X-Git-Url: https://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournald-server.c;h=52111f75a03ceb808084d0eb1ca2ccd1a3313d3e;hp=43ffe75560efb4e5c85de6bb1d9fbf1e96e3402a;hb=b1389b0d0805392570085acc7cb10eafcf885405;hpb=93b73b064c663d6248bebfbbbd82989b5ca10fc5 diff --git a/src/journal/journald-server.c b/src/journal/journald-server.c index 43ffe7556..52111f75a 100644 --- a/src/journal/journald-server.c +++ b/src/journal/journald-server.c @@ -24,35 +24,33 @@ #include #include #include +#include #include -#include -#include -#include - -#ifdef HAVE_LOGIND -#include -#endif +#include "sd-journal.h" +#include "sd-messages.h" +#include "sd-daemon.h" +#include "fileio.h" #include "mkdir.h" #include "hashmap.h" #include "journal-file.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 @@ -66,12 +64,14 @@ #define USER_JOURNALS_MAX 1024 -#define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC) -#define DEFAULT_RATE_LIMIT_BURST 200 +#define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE) +#define DEFAULT_RATE_LIMIT_INTERVAL (30*USEC_PER_SEC) +#define DEFAULT_RATE_LIMIT_BURST 1000 +#define DEFAULT_MAX_FILE_USEC USEC_PER_MONTH #define RECHECK_AVAILABLE_SPACE_USEC (30*USEC_PER_SEC) -static const char* const storage_table[] = { +static const char* const storage_table[_STORAGE_MAX] = { [STORAGE_AUTO] = "auto", [STORAGE_VOLATILE] = "volatile", [STORAGE_PERSISTENT] = "persistent", @@ -81,29 +81,31 @@ static const char* const storage_table[] = { DEFINE_STRING_TABLE_LOOKUP(storage, Storage); DEFINE_CONFIG_PARSE_ENUM(config_parse_storage, storage, Storage, "Failed to parse storage setting"); -static const char* const split_mode_table[] = { - [SPLIT_NONE] = "none", +static const char* const split_mode_table[_SPLIT_MAX] = { + [SPLIT_LOGIN] = "login", [SPLIT_UID] = "uid", - [SPLIT_LOGIN] = "login" + [SPLIT_NONE] = "none", }; DEFINE_STRING_TABLE_LOOKUP(split_mode, SplitMode); DEFINE_CONFIG_PARSE_ENUM(config_parse_split_mode, split_mode, SplitMode, "Failed to parse split mode setting"); -static uint64_t available_space(Server *s) { - char ids[33], *p; - const char *f; +static uint64_t available_space(Server *s, bool verbose) { + char ids[33]; + _cleanup_free_ char *p = NULL; sd_id128_t machine; struct statvfs ss; - uint64_t sum = 0, avail = 0, ss_avail = 0; + uint64_t sum = 0, ss_avail = 0, avail = 0; int r; - DIR *d; + _cleanup_closedir_ DIR *d = NULL; usec_t ts; + const char *f; JournalMetrics *m; ts = now(CLOCK_MONOTONIC); - if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts) + if (s->cached_available_space_timestamp + RECHECK_AVAILABLE_SPACE_USEC > ts + && !verbose) return s->cached_available_space; r = sd_id128_get_machine(&machine); @@ -125,22 +127,20 @@ static uint64_t available_space(Server *s) { return 0; d = opendir(p); - free(p); - if (!d) return 0; if (fstatvfs(dirfd(d), &ss) < 0) - goto finish; + return 0; for (;;) { struct stat st; struct dirent *de; - union dirent_storage buf; - r = readdir_r(d, &buf.de, &de); - if (r != 0) - break; + errno = 0; + de = readdir(d); + if (!de && errno != 0) + return 0; if (!de) break; @@ -158,41 +158,37 @@ static uint64_t available_space(Server *s) { sum += (uint64_t) st.st_blocks * 512UL; } - avail = sum >= m->max_use ? 0 : m->max_use - sum; - ss_avail = ss.f_bsize * ss.f_bavail; - ss_avail = ss_avail < m->keep_free ? 0 : ss_avail - m->keep_free; - - if (ss_avail < avail) - avail = ss_avail; - - s->cached_available_space = avail; - s->cached_available_space_timestamp = ts; - -finish: - closedir(d); - - return avail; -} + /* If we reached a high mark, we will always allow this much + * again, unless usage goes above max_use. This watermark + * value is cached so that we don't give up space on pressure, + * but hover below the maximum usage. */ -static void server_read_file_gid(Server *s) { - const char *adm = "adm"; - int r; + if (m->use < sum) + m->use = sum; - assert(s); + avail = LESS_BY(ss_avail, m->keep_free); - if (s->file_gid_valid) - return; + s->cached_available_space = LESS_BY(MIN(m->max_use, avail), sum); + s->cached_available_space_timestamp = ts; - r = get_group_creds(&adm, &s->file_gid); - if (r < 0) - log_warning("Failed to resolve 'adm' group: %s", strerror(-r)); + if (verbose) { + char fb1[FORMAT_BYTES_MAX], fb2[FORMAT_BYTES_MAX], fb3[FORMAT_BYTES_MAX], + fb4[FORMAT_BYTES_MAX], fb5[FORMAT_BYTES_MAX]; + + server_driver_message(s, SD_MESSAGE_JOURNAL_USAGE, + "%s journal is using %s (max allowed %s, " + "trying to leave %s free of %s available → current limit %s).", + s->system_journal ? "Permanent" : "Runtime", + format_bytes(fb1, sizeof(fb1), sum), + format_bytes(fb2, sizeof(fb2), m->max_use), + format_bytes(fb3, sizeof(fb3), m->keep_free), + format_bytes(fb4, sizeof(fb4), ss_avail), + format_bytes(fb5, sizeof(fb5), s->cached_available_space + sum)); + } - /* 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; + return s->cached_available_space; } void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { @@ -205,14 +201,12 @@ void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { 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 (uid <= SYSTEM_UID_MAX) return; acl = acl_get_fd(f->fd); @@ -232,9 +226,11 @@ void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { } } + /* We do not recalculate the mask unconditionally here, + * so that the fchmod() mask above stays intact. */ if (acl_get_permset(entry, &permset) < 0 || acl_add_perm(permset, ACL_READ) < 0 || - acl_calc_mask(&acl) < 0) { + calc_acl_mask_if_needed(&acl) < 0) { log_warning("Failed to patch ACL on %s, ignoring: %m", f->path); goto finish; } @@ -248,7 +244,7 @@ finish: } static JournalFile* find_journal(Server *s, uid_t uid) { - char *p; + _cleanup_free_ char *p = NULL; int r; JournalFile *f; sd_id128_t machine; @@ -263,7 +259,7 @@ static JournalFile* find_journal(Server *s, uid_t uid) { if (s->runtime_journal) return s->runtime_journal; - if (uid <= 0) + if (uid <= SYSTEM_UID_MAX) return s->system_journal; r = sd_id128_get_machine(&machine); @@ -274,8 +270,8 @@ static JournalFile* find_journal(Server *s, uid_t uid) { if (f) return f; - if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-%lu.journal", - SD_ID128_FORMAT_VAL(machine), (unsigned long) uid) < 0) + if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/user-"UID_FMT".journal", + SD_ID128_FORMAT_VAL(machine), uid) < 0) return s->system_journal; while (hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) { @@ -285,9 +281,7 @@ static JournalFile* find_journal(Server *s, uid_t uid) { journal_file_close(f); } - r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, s->system_journal, &f); - free(p); - + r = journal_file_open_reliably(p, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &f); if (r < 0) return s->system_journal; @@ -302,6 +296,27 @@ static JournalFile* find_journal(Server *s, uid_t uid) { return f; } +static int do_rotate(Server *s, JournalFile **f, const char* name, + bool seal, uint32_t uid) { + int r; + assert(s); + + if (!*f) + return -EINVAL; + + r = journal_file_rotate(f, s->compress, seal); + if (r < 0) + if (*f) + log_error("Failed to rotate %s: %s", + (*f)->path, strerror(-r)); + else + log_error("Failed to create new %s journal: %s", + name, strerror(-r)); + else + server_fix_perms(s, *f, uid); + return r; +} + void server_rotate(Server *s) { JournalFile *f; void *k; @@ -310,45 +325,61 @@ void server_rotate(Server *s) { log_debug("Rotating..."); - if (s->runtime_journal) { - r = journal_file_rotate(&s->runtime_journal, s->compress, false); - if (r < 0) - if (s->runtime_journal) - log_error("Failed to rotate %s: %s", s->runtime_journal->path, strerror(-r)); - else - log_error("Failed to create new runtime journal: %s", strerror(-r)); - else - server_fix_perms(s, s->runtime_journal, 0); + do_rotate(s, &s->runtime_journal, "runtime", false, 0); + do_rotate(s, &s->system_journal, "system", s->seal, 0); + + HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { + r = do_rotate(s, &f, "user", s->seal, PTR_TO_UINT32(k)); + if (r >= 0) + hashmap_replace(s->user_journals, k, f); + else if (!f) + /* Old file has been closed and deallocated */ + hashmap_remove(s->user_journals, k); } +} + +void server_sync(Server *s) { + JournalFile *f; + void *k; + Iterator i; + int r; if (s->system_journal) { - r = journal_file_rotate(&s->system_journal, s->compress, s->seal); + r = journal_file_set_offline(s->system_journal); if (r < 0) - if (s->system_journal) - log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r)); - else - log_error("Failed to create new system journal: %s", strerror(-r)); - - else - server_fix_perms(s, s->system_journal, 0); + log_error("Failed to sync system journal: %s", strerror(-r)); } HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { - r = journal_file_rotate(&f, s->compress, s->seal); + r = journal_file_set_offline(f); if (r < 0) - if (f) - log_error("Failed to rotate %s: %s", f->path, strerror(-r)); - else - log_error("Failed to create user journal: %s", strerror(-r)); - else { - hashmap_replace(s->user_journals, k, f); - server_fix_perms(s, f, PTR_TO_UINT32(k)); - } + log_error("Failed to sync user journal: %s", strerror(-r)); } + + 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; } -void server_vacuum(Server *s) { +static void do_vacuum(Server *s, char *ids, JournalFile *f, const char* path, + JournalMetrics *metrics) { char *p; + int r; + + if (!f) + return; + + p = strappenda(path, ids); + r = journal_directory_vacuum(p, metrics->max_use, s->max_retention_usec, &s->oldest_file_usec); + if (r < 0 && r != -ENOENT) + log_error("Failed to vacuum %s: %s", p, strerror(-r)); +} + +void server_vacuum(Server *s) { char ids[33]; sd_id128_t machine; int r; @@ -362,78 +393,56 @@ void server_vacuum(Server *s) { log_error("Failed to get machine ID: %s", strerror(-r)); return; } - sd_id128_to_string(machine, ids); - if (s->system_journal) { - p = strappend("/var/log/journal/", ids); - if (!p) { - log_oom(); - return; - } + do_vacuum(s, ids, s->system_journal, "/var/log/journal/", &s->system_metrics); + do_vacuum(s, ids, s->runtime_journal, "/run/log/journal/", &s->runtime_metrics); - r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec); - if (r < 0 && r != -ENOENT) - log_error("Failed to vacuum %s: %s", p, strerror(-r)); - free(p); - } + s->cached_available_space_timestamp = 0; +} - if (s->runtime_journal) { - p = strappend("/run/log/journal/", ids); - if (!p) { - log_oom(); - return; - } +static void server_cache_machine_id(Server *s) { + sd_id128_t id; + int r; - r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec); - if (r < 0 && r != -ENOENT) - log_error("Failed to vacuum %s: %s", p, strerror(-r)); - free(p); - } + assert(s); - s->cached_available_space_timestamp = 0; + r = sd_id128_get_machine(&id); + if (r < 0) + return; + + sd_id128_to_string(id, stpcpy(s->machine_id_field, "_MACHINE_ID=")); } -static char *shortened_cgroup_path(pid_t pid) { +static void server_cache_boot_id(Server *s) { + sd_id128_t id; int r; - char *process_path, *init_path, *path; - assert(pid > 0); + assert(s); - r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, pid, &process_path); + r = sd_id128_get_boot(&id); if (r < 0) - return NULL; + return; - r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, 1, &init_path); - if (r < 0) { - free(process_path); - return NULL; - } + sd_id128_to_string(id, stpcpy(s->boot_id_field, "_BOOT_ID=")); +} - if (endswith(init_path, "/system")) - init_path[strlen(init_path) - 7] = 0; - else if (streq(init_path, "/")) - init_path[0] = 0; +static void server_cache_hostname(Server *s) { + _cleanup_free_ char *t = NULL; + char *x; - if (startswith(process_path, init_path)) { - char *p; + assert(s); - 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; - } + t = gethostname_malloc(); + if (!t) + return; - free(process_path); - free(init_path); + x = strappend("_HOSTNAME=", t); + if (!x) + return; - return path; + free(s->hostname_field); + s->hostname_field = x; } bool shall_try_append_again(JournalFile *f, int r) { @@ -465,7 +474,7 @@ bool shall_try_append_again(JournalFile *f, int r) { return true; } -static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n) { +static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned n, int priority) { JournalFile *f; bool vacuumed = false; int r; @@ -490,11 +499,18 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned } r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL); - if (r >= 0) + if (r >= 0) { + server_schedule_sync(s, priority); return; + } if (vacuumed || !shall_try_append_again(f, r)) { - log_error("Failed to write entry, ignoring: %s", strerror(-r)); + size_t size = 0; + unsigned i; + for (i = 0; i < n; i++) + size += iovec[i].iov_len; + + log_error("Failed to write entry (%d items, %zu bytes), ignoring: %s", n, size, strerror(-r)); return; } @@ -507,8 +523,15 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned log_debug("Retrying write."); r = journal_file_append_entry(f, NULL, iovec, n, &s->seqnum, NULL, NULL); - if (r < 0) - log_error("Failed to write entry, ignoring: %s", strerror(-r)); + if (r < 0) { + size_t size = 0; + unsigned i; + for (i = 0; i < n; i++) + size += iovec[i].iov_len; + + log_error("Failed to write entry (%d items, %zu bytes) despite vacuuming, ignoring: %s", n, size, strerror(-r)); + } else + server_schedule_sync(s, priority); } static void dispatch_message_real( @@ -517,186 +540,282 @@ static void dispatch_message_real( 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, *session = NULL, - *owner_uid = NULL, *unit = NULL, *selinux_context = NULL; - - char idbuf[33]; - sd_id128_t id; + const char *unit_id, + int priority, + pid_t object_pid) { + + char pid[sizeof("_PID=") + DECIMAL_STR_MAX(pid_t)], + uid[sizeof("_UID=") + DECIMAL_STR_MAX(uid_t)], + gid[sizeof("_GID=") + DECIMAL_STR_MAX(gid_t)], + owner_uid[sizeof("_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)], + source_time[sizeof("_SOURCE_REALTIME_TIMESTAMP=") + DECIMAL_STR_MAX(usec_t)], + o_uid[sizeof("OBJECT_UID=") + DECIMAL_STR_MAX(uid_t)], + o_gid[sizeof("OBJECT_GID=") + DECIMAL_STR_MAX(gid_t)], + o_owner_uid[sizeof("OBJECT_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)]; + uid_t object_uid; + gid_t object_gid; + char *x; int r; - char *t; - uid_t loginuid = 0, realuid = 0; + char *t, *c; + uid_t realuid = 0, owner = 0, journal_uid; + bool owner_valid = false; +#ifdef HAVE_AUDIT + char audit_session[sizeof("_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], + audit_loginuid[sizeof("_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)], + o_audit_session[sizeof("OBJECT_AUDIT_SESSION=") + DECIMAL_STR_MAX(uint32_t)], + o_audit_loginuid[sizeof("OBJECT_AUDIT_LOGINUID=") + DECIMAL_STR_MAX(uid_t)]; + + uint32_t audit; + uid_t loginuid; +#endif assert(s); assert(iovec); assert(n > 0); - assert(n + N_IOVEC_META_FIELDS <= m); + assert(n + N_IOVEC_META_FIELDS + (object_pid ? N_IOVEC_OBJECT_FIELDS : 0) <= m); if (ucred) { - uint32_t audit; -#ifdef HAVE_LOGIND - uid_t owner; -#endif - realuid = ucred->uid; - if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0) - IOVEC_SET_STRING(iovec[n++], pid); + sprintf(pid, "_PID="PID_FMT, ucred->pid); + IOVEC_SET_STRING(iovec[n++], pid); - if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0) - IOVEC_SET_STRING(iovec[n++], uid); + sprintf(uid, "_UID="UID_FMT, ucred->uid); + IOVEC_SET_STRING(iovec[n++], uid); - if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0) - IOVEC_SET_STRING(iovec[n++], gid); + sprintf(gid, "_GID="GID_FMT, ucred->gid); + IOVEC_SET_STRING(iovec[n++], gid); r = get_process_comm(ucred->pid, &t); if (r >= 0) { - comm = strappend("_COMM=", t); + x = strappenda("_COMM=", t); free(t); - - if (comm) - IOVEC_SET_STRING(iovec[n++], comm); + IOVEC_SET_STRING(iovec[n++], x); } r = get_process_exe(ucred->pid, &t); if (r >= 0) { - exe = strappend("_EXE=", t); + x = strappenda("_EXE=", t); free(t); - - if (exe) - IOVEC_SET_STRING(iovec[n++], exe); + IOVEC_SET_STRING(iovec[n++], x); } - r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t); + r = get_process_cmdline(ucred->pid, 0, false, &t); if (r >= 0) { - cmdline = strappend("_CMDLINE=", t); + x = strappenda("_CMDLINE=", t); free(t); + IOVEC_SET_STRING(iovec[n++], x); + } - if (cmdline) - IOVEC_SET_STRING(iovec[n++], cmdline); + r = get_process_capeff(ucred->pid, &t); + if (r >= 0) { + x = strappenda("_CAP_EFFECTIVE=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); } +#ifdef HAVE_AUDIT r = audit_session_from_pid(ucred->pid, &audit); - if (r >= 0) - if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit) >= 0) - IOVEC_SET_STRING(iovec[n++], audit_session); + if (r >= 0) { + sprintf(audit_session, "_AUDIT_SESSION=%"PRIu32, audit); + IOVEC_SET_STRING(iovec[n++], audit_session); + } r = audit_loginuid_from_pid(ucred->pid, &loginuid); - if (r >= 0) - if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0) - IOVEC_SET_STRING(iovec[n++], audit_loginuid); + if (r >= 0) { + sprintf(audit_loginuid, "_AUDIT_LOGINUID="UID_FMT, loginuid); + IOVEC_SET_STRING(iovec[n++], audit_loginuid); + } +#endif + + r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &c); + if (r >= 0) { + char *session = NULL; + + x = strappenda("_SYSTEMD_CGROUP=", c); + IOVEC_SET_STRING(iovec[n++], x); + + r = cg_path_get_session(c, &t); + if (r >= 0) { + session = strappenda("_SYSTEMD_SESSION=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], session); + } + + if (cg_path_get_owner_uid(c, &owner) >= 0) { + owner_valid = true; - t = shortened_cgroup_path(ucred->pid); - if (t) { - cgroup = strappend("_SYSTEMD_CGROUP=", t); + sprintf(owner_uid, "_SYSTEMD_OWNER_UID="UID_FMT, owner); + IOVEC_SET_STRING(iovec[n++], owner_uid); + } + + if (cg_path_get_unit(c, &t) >= 0) { + x = strappenda("_SYSTEMD_UNIT=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } else if (unit_id && !session) { + x = strappenda("_SYSTEMD_UNIT=", unit_id); + IOVEC_SET_STRING(iovec[n++], x); + } + + if (cg_path_get_user_unit(c, &t) >= 0) { + x = strappenda("_SYSTEMD_USER_UNIT=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } else if (unit_id && session) { + x = strappenda("_SYSTEMD_USER_UNIT=", unit_id); + IOVEC_SET_STRING(iovec[n++], x); + } + + if (cg_path_get_slice(c, &t) >= 0) { + x = strappenda("_SYSTEMD_SLICE=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } + + free(c); + } else if (unit_id) { + x = strappenda("_SYSTEMD_UNIT=", unit_id); + IOVEC_SET_STRING(iovec[n++], x); + } + +#ifdef HAVE_SELINUX + if (use_selinux()) { + if (label) { + x = alloca(strlen("_SELINUX_CONTEXT=") + label_len + 1); + + *((char*) mempcpy(stpcpy(x, "_SELINUX_CONTEXT="), label, label_len)) = 0; + IOVEC_SET_STRING(iovec[n++], x); + } else { + security_context_t con; + + if (getpidcon(ucred->pid, &con) >= 0) { + x = strappenda("_SELINUX_CONTEXT=", con); + + freecon(con); + IOVEC_SET_STRING(iovec[n++], x); + } + } + } +#endif + } + assert(n <= m); + + if (object_pid) { + r = get_process_uid(object_pid, &object_uid); + if (r >= 0) { + sprintf(o_uid, "OBJECT_UID="UID_FMT, object_uid); + IOVEC_SET_STRING(iovec[n++], o_uid); + } + + r = get_process_gid(object_pid, &object_gid); + if (r >= 0) { + sprintf(o_gid, "OBJECT_GID="GID_FMT, object_gid); + IOVEC_SET_STRING(iovec[n++], o_gid); + } + + r = get_process_comm(object_pid, &t); + if (r >= 0) { + x = strappenda("OBJECT_COMM=", t); free(t); + IOVEC_SET_STRING(iovec[n++], x); + } - if (cgroup) - IOVEC_SET_STRING(iovec[n++], cgroup); + r = get_process_exe(object_pid, &t); + if (r >= 0) { + x = strappenda("OBJECT_EXE=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); } -#ifdef HAVE_LOGIND - if (sd_pid_get_session(ucred->pid, &t) >= 0) { - session = strappend("_SYSTEMD_SESSION=", t); + r = get_process_cmdline(object_pid, 0, false, &t); + if (r >= 0) { + x = strappenda("OBJECT_CMDLINE=", t); free(t); + IOVEC_SET_STRING(iovec[n++], x); + } - if (session) - IOVEC_SET_STRING(iovec[n++], session); +#ifdef HAVE_AUDIT + r = audit_session_from_pid(object_pid, &audit); + if (r >= 0) { + sprintf(o_audit_session, "OBJECT_AUDIT_SESSION=%"PRIu32, audit); + IOVEC_SET_STRING(iovec[n++], o_audit_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); + r = audit_loginuid_from_pid(object_pid, &loginuid); + if (r >= 0) { + sprintf(o_audit_loginuid, "OBJECT_AUDIT_LOGINUID="UID_FMT, loginuid); + IOVEC_SET_STRING(iovec[n++], o_audit_loginuid); + } #endif - 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); + r = cg_pid_get_path_shifted(object_pid, s->cgroup_root, &c); + if (r >= 0) { + x = strappenda("OBJECT_SYSTEMD_CGROUP=", c); + IOVEC_SET_STRING(iovec[n++], x); + + r = cg_path_get_session(c, &t); + if (r >= 0) { + x = strappenda("OBJECT_SYSTEMD_SESSION=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } -#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); + if (cg_path_get_owner_uid(c, &owner) >= 0) { + sprintf(o_owner_uid, "OBJECT_SYSTEMD_OWNER_UID="UID_FMT, owner); + IOVEC_SET_STRING(iovec[n++], o_owner_uid); } - } 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); + if (cg_path_get_unit(c, &t) >= 0) { + x = strappenda("OBJECT_SYSTEMD_UNIT=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } - freecon(con); + if (cg_path_get_user_unit(c, &t) >= 0) { + x = strappenda("OBJECT_SYSTEMD_USER_UNIT=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); } + + free(c); } -#endif } + assert(n <= m); if (tv) { - if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu", - (unsigned long long) timeval_load(tv)) >= 0) - IOVEC_SET_STRING(iovec[n++], source_time); + sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu", (unsigned long long) timeval_load(tv)); + IOVEC_SET_STRING(iovec[n++], source_time); } /* Note that strictly speaking storing the boot id here is * redundant since the entry includes this in-line * anyway. However, we need this indexed, too. */ - r = sd_id128_get_boot(&id); - if (r >= 0) - if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0) - IOVEC_SET_STRING(iovec[n++], boot_id); + if (!isempty(s->boot_id_field)) + IOVEC_SET_STRING(iovec[n++], s->boot_id_field); - r = sd_id128_get_machine(&id); - if (r >= 0) - if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0) - IOVEC_SET_STRING(iovec[n++], machine_id); + if (!isempty(s->machine_id_field)) + IOVEC_SET_STRING(iovec[n++], s->machine_id_field); - t = gethostname_malloc(); - if (t) { - hostname = strappend("_HOSTNAME=", t); - free(t); - if (hostname) - IOVEC_SET_STRING(iovec[n++], hostname); - } + if (!isempty(s->hostname_field)) + IOVEC_SET_STRING(iovec[n++], s->hostname_field); assert(n <= m); - write_to_journal(s, - s->split_mode == SPLIT_NONE ? 0 : - (s->split_mode == SPLIT_UID ? realuid : - (realuid == 0 ? 0 : loginuid)), iovec, n); - - free(pid); - free(uid); - free(gid); - free(comm); - free(exe); - free(cmdline); - free(source_time); - free(boot_id); - free(machine_id); - free(hostname); - free(audit_session); - free(audit_loginuid); - free(cgroup); - free(session); - free(owner_uid); - free(unit); - free(selinux_context); + if (s->split_mode == SPLIT_UID && realuid > 0) + /* Split up strictly by any UID */ + journal_uid = realuid; + else if (s->split_mode == SPLIT_LOGIN && realuid > 0 && owner_valid && owner > 0) + /* Split up by login UIDs. We do this only if the + * realuid is not root, in order not to accidentally + * leak privileged information to the user that is + * logged by a privileged process that is part of an + * unprivileged session.*/ + journal_uid = owner; + else + journal_uid = 0; + + write_to_journal(s, journal_uid, iovec, n, priority); } void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) { @@ -705,7 +824,7 @@ void server_driver_message(Server *s, sd_id128_t message_id, const char *format, struct iovec iovec[N_IOVEC_META_FIELDS + 4]; int n = 0; va_list ap; - struct ucred ucred; + struct ucred ucred = {}; assert(s); assert(format); @@ -726,12 +845,11 @@ void server_driver_message(Server *s, sd_id128_t message_id, const char *format, 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); + dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0); } void server_dispatch_message( @@ -741,10 +859,12 @@ void server_dispatch_message( struct timeval *tv, const char *label, size_t label_len, const char *unit_id, - int priority) { + int priority, + pid_t object_pid) { - int rl; - char *path = NULL, *c; + int rl, r; + _cleanup_free_ char *path = NULL; + char *c; assert(s); assert(iovec || n == 0); @@ -755,11 +875,16 @@ void server_dispatch_message( if (LOG_PRI(priority) > s->max_level_store) return; + /* Stop early in case the information will not be stored + * in a journal. */ + if (s->storage == STORAGE_NONE) + return; + if (!ucred) goto finish; - path = shortened_cgroup_path(ucred->pid); - if (!path) + r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &path); + if (r < 0) goto finish; /* example: /user/lennart/3/foobar @@ -778,21 +903,19 @@ void server_dispatch_message( } } - rl = journal_rate_limit_test(s->rate_limit, path, priority & LOG_PRIMASK, available_space(s)); + rl = journal_rate_limit_test(s->rate_limit, path, + priority & LOG_PRIMASK, available_space(s, false)); - if (rl == 0) { - free(path); + if (rl == 0) return; - } /* Write a suppression message if we suppressed something */ if (rl > 1) - server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, "Suppressed %u messages from %s", rl - 1, path); - - free(path); + server_driver_message(s, SD_MESSAGE_JOURNAL_DROPPED, + "Suppressed %u messages from %s", rl - 1, path); finish: - dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id); + dispatch_message_real(s, iovec, n, m, ucred, tv, label, label_len, unit_id, priority, object_pid); } @@ -803,8 +926,10 @@ static int system_journal_open(Server *s) { char ids[33]; r = sd_id128_get_machine(&machine); - if (r < 0) + if (r < 0) { + log_error("Failed to get machine id: %s", strerror(-r)); return r; + } sd_id128_to_string(machine, ids); @@ -821,29 +946,15 @@ static int system_journal_open(Server *s) { if (s->storage == STORAGE_PERSISTENT) (void) mkdir("/var/log/journal/", 0755); - fn = strappend("/var/log/journal/", ids); - if (!fn) - return -ENOMEM; - + fn = strappenda("/var/log/journal/", ids); (void) mkdir(fn, 0755); - free(fn); - - fn = strjoin("/var/log/journal/", ids, "/system.journal", NULL); - if (!fn) - return -ENOMEM; + fn = strappenda(fn, "/system.journal"); r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, s->seal, &s->system_metrics, s->mmap, NULL, &s->system_journal); - free(fn); - - if (r >= 0) { - char fb[FORMAT_BYTES_MAX]; + if (r >= 0) server_fix_perms(s, s->system_journal, 0); - server_driver_message(s, SD_ID128_NULL, "Allowing system journal files to grow to %s.", - format_bytes(fb, sizeof(fb), s->system_metrics.max_use)); - - } else if (r < 0) { - + else if (r < 0) { if (r != -ENOENT && r != -EROFS) log_warning("Failed to open system journal: %s", strerror(-r)); @@ -879,7 +990,10 @@ static int system_journal_open(Server *s) { /* OK, we really need the runtime journal, so create * it if necessary. */ - (void) mkdir_parents(fn, 0755); + (void) mkdir("/run/log", 0755); + (void) mkdir("/run/log/journal", 0755); + (void) mkdir_parents(fn, 0750); + r = journal_file_open_reliably(fn, O_RDWR|O_CREAT, 0640, s->compress, false, &s->runtime_metrics, s->mmap, NULL, &s->runtime_journal); free(fn); @@ -889,22 +1003,22 @@ static int system_journal_open(Server *s) { } } - if (s->runtime_journal) { - char fb[FORMAT_BYTES_MAX]; - + if (s->runtime_journal) server_fix_perms(s, s->runtime_journal, 0); - server_driver_message(s, SD_ID128_NULL, "Allowing runtime journal files to grow to %s.", - format_bytes(fb, sizeof(fb), s->runtime_metrics.max_use)); - } } + available_space(s, true); + return r; } 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); @@ -922,11 +1036,11 @@ int server_flush_to_var(Server *s) { log_debug("Flushing to /var..."); + start = now(CLOCK_MONOTONIC); + r = sd_id128_get_machine(&machine); - if (r < 0) { - log_error("Failed to get machine id: %s", strerror(-r)); + if (r < 0) return r; - } r = sd_journal_open(&j, SD_JOURNAL_RUNTIME_ONLY); if (r < 0) { @@ -943,6 +1057,8 @@ int server_flush_to_var(Server *s) { 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)); @@ -961,6 +1077,12 @@ int server_flush_to_var(Server *s) { server_rotate(s); server_vacuum(s); + if (!s->system_journal) { + log_notice("Didn't flush runtime journal since rotation of system journal wasn't successful."); + r = -EIO; + goto finish; + } + log_debug("Retrying write."); r = journal_file_copy_entry(f, s->system_journal, o, f->current_offset, NULL, NULL, NULL); if (r < 0) { @@ -978,236 +1100,160 @@ finish: if (r >= 0) rm_rf("/run/log/journal", false, true, false); - if (j) - sd_journal_close(j); + 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; } - log_info("Received SIG%s", signal_to_string(sfsi.ssi_signo)); - - 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; - } - - return 0; - - } 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; - - } else if (ev->data.fd == s->native_fd || - ev->data.fd == s->syslog_fd) { + 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."); - 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); - } - } - - if (ev->data.fd == s->syslog_fd) { - char *e; - - if (n > 0 && n_fds == 0) { - e = memchr(s->buffer, '\n', n); - if (e) - *e = 0; - else - 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."); + close_many(fds, n_fds); + } +} - } 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_received_signal(LOG_INFO, si); - log_error("Unknown event."); + sd_event_exit(s->event, 0); return 0; } -static int open_signalfd(Server *s) { +static int setup_signals(Server *s) { sigset_t mask; - struct epoll_event ev; + int r; assert(s); @@ -1215,46 +1261,43 @@ static int open_signalfd(Server *s) { sigset_add_many(&mask, SIGINT, SIGTERM, SIGUSR1, SIGUSR2, -1); assert_se(sigprocmask(SIG_SETMASK, &mask, NULL) == 0); - s->signal_fd = signalfd(-1, &mask, SFD_NONBLOCK|SFD_CLOEXEC); - if (s->signal_fd < 0) { - log_error("signalfd(): %m"); - return -errno; - } + r = sd_event_add_signal(s->event, &s->sigusr1_event_source, SIGUSR1, dispatch_sigusr1, s); + if (r < 0) + return r; - zero(ev); - ev.events = EPOLLIN; - ev.data.fd = s->signal_fd; + r = sd_event_add_signal(s->event, &s->sigusr2_event_source, SIGUSR2, dispatch_sigusr2, s); + if (r < 0) + return r; - if (epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->signal_fd, &ev) < 0) { - log_error("epoll_ctl(): %m"); - return -errno; - } + r = sd_event_add_signal(s->event, &s->sigterm_event_source, SIGTERM, dispatch_sigterm, s); + if (r < 0) + return r; + + r = sd_event_add_signal(s->event, &s->sigint_event_source, SIGINT, dispatch_sigterm, s); + if (r < 0) + return r; return 0; } static int server_parse_proc_cmdline(Server *s) { - char *line, *w, *state; - int r; + _cleanup_free_ char *line = NULL; + const char *w, *state; size_t l; + int r; - if (detect_container(NULL) > 0) - return 0; - - r = read_one_line_file("/proc/cmdline", &line); - if (r < 0) { + r = proc_cmdline(&line); + if (r < 0) log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r)); + if (r <= 0) return 0; - } FOREACH_WORD_QUOTED(w, l, line, state) { - char *word; + _cleanup_free_ char *word; word = strndup(w, l); - if (!word) { - r = -ENOMEM; - goto finish; - } + if (!word) + return -ENOMEM; if (startswith(word, "systemd.journald.forward_to_syslog=")) { r = parse_boolean(word + 35); @@ -1274,43 +1317,130 @@ static int server_parse_proc_cmdline(Server *s) { log_warning("Failed to parse forward to console switch %s. Ignoring.", word + 36); else s->forward_to_console = r; + } else if (startswith(word, "systemd.journald.forward_to_wall=")) { + r = parse_boolean(word + 33); + if (r < 0) + log_warning("Failed to parse forward to wall switch %s. Ignoring.", word + 33); + else + s->forward_to_wall = r; } else if (startswith(word, "systemd.journald")) log_warning("Invalid systemd.journald parameter. Ignoring."); - - free(word); } + /* do not warn about state here, since probably systemd already did */ - r = 0; - -finish: - free(line); - return r; + return 0; } static int server_parse_config_file(Server *s) { - FILE *f; - const char *fn; + assert(s); + + return config_parse(NULL, "/etc/systemd/journald.conf", NULL, + "Journal\0", + config_item_perf_lookup, journald_gperf_lookup, + false, false, true, s); +} + +static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) { + Server *s = userdata; + + assert(s); + + server_sync(s); + return 0; +} + +int server_schedule_sync(Server *s, int priority) { int r; assert(s); - fn = "/etc/systemd/journald.conf"; - f = fopen(fn, "re"); - if (!f) { - if (errno == ENOENT) - return 0; + if (priority <= LOG_CRIT) { + /* Immediately sync to disk when this is of priority CRIT, ALERT, EMERG */ + server_sync(s); + return 0; + } + + if (s->sync_scheduled) + return 0; + + if (s->sync_interval_usec > 0) { + usec_t when; - log_warning("Failed to open configuration file %s: %m", fn); + r = sd_event_now(s->event, CLOCK_MONOTONIC, &when); + if (r < 0) + return r; + + when += s->sync_interval_usec; + + if (!s->sync_event_source) { + r = sd_event_add_time( + s->event, + &s->sync_event_source, + CLOCK_MONOTONIC, + when, 0, + server_dispatch_sync, s); + if (r < 0) + return r; + + r = sd_event_source_set_priority(s->sync_event_source, SD_EVENT_PRIORITY_IMPORTANT); + } else { + r = sd_event_source_set_time(s->sync_event_source, when); + if (r < 0) + return r; + + r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_ONESHOT); + } + if (r < 0) + return r; + + s->sync_scheduled = true; + } + + return 0; +} + +static int dispatch_hostname_change(sd_event_source *es, int fd, uint32_t revents, void *userdata) { + Server *s = userdata; + + assert(s); + + server_cache_hostname(s); + return 0; +} + +static int server_open_hostname(Server *s) { + int r; + + assert(s); + + s->hostname_fd = open("/proc/sys/kernel/hostname", O_RDONLY|O_CLOEXEC|O_NDELAY|O_NOCTTY); + if (s->hostname_fd < 0) { + log_error("Failed to open /proc/sys/kernel/hostname: %m"); return -errno; } - r = 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)); + r = sd_event_add_io(s->event, &s->hostname_event_source, s->hostname_fd, 0, dispatch_hostname_change, s); + if (r < 0) { + /* kernels prior to 3.2 don't support polling this file. Ignore + * the failure. */ + if (r == -EPERM) { + log_warning("Failed to register hostname fd in event loop: %s. Ignoring.", + strerror(-r)); + s->hostname_fd = safe_close(s->hostname_fd); + return 0; + } - fclose(f); + log_error("Failed to register hostname fd in event loop: %s", strerror(-r)); + return r; + } - return r; + r = sd_event_source_set_priority(s->hostname_event_source, SD_EVENT_PRIORITY_IMPORTANT-10); + if (r < 0) { + log_error("Failed to adjust priority of host name event source: %s", strerror(-r)); + return r; + } + + return 0; } int server_init(Server *s) { @@ -1319,29 +1449,40 @@ int server_init(Server *s) { assert(s); zero(*s); - s->syslog_fd = s->native_fd = s->stdout_fd = s->signal_fd = s->epoll_fd = s->dev_kmsg_fd = -1; + s->syslog_fd = s->native_fd = s->stdout_fd = s->dev_kmsg_fd = s->hostname_fd = -1; s->compress = true; s->seal = true; + s->sync_interval_usec = DEFAULT_SYNC_INTERVAL_USEC; + s->sync_scheduled = false; + s->rate_limit_interval = DEFAULT_RATE_LIMIT_INTERVAL; s->rate_limit_burst = DEFAULT_RATE_LIMIT_BURST; - s->forward_to_syslog = true; + s->forward_to_wall = true; + + s->max_file_usec = DEFAULT_MAX_FILE_USEC; s->max_level_store = LOG_DEBUG; s->max_level_syslog = LOG_DEBUG; s->max_level_kmsg = LOG_NOTICE; s->max_level_console = LOG_INFO; + s->max_level_wall = LOG_EMERG; memset(&s->system_metrics, 0xFF, sizeof(s->system_metrics)); memset(&s->runtime_metrics, 0xFF, sizeof(s->runtime_metrics)); server_parse_config_file(s); server_parse_proc_cmdline(s); + if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) { + log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0", + s->rate_limit_interval, s->rate_limit_burst); + s->rate_limit_interval = s->rate_limit_burst = 0; + } mkdir_p("/run/systemd/journal", 0755); - s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func); + s->user_journals = hashmap_new(NULL); if (!s->user_journals) return log_oom(); @@ -1349,12 +1490,14 @@ int server_init(Server *s) { 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)); @@ -1381,7 +1524,8 @@ int server_init(Server *s) { s->stdout_fd = fd; - } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0) { + } else if (sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/dev/log", 0) > 0 || + sd_is_socket_unix(fd, SOCK_DGRAM, -1, "/run/systemd/journal/dev-log", 0) > 0) { if (s->syslog_fd >= 0) { log_error("Too many /dev/log sockets passed."); @@ -1416,7 +1560,11 @@ int server_init(Server *s) { if (r < 0) return r; - r = open_signalfd(s); + r = server_open_hostname(s); + if (r < 0) + return r; + + r = setup_signals(s); if (r < 0) return r; @@ -1428,6 +1576,14 @@ int server_init(Server *s) { if (!s->rate_limit) return -ENOMEM; + r = cg_get_root_path(&s->cgroup_root); + if (r < 0) + return r; + + server_cache_hostname(s); + server_cache_boot_id(s); + server_cache_machine_id(s); + r = system_journal_open(s); if (r < 0) return r; @@ -1469,23 +1625,23 @@ void server_done(Server *s) { hashmap_free(s->user_journals); - if (s->epoll_fd >= 0) - close_nointr_nofail(s->epoll_fd); - - if (s->signal_fd >= 0) - close_nointr_nofail(s->signal_fd); - - if (s->syslog_fd >= 0) - close_nointr_nofail(s->syslog_fd); - - if (s->native_fd >= 0) - close_nointr_nofail(s->native_fd); - - if (s->stdout_fd >= 0) - close_nointr_nofail(s->stdout_fd); - - if (s->dev_kmsg_fd >= 0) - close_nointr_nofail(s->dev_kmsg_fd); + sd_event_source_unref(s->syslog_event_source); + sd_event_source_unref(s->native_event_source); + sd_event_source_unref(s->stdout_event_source); + sd_event_source_unref(s->dev_kmsg_event_source); + sd_event_source_unref(s->sync_event_source); + sd_event_source_unref(s->sigusr1_event_source); + sd_event_source_unref(s->sigusr2_event_source); + sd_event_source_unref(s->sigterm_event_source); + sd_event_source_unref(s->sigint_event_source); + sd_event_source_unref(s->hostname_event_source); + sd_event_unref(s->event); + + safe_close(s->syslog_fd); + safe_close(s->native_fd); + safe_close(s->stdout_fd); + safe_close(s->dev_kmsg_fd); + safe_close(s->hostname_fd); if (s->rate_limit) journal_rate_limit_free(s->rate_limit); @@ -1495,6 +1651,7 @@ void server_done(Server *s) { free(s->buffer); free(s->tty_path); + free(s->cgroup_root); if (s->mmap) mmap_cache_unref(s->mmap);