { 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; @@ -1172,7 +2322,7 @@ static int process_event(Server *s, struct epoll_event *ev) { 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) @@ -1189,23 +2339,44 @@ static int process_event(Server *s, struct epoll_event *ev) { 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; @@ -1247,66 +2418,6 @@ static int process_event(Server *s, struct epoll_event *ev) { return 0; } -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) - return r; - - /* First try to create the machine path, but not the prefix */ - fn = strappend("/var/log/journal/", sd_id128_to_string(machine, ids)); - if (!fn) - return -ENOMEM; - (void) mkdir(fn, 0755); - free(fn); - - /* The create the system journal file */ - fn = join("/var/log/journal/", ids, "/system.journal", NULL); - if (!fn) - return -ENOMEM; - - r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->system_journal); - free(fn); - - if (r >= 0) { - s->system_journal->metrics = s->metrics; - s->system_journal->compress = s->compress; - - fix_perms(s->system_journal, 0); - return r; - } - - if (r < 0 && r != -ENOENT) { - log_error("Failed to open system journal: %s", strerror(-r)); - return r; - } - - /* /var didn't work, so try /run, but this time we - * create the prefix too */ - fn = join("/run/log/journal/", ids, "/system.journal", NULL); - if (!fn) - return -ENOMEM; - - (void) mkdir_parents(fn, 0755); - r = journal_file_open(fn, O_RDWR|O_CREAT, 0640, NULL, &s->runtime_journal); - free(fn); - - if (r < 0) { - log_error("Failed to open runtime journal: %s", strerror(-r)); - return r; - } - - s->runtime_journal->metrics = s->metrics; - s->runtime_journal->compress = s->compress; - - fix_perms(s->runtime_journal, 0); - return r; -} - static int open_syslog_socket(Server *s) { union sockaddr_union sa; int one, r; @@ -1316,7 +2427,7 @@ static int open_syslog_socket(Server *s) { if (s->syslog_fd < 0) { - s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0); + s->syslog_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); if (s->syslog_fd < 0) { log_error("socket() failed: %m"); return -errno; @@ -1324,7 +2435,7 @@ static int open_syslog_socket(Server *s) { zero(sa); sa.un.sun_family = AF_UNIX; - strncpy(sa.un.sun_path, "/run/systemd/syslog", sizeof(sa.un.sun_path)); + strncpy(sa.un.sun_path, "/dev/log", sizeof(sa.un.sun_path)); unlink(sa.un.sun_path); @@ -1335,7 +2446,8 @@ static int open_syslog_socket(Server *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)); @@ -1344,6 +2456,13 @@ static int open_syslog_socket(Server *s) { 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) { @@ -1371,7 +2490,7 @@ static int open_native_socket(Server*s) { if (s->native_fd < 0) { - s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0); + s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); if (s->native_fd < 0) { log_error("socket() failed: %m"); return -errno; @@ -1379,7 +2498,7 @@ static int open_native_socket(Server*s) { zero(sa); sa.un.sun_family = AF_UNIX; - strncpy(sa.un.sun_path, "/run/systemd/journal", sizeof(sa.un.sun_path)); + strncpy(sa.un.sun_path, "/run/systemd/journal/socket", sizeof(sa.un.sun_path)); unlink(sa.un.sun_path); @@ -1390,7 +2509,8 @@ static int open_native_socket(Server*s) { } 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)); @@ -1399,6 +2519,13 @@ static int open_native_socket(Server*s) { 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) { @@ -1426,7 +2553,7 @@ static int open_stdout_socket(Server *s) { if (s->stdout_fd < 0) { - s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0); + s->stdout_fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0); if (s->stdout_fd < 0) { log_error("socket() failed: %m"); return -errno; @@ -1434,7 +2561,7 @@ static int open_stdout_socket(Server *s) { zero(sa); sa.un.sun_family = AF_UNIX; - strncpy(sa.un.sun_path, "/run/systemd/stdout", sizeof(sa.un.sun_path)); + strncpy(sa.un.sun_path, "/run/systemd/journal/stdout", sizeof(sa.un.sun_path)); unlink(sa.un.sun_path); @@ -1450,7 +2577,8 @@ static int open_stdout_socket(Server *s) { log_error("liste() failed: %m"); return -errno; } - } + } else + fd_nonblock(s->stdout_fd, 1); zero(ev); ev.events = EPOLLIN; @@ -1463,6 +2591,31 @@ static int open_stdout_socket(Server *s) { return 0; } +static int open_proc_kmsg(Server *s) { + struct epoll_event ev; + + assert(s); + + if (!s->import_proc_kmsg) + return 0; + + 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; + } + + zero(ev); + ev.events = EPOLLIN; + ev.data.fd = s->proc_kmsg_fd; + if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->proc_kmsg_fd, &ev) < 0) { + log_error("Failed to add /proc/kmsg fd to epoll object: %m"); + return -errno; + } + + return 0; +} + static int open_signalfd(Server *s) { sigset_t mask; struct epoll_event ev; @@ -1470,7 +2623,7 @@ static int open_signalfd(Server *s) { assert(s); assert_se(sigemptyset(&mask) == 0); - sigset_add_many(&mask, SIGINT, SIGTERM, -1); + 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); @@ -1491,19 +2644,111 @@ static int open_signalfd(Server *s) { return 0; } +static int server_parse_proc_cmdline(Server *s) { + char *line, *w, *state; + int r; + size_t l; + + if (detect_container(NULL) > 0) + return 0; + + r = read_one_line_file("/proc/cmdline", &line); + if (r < 0) { + log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r)); + return 0; + } + + FOREACH_WORD_QUOTED(w, l, line, state) { + char *word; + + word = strndup(w, l); + if (!word) { + r = -ENOMEM; + goto finish; + } + + 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."); + + free(word); + } + + r = 0; + +finish: + free(line); + return r; +} + +static int server_parse_config_file(Server *s) { + FILE *f; + const char *fn; + int r; + + assert(s); + + fn = "/etc/systemd/journald.conf"; + 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(fn, f, "Journal\0", config_item_perf_lookup, (void*) journald_gperf_lookup, false, s); + if (r < 0) + log_warning("Failed to parse configuration file: %s", strerror(-r)); + + fclose(f); + + return r; +} + static int server_init(Server *s) { int n, r, fd; assert(s); zero(*s); - s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = -1; - s->metrics.max_size = DEFAULT_MAX_SIZE; - s->metrics.min_size = DEFAULT_MIN_SIZE; - s->metrics.keep_free = DEFAULT_KEEP_FREE; - s->max_use = DEFAULT_MAX_USE; + s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = s->proc_kmsg_fd = -1; s->compress = true; + s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL; + s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST; + + s->forward_to_syslog = true; + + s->max_level_store = LOG_DEBUG; + s->max_level_syslog = LOG_DEBUG; + s->max_level_kmsg = LOG_NOTICE; + s->max_level_console = LOG_INFO; + + 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); + s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func); if (!s->user_journals) { log_error("Out of memory."); @@ -1524,7 +2769,7 @@ static int server_init(Server *s) { for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) { - if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/native", 0) > 0) { + if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/socket", 0) > 0) { if (s->native_fd >= 0) { log_error("Too many native sockets passed."); @@ -1533,7 +2778,7 @@ static int server_init(Server *s) { s->native_fd = fd; - } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/stdout", 0) > 0) { + } else if (sd_is_socket_unix(fd, SOCK_STREAM, 1, "/run/systemd/journal/stdout", 0) > 0) { if (s->stdout_fd >= 0) { log_error("Too many stdout sockets passed."); @@ -1569,7 +2814,7 @@ static int server_init(Server *s) { if (r < 0) return r; - r = system_journal_open(s); + r = open_proc_kmsg(s); if (r < 0) return r; @@ -1577,10 +2822,14 @@ static int server_init(Server *s) { if (r < 0) return r; - s->rate_limit = journal_rate_limit_new(DEFAULT_RATE_LIMIT_INTERVAL, DEFAULT_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 = system_journal_open(s); + if (r < 0) + return r; + return 0; } @@ -1617,8 +2866,14 @@ static void server_done(Server *s) { if (s->stdout_fd >= 0) close_nointr_nofail(s->stdout_fd); + if (s->proc_kmsg_fd >= 0) + close_nointr_nofail(s->proc_kmsg_fd); + if (s->rate_limit) journal_rate_limit_free(s->rate_limit); + + free(s->buffer); + free(s->tty_path); } int main(int argc, char *argv[]) { @@ -1635,8 +2890,8 @@ int main(int argc, char *argv[]) { return EXIT_FAILURE; } - log_set_target(LOG_TARGET_CONSOLE); - log_set_max_level(LOG_DEBUG); + log_set_target(LOG_TARGET_SAFE); + log_set_facility(LOG_SYSLOG); log_parse_environment(); log_open(); @@ -1646,7 +2901,12 @@ int main(int argc, char *argv[]) { if (r < 0) goto finish; + server_vacuum(&server); + server_flush_to_var(&server); + server_flush_proc_kmsg(&server); + log_debug("systemd-journald running as pid %lu", (unsigned long) getpid()); + driver_message(&server, SD_MESSAGE_JOURNAL_START, "Journal started"); sd_notify(false, "READY=1\n" @@ -1675,6 +2935,7 @@ int main(int argc, char *argv[]) { } log_debug("systemd-journald stopped as pid %lu", (unsigned long) getpid()); + driver_message(&server, SD_MESSAGE_JOURNAL_STOP, "Journal stopped"); finish: sd_notify(false,