X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournald-server.c;h=019c3a649acae87f7dfd68292e479e3af701b4ab;hp=de9604024242beae5c0c3384579d1a12bab1b9c8;hb=ea69bd41c5923f4f278a09bb7d8cb1abcfa122e1;hpb=e40ec7aec5e64cd0cfa5fc556d6a9747229b5794 diff --git a/src/journal/journald-server.c b/src/journal/journald-server.c index de9604024..019c3a649 100644 --- a/src/journal/journald-server.c +++ b/src/journal/journald-server.c @@ -27,10 +27,10 @@ #include #include -#include -#include -#include +#include "sd-journal.h" +#include "sd-messages.h" +#include "sd-daemon.h" #include "fileio.h" #include "mkdir.h" #include "hashmap.h" @@ -38,19 +38,20 @@ #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-audit.h" +#include "journald-server.h" #ifdef HAVE_ACL #include @@ -65,12 +66,13 @@ #define USER_JOURNALS_MAX 1024 #define DEFAULT_SYNC_INTERVAL_USEC (5*USEC_PER_MINUTE) -#define DEFAULT_RATE_LIMIT_INTERVAL (10*USEC_PER_SEC) -#define DEFAULT_RATE_LIMIT_BURST 200 +#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", @@ -80,30 +82,31 @@ 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) { +static uint64_t available_space(Server *s, bool verbose) { char ids[33]; _cleanup_free_ char *p = NULL; - const char *f; 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; _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); @@ -134,11 +137,11 @@ static uint64_t available_space(Server *s) { 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; @@ -156,38 +159,37 @@ 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 we reached a high mark, we will always allow this much + * again, unless usage goes above max_use. This watermark + * value is cached so that we don't give up space on pressure, + * but hover below the maximum usage. */ - if (ss_avail < avail) - avail = ss_avail; + if (m->use < sum) + m->use = sum; - s->cached_available_space = avail; - s->cached_available_space_timestamp = ts; - - return avail; -} + avail = LESS_BY(ss_avail, m->keep_free); -static void server_read_file_gid(Server *s) { - const char *g = "systemd-journal"; - int r; - - assert(s); - - 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(&g, &s->file_gid); - if (r < 0) - log_warning("Failed to resolve '%s' group: %s", g, 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) { @@ -200,19 +202,17 @@ 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_errno(r, "Failed to fix access mode on %s, ignoring: %m", f->path); #ifdef HAVE_ACL - if (uid <= 0) + if (uid <= SYSTEM_UID_MAX) return; acl = acl_get_fd(f->fd); if (!acl) { - log_warning("Failed to read ACL on %s, ignoring: %m", f->path); + log_warning_errno(errno, "Failed to read ACL on %s, ignoring: %m", f->path); return; } @@ -222,7 +222,7 @@ void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { if (acl_create_entry(&acl, &entry) < 0 || acl_set_tag_type(entry, ACL_USER) < 0 || acl_set_qualifier(entry, &uid) < 0) { - log_warning("Failed to patch ACL on %s, ignoring: %m", f->path); + log_warning_errno(errno, "Failed to patch ACL on %s, ignoring: %m", f->path); goto finish; } } @@ -232,12 +232,12 @@ void server_fix_perms(Server *s, JournalFile *f, uid_t uid) { if (acl_get_permset(entry, &permset) < 0 || acl_add_perm(permset, ACL_READ) < 0 || calc_acl_mask_if_needed(&acl) < 0) { - log_warning("Failed to patch ACL on %s, ignoring: %m", f->path); + log_warning_errno(errno, "Failed to patch ACL on %s, ignoring: %m", f->path); goto finish; } if (acl_set_fd(f->fd, acl) < 0) - log_warning("Failed to set ACL on %s, ignoring: %m", f->path); + log_warning_errno(errno, "Failed to set ACL on %s, ignoring: %m", f->path); finish: acl_free(acl); @@ -260,24 +260,24 @@ 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); if (r < 0) return s->system_journal; - f = hashmap_get(s->user_journals, UINT32_TO_PTR(uid)); + f = ordered_hashmap_get(s->user_journals, UINT32_TO_PTR(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) { + while (ordered_hashmap_size(s->user_journals) >= USER_JOURNALS_MAX) { /* Too many open? Then let's close one */ - f = hashmap_steal_first(s->user_journals); + f = ordered_hashmap_steal_first(s->user_journals); assert(f); journal_file_close(f); } @@ -288,7 +288,7 @@ static JournalFile* find_journal(Server *s, uid_t uid) { server_fix_perms(s, f, uid); - r = hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f); + r = ordered_hashmap_put(s->user_journals, UINT32_TO_PTR(uid), f); if (r < 0) { journal_file_close(f); return s->system_journal; @@ -297,6 +297,30 @@ 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_errno(r, "Failed to rotate %s: %m", (*f)->path); + else + log_error_errno(r, "Failed to create new %s journal: %m", name); + else + server_fix_perms(s, *f, uid); + return r; +} + void server_rotate(Server *s) { JournalFile *f; void *k; @@ -305,40 +329,16 @@ 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); - } - - if (s->system_journal) { - r = journal_file_rotate(&s->system_journal, s->compress, s->seal); - if (r < 0) - if (s->system_journal) - log_error("Failed to rotate %s: %s", s->system_journal->path, strerror(-r)); - else - log_error("Failed to create new system journal: %s", strerror(-r)); - - else - server_fix_perms(s, s->system_journal, 0); - } + do_rotate(s, &s->runtime_journal, "runtime", false, 0); + do_rotate(s, &s->system_journal, "system", s->seal, 0); - HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { - r = journal_file_rotate(&f, s->compress, s->seal); - if (r < 0) - if (f) - log_error("Failed to rotate %s: %s", f->path, strerror(-r)); - else - log_error("Failed to create user journal: %s", strerror(-r)); - else { - hashmap_replace(s->user_journals, k, f); - server_fix_perms(s, f, PTR_TO_UINT32(k)); - } + ORDERED_HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { + r = do_rotate(s, &f, "user", s->seal, PTR_TO_UINT32(k)); + if (r >= 0) + ordered_hashmap_replace(s->user_journals, k, f); + else if (!f) + /* Old file has been closed and deallocated */ + ordered_hashmap_remove(s->user_journals, k); } } @@ -348,27 +348,46 @@ void server_sync(Server *s) { Iterator i; int r; - static const struct itimerspec sync_timer_disable = {}; - if (s->system_journal) { r = journal_file_set_offline(s->system_journal); if (r < 0) - log_error("Failed to sync system journal: %s", strerror(-r)); + log_error_errno(r, "Failed to sync system journal: %m"); } - HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { + ORDERED_HASHMAP_FOREACH_KEY(f, k, s->user_journals, i) { r = journal_file_set_offline(f); if (r < 0) - log_error("Failed to sync user journal: %s", strerror(-r)); + log_error_errno(r, "Failed to sync user journal: %m"); } - r = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_disable, NULL); - if (r < 0) - log_error("Failed to disable max timer: %m"); + if (s->sync_event_source) { + r = sd_event_source_set_enabled(s->sync_event_source, SD_EVENT_OFF); + if (r < 0) + log_error_errno(r, "Failed to disable sync timer source: %m"); + } s->sync_scheduled = false; } +static void do_vacuum( + Server *s, + const char *id, + JournalFile *f, + const char* path, + JournalMetrics *metrics) { + + const char *p; + int r; + + if (!f) + return; + + p = strappenda(path, id); + r = journal_directory_vacuum(p, metrics->max_use, s->max_retention_usec, &s->oldest_file_usec, false); + if (r < 0 && r != -ENOENT) + log_error_errno(r, "Failed to vacuum %s: %m", p); +} + void server_vacuum(Server *s) { char ids[33]; sd_id128_t machine; @@ -380,37 +399,68 @@ void server_vacuum(Server *s) { r = sd_id128_get_machine(&machine); if (r < 0) { - log_error("Failed to get machine ID: %s", strerror(-r)); + log_error_errno(r, "Failed to get machine ID: %m"); return; } - sd_id128_to_string(machine, ids); - if (s->system_journal) { - char *p = strappenda("/var/log/journal/", ids); + do_vacuum(s, ids, s->system_journal, "/var/log/journal/", &s->system_metrics); + do_vacuum(s, ids, s->runtime_journal, "/run/log/journal/", &s->runtime_metrics); - r = journal_directory_vacuum(p, s->system_metrics.max_use, s->system_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec); - if (r < 0 && r != -ENOENT) - log_error("Failed to vacuum %s: %s", p, strerror(-r)); - } + s->cached_available_space_timestamp = 0; +} - if (s->runtime_journal) { - char *p = strappenda("/run/log/journal/", ids); +static void server_cache_machine_id(Server *s) { + sd_id128_t id; + int r; - r = journal_directory_vacuum(p, s->runtime_metrics.max_use, s->runtime_metrics.keep_free, s->max_retention_usec, &s->oldest_file_usec); - if (r < 0 && r != -ENOENT) - log_error("Failed to vacuum %s: %s", p, strerror(-r)); - } + assert(s); - s->cached_available_space_timestamp = 0; + r = sd_id128_get_machine(&id); + if (r < 0) + return; + + sd_id128_to_string(id, stpcpy(s->machine_id_field, "_MACHINE_ID=")); +} + +static void server_cache_boot_id(Server *s) { + sd_id128_t id; + int r; + + assert(s); + + r = sd_id128_get_boot(&id); + if (r < 0) + return; + + sd_id128_to_string(id, stpcpy(s->boot_id_field, "_BOOT_ID=")); } -bool shall_try_append_again(JournalFile *f, int r) { +static void server_cache_hostname(Server *s) { + _cleanup_free_ char *t = NULL; + char *x; + + assert(s); + + t = gethostname_malloc(); + if (!t) + return; + + x = strappend("_HOSTNAME=", t); + if (!x) + return; + + free(s->hostname_field); + s->hostname_field = x; +} + +static bool shall_try_append_again(JournalFile *f, int r) { /* -E2BIG Hit configured limit -EFBIG Hit fs limit -EDQUOT Quota limit hit -ENOSPC Disk full + -EIO I/O error of some kind (mmap) -EHOSTDOWN Other machine -EBUSY Unclean shutdown -EPROTONOSUPPORT Unsupported feature @@ -428,13 +478,15 @@ bool shall_try_append_again(JournalFile *f, int r) { log_info("%s: Unsupported feature, rotating.", f->path); else if (r == -EBADMSG || r == -ENODATA || r == ESHUTDOWN) log_warning("%s: Journal file corrupted, rotating.", f->path); + else if (r == -EIO) + log_warning("%s: IO error, rotating.", f->path); else return false; 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; @@ -460,7 +512,7 @@ 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) { - server_schedule_sync(s); + server_schedule_sync(s, priority); return; } @@ -470,7 +522,7 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned 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)); + log_error_errno(r, "Failed to write entry (%d items, %zu bytes), ignoring: %m", n, size); return; } @@ -489,34 +541,41 @@ static void write_to_journal(Server *s, uid_t uid, struct iovec *iovec, unsigned 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)); - } + log_error_errno(r, "Failed to write entry (%d items, %zu bytes) despite vacuuming, ignoring: %m", n, size); + } else + server_schedule_sync(s, priority); } static void dispatch_message_real( Server *s, struct iovec *iovec, unsigned n, unsigned m, - struct ucred *ucred, - struct timeval *tv, + const struct ucred *ucred, + const struct timeval *tv, const char *label, size_t label_len, - const char *unit_id) { + const char *unit_id, + int priority, + pid_t object_pid) { - char pid[sizeof("_PID=") + DECIMAL_STR_MAX(pid_t)], + 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)], - boot_id[sizeof("_BOOT_ID=") + 32] = "_BOOT_ID=", - machine_id[sizeof("_MACHINE_ID=") + 32] = "_MACHINE_ID="; - char *comm, *exe, *cmdline, *cgroup, *session, *unit, *hostname; - sd_id128_t id; + o_uid[sizeof("OBJECT_UID=") + DECIMAL_STR_MAX(uid_t)], + o_gid[sizeof("OBJECT_GID=") + DECIMAL_STR_MAX(gid_t)], + o_owner_uid[sizeof("OBJECT_SYSTEMD_OWNER_UID=") + DECIMAL_STR_MAX(uid_t)]; + uid_t object_uid; + gid_t object_gid; + char *x; int r; 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)]; + 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; @@ -525,59 +584,68 @@ static void dispatch_message_real( 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) { realuid = ucred->uid; - sprintf(pid, "_PID=%lu", (unsigned long) ucred->pid); + sprintf(pid, "_PID="PID_FMT, ucred->pid); IOVEC_SET_STRING(iovec[n++], pid); - sprintf(uid, "_UID=%lu", (unsigned long) ucred->uid); + sprintf(uid, "_UID="UID_FMT, ucred->uid); IOVEC_SET_STRING(iovec[n++], uid); - sprintf(gid, "_GID=%lu", (unsigned long) ucred->gid); + sprintf(gid, "_GID="GID_FMT, ucred->gid); IOVEC_SET_STRING(iovec[n++], gid); r = get_process_comm(ucred->pid, &t); if (r >= 0) { - comm = strappenda("_COMM=", t); + x = strappenda("_COMM=", t); free(t); - IOVEC_SET_STRING(iovec[n++], comm); + IOVEC_SET_STRING(iovec[n++], x); } r = get_process_exe(ucred->pid, &t); if (r >= 0) { - exe = strappenda("_EXE=", t); + x = strappenda("_EXE=", t); free(t); - IOVEC_SET_STRING(iovec[n++], exe); + IOVEC_SET_STRING(iovec[n++], x); } r = get_process_cmdline(ucred->pid, 0, false, &t); if (r >= 0) { - cmdline = strappenda("_CMDLINE=", t); + x = strappenda("_CMDLINE=", t); free(t); - IOVEC_SET_STRING(iovec[n++], cmdline); + IOVEC_SET_STRING(iovec[n++], x); + } + + 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) { - sprintf(audit_session, "_AUDIT_SESSION=%lu", (unsigned long) audit); + sprintf(audit_session, "_AUDIT_SESSION=%"PRIu32, audit); IOVEC_SET_STRING(iovec[n++], audit_session); } r = audit_loginuid_from_pid(ucred->pid, &loginuid); if (r >= 0) { - sprintf(audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid); + sprintf(audit_loginuid, "_AUDIT_LOGINUID="UID_FMT, loginuid); IOVEC_SET_STRING(iovec[n++], audit_loginuid); } #endif - r = cg_pid_get_path_shifted(ucred->pid, NULL, &c); + r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &c); if (r >= 0) { - cgroup = strappenda("_SYSTEMD_CGROUP=", c); - IOVEC_SET_STRING(iovec[n++], cgroup); + char *session = NULL; + + x = strappenda("_SYSTEMD_CGROUP=", c); + IOVEC_SET_STRING(iovec[n++], x); r = cg_path_get_session(c, &t); if (r >= 0) { @@ -589,48 +657,143 @@ static void dispatch_message_real( if (cg_path_get_owner_uid(c, &owner) >= 0) { owner_valid = true; - sprintf(owner_uid, "_SYSTEMD_OWNER_UID=%lu", (unsigned long) owner); + sprintf(owner_uid, "_SYSTEMD_OWNER_UID="UID_FMT, owner); IOVEC_SET_STRING(iovec[n++], owner_uid); } if (cg_path_get_unit(c, &t) >= 0) { - unit = strappenda("_SYSTEMD_UNIT=", t); + x = strappenda("_SYSTEMD_UNIT=", t); free(t); - } else if (cg_path_get_user_unit(c, &t) >= 0) { - unit = strappenda("_SYSTEMD_USER_UNIT=", 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); - } else if (unit_id) { - if (session) - unit = strappenda("_SYSTEMD_USER_UNIT=", unit_id); - else - unit = strappenda("_SYSTEMD_UNIT=", unit_id); - } else - unit = NULL; + 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 (unit) - IOVEC_SET_STRING(iovec[n++], unit); + 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 (label) { - char *selinux_context = alloca(sizeof("_SELINUX_CONTEXT=") + label_len); + if (mac_selinux_use()) { + if (label) { + x = alloca(strlen("_SELINUX_CONTEXT=") + label_len + 1); - *((char*) mempcpy(stpcpy(selinux_context, "_SELINUX_CONTEXT="), label, label_len)) = 0; - IOVEC_SET_STRING(iovec[n++], selinux_context); - } else { - security_context_t con; + *((char*) mempcpy(stpcpy(x, "_SELINUX_CONTEXT="), label, label_len)) = 0; + IOVEC_SET_STRING(iovec[n++], x); + } else { + security_context_t con; - if (getpidcon(ucred->pid, &con) >= 0) { - char *selinux_context = strappenda("_SELINUX_CONTEXT=", con); + if (getpidcon(ucred->pid, &con) >= 0) { + x = strappenda("_SELINUX_CONTEXT=", con); - freecon(con); - IOVEC_SET_STRING(iovec[n++], selinux_context); + 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); + } + + r = get_process_exe(object_pid, &t); + if (r >= 0) { + x = strappenda("OBJECT_EXE=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } + + 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); + } + +#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); + } + + 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 + + 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); + } + + 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); + } + + if (cg_path_get_unit(c, &t) >= 0) { + x = strappenda("OBJECT_SYSTEMD_UNIT=", t); + free(t); + IOVEC_SET_STRING(iovec[n++], x); + } + + 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); + } + } + assert(n <= m); if (tv) { sprintf(source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu", (unsigned long long) timeval_load(tv)); @@ -640,24 +803,14 @@ static void dispatch_message_real( /* Note that strictly speaking storing the boot id here is * redundant since the entry includes this in-line * anyway. However, we need this indexed, too. */ - r = sd_id128_get_boot(&id); - if (r >= 0) { - sd_id128_to_string(id, boot_id + sizeof("_BOOT_ID=") - 1); - IOVEC_SET_STRING(iovec[n++], boot_id); - } + if (!isempty(s->boot_id_field)) + IOVEC_SET_STRING(iovec[n++], s->boot_id_field); - r = sd_id128_get_machine(&id); - if (r >= 0) { - sd_id128_to_string(id, machine_id + sizeof("_MACHINE_ID=") - 1); - 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 = strappenda("_HOSTNAME=", t); - free(t); - IOVEC_SET_STRING(iovec[n++], hostname); - } + if (!isempty(s->hostname_field)) + IOVEC_SET_STRING(iovec[n++], s->hostname_field); assert(n <= m); @@ -665,17 +818,16 @@ static void dispatch_message_real( /* Split up strictly by any UID */ journal_uid = realuid; else if (s->split_mode == SPLIT_LOGIN && realuid > 0 && owner_valid && owner > 0) - /* Split up by login UIDs, this avoids creation of - * individual journals for system UIDs. We do this - * only if the realuid is not root, in order not to - * accidentally leak privileged information to the - * user that is logged by a privileged process that is - * part of an unprivileged session.*/ + /* Split up by login UIDs. We do this only if the + * realuid is not root, in order not to accidentally + * leak privileged information to the user that is + * logged by a privileged process that is part of an + * unprivileged session. */ journal_uid = owner; else journal_uid = 0; - write_to_journal(s, journal_uid, iovec, n); + write_to_journal(s, journal_uid, iovec, n, priority); } void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ...) { @@ -700,7 +852,7 @@ void server_driver_message(Server *s, sd_id128_t message_id, const char *format, IOVEC_SET_STRING(iovec[n++], buffer); if (!sd_id128_equal(message_id, SD_ID128_NULL)) { - snprintf(mid, sizeof(mid), MESSAGE_ID(message_id)); + snprintf(mid, sizeof(mid), LOG_MESSAGE_ID(message_id)); char_array_0(mid); IOVEC_SET_STRING(iovec[n++], mid); } @@ -709,17 +861,18 @@ void server_driver_message(Server *s, sd_id128_t message_id, const char *format, ucred.uid = getuid(); ucred.gid = getgid(); - dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL); + dispatch_message_real(s, iovec, n, ELEMENTSOF(iovec), &ucred, NULL, NULL, 0, NULL, LOG_INFO, 0); } void server_dispatch_message( Server *s, struct iovec *iovec, unsigned n, unsigned m, - struct ucred *ucred, - struct timeval *tv, + const struct ucred *ucred, + const struct timeval *tv, const char *label, size_t label_len, const char *unit_id, - int priority) { + int priority, + pid_t object_pid) { int rl, r; _cleanup_free_ char *path = NULL; @@ -734,10 +887,15 @@ 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; - r = cg_pid_get_path_shifted(ucred->pid, NULL, &path); + r = cg_pid_get_path_shifted(ucred->pid, s->cgroup_root, &path); if (r < 0) goto finish; @@ -758,7 +916,7 @@ void server_dispatch_message( } rl = journal_rate_limit_test(s->rate_limit, path, - priority & LOG_PRIMASK, available_space(s)); + priority & LOG_PRIMASK, available_space(s, false)); if (rl == 0) return; @@ -769,11 +927,11 @@ void server_dispatch_message( "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); } -static int system_journal_open(Server *s) { +static int system_journal_open(Server *s, bool flush_requested) { int r; char *fn; sd_id128_t machine; @@ -781,13 +939,14 @@ static int system_journal_open(Server *s) { r = sd_id128_get_machine(&machine); if (r < 0) - return r; + return log_error_errno(r, "Failed to get machine id: %m"); 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) { + (flush_requested + || access("/run/systemd/journal/flushed", F_OK) >= 0)) { /* If in auto mode: first try to create the machine * path, but not the prefix. @@ -804,25 +963,11 @@ static int system_journal_open(Server *s) { 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); - if (r >= 0) { - char fb[FORMAT_BYTES_MAX]; - uint64_t avail; - + 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)); - - avail = available_space(s); - - if (s->system_metrics.max_use > avail) - server_driver_message(s, SD_ID128_NULL, "Journal size currently limited to %s due to SystemKeepFree.", - format_bytes(fb, sizeof(fb), avail)); - - } else if (r < 0) { - + else if (r < 0) { if (r != -ENOENT && r != -EROFS) - log_warning("Failed to open system journal: %s", strerror(-r)); + log_warning_errno(r, "Failed to open system journal: %m"); r = 0; } @@ -846,7 +991,7 @@ static int system_journal_open(Server *s) { if (r < 0) { if (r != -ENOENT) - log_warning("Failed to open runtime journal: %s", strerror(-r)); + log_warning_errno(r, "Failed to open runtime journal: %m"); r = 0; } @@ -856,39 +1001,33 @@ 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); - if (r < 0) { - log_error("Failed to open runtime journal: %s", strerror(-r)); - return r; - } + if (r < 0) + return log_error_errno(r, "Failed to open runtime journal: %m"); } - if (s->runtime_journal) { - char fb[FORMAT_BYTES_MAX]; - uint64_t avail; - + 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)); - - avail = available_space(s); - - if (s->system_metrics.max_use > avail) - server_driver_message(s, SD_ID128_NULL, "Journal size currently limited to %s due to RuntimeKeepFree.", - format_bytes(fb, sizeof(fb), avail)); - } } + 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); @@ -899,24 +1038,22 @@ int server_flush_to_var(Server *s) { if (!s->runtime_journal) return 0; - system_journal_open(s); + system_journal_open(s, true); if (!s->system_journal) return 0; 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) { - log_error("Failed to read runtime journal: %s", strerror(-r)); - return r; - } + if (r < 0) + return log_error_errno(r, "Failed to read runtime journal: %m"); sd_journal_set_data_threshold(j, 0); @@ -927,9 +1064,11 @@ 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)); + log_error_errno(r, "Can't read entry: %m"); goto finish; } @@ -938,7 +1077,7 @@ int server_flush_to_var(Server *s) { continue; if (!shall_try_append_again(s->system_journal, r)) { - log_error("Can't write entry: %s", strerror(-r)); + log_error_errno(r, "Can't write entry: %m"); goto finish; } @@ -954,7 +1093,7 @@ int server_flush_to_var(Server *s) { log_debug("Retrying write."); 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)); + log_error_errno(r, "Can't write entry: %m"); goto finish; } } @@ -970,247 +1109,179 @@ finish: 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 server_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 || fd == s->audit_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 = {}; + union sockaddr_union sa = {}; + struct msghdr msghdr = { + .msg_iov = &iovec, + .msg_iovlen = 1, + .msg_control = &control, + .msg_controllen = sizeof(control), + .msg_name = &sa, + .msg_namelen = sizeof(sa), + }; - if (ev->data.fd == s->signal_fd) { - struct signalfd_siginfo sfsi; ssize_t n; + int *fds = NULL; + unsigned n_fds = 0; + int v = 0; + size_t m; - if (ev->events != EPOLLIN) { - log_error("Got invalid event from epoll."); - return -EIO; - } + /* Try to get the right size, if we can. (Not all + * sockets support SIOCINQ, hence we just try, but + * don't rely on it. */ + (void) ioctl(fd, SIOCINQ, &v); + + /* Fix it up, if it is too small. We use the same fixed value as auditd here. Awful! */ + m = PAGE_ALIGN(MAX3((size_t) v + 1, + (size_t) LINE_MAX, + ALIGN(sizeof(struct nlmsghdr)) + ALIGN((size_t) MAX_AUDIT_MESSAGE_LENGTH)) + 1); - n = read(s->signal_fd, &sfsi, sizeof(sfsi)); - if (n != sizeof(sfsi)) { + if (!GREEDY_REALLOC(s->buffer, s->buffer_size, m)) + return log_oom(); - if (n >= 0) - return -EIO; + iovec.iov_base = s->buffer; + iovec.iov_len = s->buffer_size - 1; /* Leave room for trailing NUL we add later */ + n = recvmsg(fd, &msghdr, MSG_DONTWAIT|MSG_CMSG_CLOEXEC); + if (n < 0) { if (errno == EINTR || errno == EAGAIN) - return 1; + return 0; + log_error_errno(errno, "recvmsg() failed: %m"); return -errno; } - if (sfsi.ssi_signo == SIGUSR1) { - touch("/run/systemd/journal/flushed"); - server_flush_to_var(s); - server_sync(s); - return 1; - } - - if (sfsi.ssi_signo == SIGUSR2) { - server_rotate(s); - server_vacuum(s); - return 1; + 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); + } } - log_info("Received SIG%s", signal_to_string(sfsi.ssi_signo)); - - return 0; - - } else if (ev->data.fd == s->sync_timer_fd) { - int r; - uint64_t t; + /* And a trailing NUL, just in case */ + s->buffer[n] = 0; - log_debug("Got sync request from epoll."); + if (fd == s->syslog_fd) { + if (n > 0 && n_fds == 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."); - r = read(ev->data.fd, (void *)&t, sizeof(t)); - if (r < 0) - return 0; - - server_sync(s); - return 1; - - } else if (ev->data.fd == s->dev_kmsg_fd) { - int r; - - if (ev->events != EPOLLIN) { - log_error("Got invalid event from epoll."); - return -EIO; - } + } else if (fd == s->native_fd) { + 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."); - 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) { + } else { + assert(fd == s->audit_fd); - if (ev->events != EPOLLIN) { - log_error("Got invalid event from epoll."); - return -EIO; + if (n > 0 && n_fds == 0) + server_process_audit_message(s, s->buffer, n, ucred, &sa, msghdr.msg_namelen); + else if (n_fds > 0) + log_warning("Got file descriptors via audit 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; + close_many(fds, n_fds); + } +} - if (n > 0 && n_fds == 0) { - e = memchr(s->buffer, '\n', n); - if (e) - *e = 0; - else - s->buffer[n] = 0; +static int dispatch_sigusr1(sd_event_source *es, const struct signalfd_siginfo *si, void *userdata) { + Server *s = userdata; - 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."); + assert(s); - } 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."); - } + log_info("Received request to flush runtime journal from PID %"PRIu32, si->ssi_pid); - close_many(fds, n_fds); - } - - return 1; + server_flush_to_var(s); + server_sync(s); + server_vacuum(s); - } else if (ev->data.fd == s->stdout_fd) { + touch("/run/systemd/journal/flushed"); - 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); @@ -1218,36 +1289,34 @@ 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) { _cleanup_free_ char *line = NULL; - char *w, *state; - int r; + const char *w, *state; size_t l; + int r; - if (detect_container(NULL) > 0) - return 0; - - r = read_one_line_file("/proc/cmdline", &line); + r = proc_cmdline(&line); if (r < 0) { - log_warning("Failed to read /proc/cmdline, ignoring: %s", strerror(-r)); + log_warning_errno(r, "Failed to read /proc/cmdline, ignoring: %m"); return 0; } @@ -1276,79 +1345,124 @@ 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."); } + /* do not warn about state here, since probably systemd already did */ return 0; } static int server_parse_config_file(Server *s) { - static const char fn[] = "/etc/systemd/journald.conf"; - _cleanup_fclose_ FILE *f = NULL; - int r; - assert(s); - f = fopen(fn, "re"); - if (!f) { - if (errno == ENOENT) - return 0; + return config_parse_many("/etc/systemd/journald.conf", + CONF_DIRS_NULSTR("systemd/journald.conf"), + "Journal\0", + config_item_perf_lookup, journald_gperf_lookup, + false, s); +} - log_warning("Failed to open configuration file %s: %m", fn); - return -errno; - } +static int server_dispatch_sync(sd_event_source *es, usec_t t, void *userdata) { + Server *s = userdata; - r = config_parse(NULL, fn, f, "Journal\0", config_item_perf_lookup, - (void*) journald_gperf_lookup, false, false, s); - if (r < 0) - log_warning("Failed to parse configuration file: %s", strerror(-r)); + assert(s); - return r; + server_sync(s); + return 0; } -static int server_open_sync_timer(Server *s) { +int server_schedule_sync(Server *s, int priority) { int r; - struct epoll_event ev; assert(s); - s->sync_timer_fd = timerfd_create(CLOCK_MONOTONIC, TFD_CLOEXEC); - if (s->sync_timer_fd < 0) - return -errno; + if (priority <= LOG_CRIT) { + /* Immediately sync to disk when this is of priority CRIT, ALERT, EMERG */ + server_sync(s); + return 0; + } - zero(ev); - ev.events = EPOLLIN; - ev.data.fd = s->sync_timer_fd; + if (s->sync_scheduled) + return 0; - r = epoll_ctl(s->epoll_fd, EPOLL_CTL_ADD, s->sync_timer_fd, &ev); - if (r < 0) { - log_error("Failed to add idle timer fd to epoll object: %m"); - return -errno; + if (s->sync_interval_usec > 0) { + usec_t when; + + r = sd_event_now(s->event, CLOCK_MONOTONIC, &when); + if (r < 0) + return r; + + when += s->sync_interval_usec; + + 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; } -int server_schedule_sync(Server *s) { - int r; +static int dispatch_hostname_change(sd_event_source *es, int fd, uint32_t revents, void *userdata) { + Server *s = userdata; assert(s); - if (s->sync_scheduled) - return 0; + server_cache_hostname(s); + return 0; +} - if (s->sync_interval_usec) { - struct itimerspec sync_timer_enable = {}; +static int server_open_hostname(Server *s) { + int r; - timespec_store(&sync_timer_enable.it_value, s->sync_interval_usec); + assert(s); - r = timerfd_settime(s->sync_timer_fd, 0, &sync_timer_enable, NULL); - if (r < 0) - return -errno; + s->hostname_fd = open("/proc/sys/kernel/hostname", O_RDONLY|O_CLOEXEC|O_NDELAY|O_NOCTTY); + if (s->hostname_fd < 0) + return log_error_errno(errno, "Failed to open /proc/sys/kernel/hostname: %m"); + + 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; + } + + return log_error_errno(r, "Failed to register hostname fd in event loop: %m"); } - s->sync_scheduled = true; + r = sd_event_source_set_priority(s->hostname_event_source, SD_EVENT_PRIORITY_IMPORTANT-10); + if (r < 0) + return log_error_errno(r, "Failed to adjust priority of host name event source: %m"); return 0; } @@ -1359,8 +1473,7 @@ int server_init(Server *s) { assert(s); zero(*s); - s->sync_timer_fd = s->syslog_fd = s->native_fd = s->stdout_fd = - s->signal_fd = s->epoll_fd = s->dev_kmsg_fd = -1; + s->syslog_fd = s->native_fd = s->stdout_fd = s->dev_kmsg_fd = s->audit_fd = s->hostname_fd = -1; s->compress = true; s->seal = true; @@ -1370,12 +1483,15 @@ int server_init(Server *s) { 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)); @@ -1383,15 +1499,14 @@ int server_init(Server *s) { server_parse_config_file(s); server_parse_proc_cmdline(s); if (!!s->rate_limit_interval ^ !!s->rate_limit_burst) { - log_debug("Setting both rate limit interval and burst from %llu,%u to 0,0", - (long long unsigned) s->rate_limit_interval, - s->rate_limit_burst); + log_debug("Setting both rate limit interval and burst from "USEC_FMT",%u to 0,0", + s->rate_limit_interval, s->rate_limit_burst); s->rate_limit_interval = s->rate_limit_burst = 0; } mkdir_p("/run/systemd/journal", 0755); - s->user_journals = hashmap_new(trivial_hash_func, trivial_compare_func); + s->user_journals = ordered_hashmap_new(NULL); if (!s->user_journals) return log_oom(); @@ -1399,17 +1514,15 @@ 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) + return log_error_errno(r, "Failed to create event loop: %m"); + + 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)); - return n; - } + if (n < 0) + return log_error_errno(n, "Failed to read listening file descriptors from environment: %m"); for (fd = SD_LISTEN_FDS_START; fd < SD_LISTEN_FDS_START + n; fd++) { @@ -1431,7 +1544,8 @@ 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."); @@ -1440,9 +1554,24 @@ int server_init(Server *s) { s->syslog_fd = fd; + } else if (sd_is_socket(fd, AF_NETLINK, SOCK_RAW, -1) > 0) { + + if (s->audit_fd >= 0) { + log_error("Too many audit sockets passed."); + return -EINVAL; + } + + s->audit_fd = fd; + } else { - log_error("Unknown socket passed."); - return -EINVAL; + log_warning("Unknown socket passed as file descriptor %d, ignoring.", fd); + + /* Let's close the fd, better be safe than + sorry. The fd might reference some resource + that we really want to release if we don't + make use of it. */ + + safe_close(fd); } } @@ -1462,15 +1591,19 @@ int server_init(Server *s) { if (r < 0) return r; + r = server_open_audit(s); + if (r < 0) + return r; + r = server_open_kernel_seqnum(s); if (r < 0) return r; - r = server_open_sync_timer(s); + r = server_open_hostname(s); if (r < 0) return r; - r = open_signalfd(s); + r = setup_signals(s); if (r < 0) return r; @@ -1478,12 +1611,19 @@ int server_init(Server *s) { if (!s->udev) return -ENOMEM; - s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, - s->rate_limit_burst); + s->rate_limit = journal_rate_limit_new(s->rate_limit_interval, s->rate_limit_burst); if (!s->rate_limit) return -ENOMEM; - r = system_journal_open(s); + 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, false); if (r < 0) return r; @@ -1501,7 +1641,7 @@ void server_maybe_append_tags(Server *s) { if (s->system_journal) journal_file_maybe_append_tag(s->system_journal, n); - HASHMAP_FOREACH(f, s->user_journals, i) + ORDERED_HASHMAP_FOREACH(f, s->user_journals, i) journal_file_maybe_append_tag(f, n); #endif } @@ -1519,31 +1659,30 @@ void server_done(Server *s) { if (s->runtime_journal) journal_file_close(s->runtime_journal); - while ((f = hashmap_steal_first(s->user_journals))) + while ((f = ordered_hashmap_steal_first(s->user_journals))) journal_file_close(f); - hashmap_free(s->user_journals); - - if (s->epoll_fd >= 0) - close_nointr_nofail(s->epoll_fd); - - if (s->signal_fd >= 0) - close_nointr_nofail(s->signal_fd); - - if (s->syslog_fd >= 0) - close_nointr_nofail(s->syslog_fd); - - if (s->native_fd >= 0) - close_nointr_nofail(s->native_fd); - - if (s->stdout_fd >= 0) - close_nointr_nofail(s->stdout_fd); - - if (s->dev_kmsg_fd >= 0) - close_nointr_nofail(s->dev_kmsg_fd); - - if (s->sync_timer_fd >= 0) - close_nointr_nofail(s->sync_timer_fd); + ordered_hashmap_free(s->user_journals); + + 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->audit_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->audit_fd); + safe_close(s->hostname_fd); if (s->rate_limit) journal_rate_limit_free(s->rate_limit); @@ -1553,6 +1692,8 @@ void server_done(Server *s) { free(s->buffer); free(s->tty_path); + free(s->cgroup_root); + free(s->hostname_field); if (s->mmap) mmap_cache_unref(s->mmap);