+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
+
+ return generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ p,
+ test_object_offset,
+ direction,
+ ret, offset, NULL);
+}
+
+int journal_file_move_to_entry_by_monotonic_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ sd_id128_t boot_id,
+ uint64_t monotonic,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ char t[9+32+1] = "_BOOT_ID=";
+ Object *o, *d;
+ int r;
+ uint64_t b, z;
+
+ assert(f);
+
+ /* First, seek by time */
+ sd_id128_to_string(boot_id, t + 9);
+ r = journal_file_find_data_object(f, t, strlen(t), &o, &b);
+ if (r < 0)
+ return r;
+ if (r == 0)
+ return -ENOENT;
+
+ r = generic_array_bisect_plus_one(f,
+ le64toh(o->data.entry_offset),
+ le64toh(o->data.entry_array_offset),
+ le64toh(o->data.n_entries),
+ monotonic,
+ test_object_monotonic,
+ direction,
+ NULL, &z, NULL);
+ if (r <= 0)
+ return r;
+
+ /* And now, continue seeking until we find an entry that
+ * exists in both bisection arrays */
+
+ for (;;) {
+ Object *qo;
+ uint64_t p, q;
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
+
+ r = generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ z,
+ test_object_offset,
+ direction,
+ NULL, &p, NULL);
+ if (r <= 0)
+ return r;
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, b, &o);
+ if (r < 0)
+ return r;
+
+ r = generic_array_bisect_plus_one(f,
+ le64toh(o->data.entry_offset),
+ le64toh(o->data.entry_array_offset),
+ le64toh(o->data.n_entries),
+ p,
+ test_object_offset,
+ direction,
+ &qo, &q, NULL);
+
+ if (r <= 0)
+ return r;
+
+ if (p == q) {
+ if (ret)
+ *ret = qo;
+ if (offset)
+ *offset = q;
+
+ return 1;
+ }
+
+ z = q;
+ }
+
+ return 0;
+}
+
+int journal_file_move_to_entry_by_seqnum_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ uint64_t seqnum,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ Object *d;
+ int r;
+
+ assert(f);
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
+
+ return generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ seqnum,
+ test_object_seqnum,
+ direction,
+ ret, offset, NULL);
+}
+
+int journal_file_move_to_entry_by_realtime_for_data(
+ JournalFile *f,
+ uint64_t data_offset,
+ uint64_t realtime,
+ direction_t direction,
+ Object **ret, uint64_t *offset) {
+
+ Object *d;
+ int r;
+
+ assert(f);
+
+ r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+ if (r < 0)
+ return r;
+
+ return generic_array_bisect_plus_one(f,
+ le64toh(d->data.entry_offset),
+ le64toh(d->data.entry_array_offset),
+ le64toh(d->data.n_entries),
+ realtime,
+ test_object_realtime,
+ direction,
+ ret, offset, NULL);
+}
+
+static void *fsprg_state(JournalFile *f) {
+ uint64_t a, b;
+ assert(f);
+
+ if (!f->authenticate)
+ return NULL;
+
+ a = le64toh(f->fsprg_header->header_size);
+ b = le64toh(f->fsprg_header->state_size);
+
+ if (a + b > f->fsprg_size)
+ return NULL;
+
+ return (uint8_t*) f->fsprg_header + a;
+}
+
+static uint64_t journal_file_tag_seqnum(JournalFile *f) {
+ uint64_t r;
+
+ assert(f);
+
+ r = le64toh(f->header->n_tags) + 1;
+ f->header->n_tags = htole64(r);
+
+ return r;
+}
+
+int journal_file_append_tag(JournalFile *f) {
+ Object *o;
+ uint64_t p;
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ if (!f->hmac_running)
+ return 0;
+
+ log_debug("Writing tag for epoch %llu\n", (unsigned long long) FSPRG_GetEpoch(fsprg_state(f)));
+
+ assert(f->hmac);
+
+ r = journal_file_append_object(f, OBJECT_TAG, sizeof(struct TagObject), &o, &p);
+ if (r < 0)
+ return r;
+
+ o->tag.seqnum = htole64(journal_file_tag_seqnum(f));
+
+ /* Add the tag object itself, so that we can protect its
+ * header. This will exclude the actual hash value in it */
+ r = journal_file_hmac_put_object(f, OBJECT_TAG, p);
+ if (r < 0)
+ return r;
+
+ /* Get the HMAC tag and store it in the object */
+ memcpy(o->tag.tag, gcry_md_read(f->hmac, 0), TAG_LENGTH);
+ f->hmac_running = false;
+
+ return 0;
+}
+
+static int journal_file_hmac_start(JournalFile *f) {
+ uint8_t key[256 / 8]; /* Let's pass 256 bit from FSPRG to HMAC */
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ if (f->hmac_running)
+ return 0;
+
+ /* Prepare HMAC for next cycle */
+ gcry_md_reset(f->hmac);
+ FSPRG_GetKey(fsprg_state(f), key, sizeof(key), 0);
+ gcry_md_setkey(f->hmac, key, sizeof(key));
+
+ f->hmac_running = true;
+
+ return 0;
+}
+
+static int journal_file_get_epoch(JournalFile *f, uint64_t realtime, uint64_t *epoch) {
+ uint64_t t;
+
+ assert(f);
+ assert(epoch);
+ assert(f->authenticate);
+
+ if (le64toh(f->fsprg_header->fsprg_start_usec) == 0 ||
+ le64toh(f->fsprg_header->fsprg_interval_usec) == 0)
+ return -ENOTSUP;
+
+ if (realtime < le64toh(f->fsprg_header->fsprg_start_usec))
+ return -ESTALE;
+
+ t = realtime - le64toh(f->fsprg_header->fsprg_start_usec);
+ t = t / le64toh(f->fsprg_header->fsprg_interval_usec);
+
+ *epoch = t;
+ return 0;
+}
+
+static int journal_file_need_evolve(JournalFile *f, uint64_t realtime) {
+ uint64_t goal, epoch;
+ int r;
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_get_epoch(f, realtime, &goal);
+ if (r < 0)
+ return r;
+
+ epoch = FSPRG_GetEpoch(fsprg_state(f));
+ if (epoch > goal)
+ return -ESTALE;
+
+ return epoch != goal;
+}
+
+static int journal_file_evolve(JournalFile *f, uint64_t realtime) {
+ uint64_t goal, epoch;
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_get_epoch(f, realtime, &goal);
+ if (r < 0)
+ return r;
+
+ epoch = FSPRG_GetEpoch(fsprg_state(f));
+ if (epoch < goal)
+ log_debug("Evolving FSPRG key from epoch %llu to %llu.", (unsigned long long) epoch, (unsigned long long) goal);
+
+ for (;;) {
+ if (epoch > goal)
+ return -ESTALE;
+ if (epoch == goal)
+ return 0;
+
+ FSPRG_Evolve(fsprg_state(f));
+ epoch = FSPRG_GetEpoch(fsprg_state(f));
+ }
+}
+
+static int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime) {
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_need_evolve(f, realtime);
+ if (r <= 0)
+ return 0;
+
+ r = journal_file_append_tag(f);
+ if (r < 0)
+ return r;
+
+ r = journal_file_evolve(f, realtime);
+ if (r < 0)
+ return r;
+
+ r = journal_file_hmac_start(f);
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
+static int journal_file_hmac_put_object(JournalFile *f, int type, uint64_t p) {
+ int r;
+ Object *o;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_hmac_start(f);
+ if (r < 0)
+ return r;
+
+ r = journal_file_move_to_object(f, type, p, &o);
+ if (r < 0)
+ return r;
+
+ gcry_md_write(f->hmac, o, offsetof(ObjectHeader, payload));
+
+ switch (o->object.type) {
+
+ case OBJECT_DATA:
+ /* All but: hash and payload are mutable */
+ gcry_md_write(f->hmac, &o->data.hash, sizeof(o->data.hash));
+ gcry_md_write(f->hmac, o->data.payload, le64toh(o->object.size) - offsetof(DataObject, payload));
+ break;
+
+ case OBJECT_ENTRY:
+ /* All */
+ gcry_md_write(f->hmac, &o->entry.seqnum, le64toh(o->object.size) - offsetof(EntryObject, seqnum));
+ break;
+
+ case OBJECT_FIELD_HASH_TABLE:
+ case OBJECT_DATA_HASH_TABLE:
+ case OBJECT_ENTRY_ARRAY:
+ /* Nothing: everything is mutable */
+ break;
+
+ case OBJECT_TAG:
+ /* All but the tag itself */
+ gcry_md_write(f->hmac, &o->tag.seqnum, sizeof(o->tag.seqnum));
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
+static int journal_file_hmac_put_header(JournalFile *f) {
+ int r;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = journal_file_hmac_start(f);
+ if (r < 0)
+ return r;
+
+ /* All but state+reserved, boot_id, arena_size,
+ * tail_object_offset, n_objects, n_entries, tail_seqnum,
+ * head_entry_realtime, tail_entry_realtime,
+ * tail_entry_monotonic, n_data, n_fields, header_tag */
+
+ gcry_md_write(f->hmac, f->header->signature, offsetof(Header, state) - offsetof(Header, signature));
+ gcry_md_write(f->hmac, &f->header->file_id, offsetof(Header, boot_id) - offsetof(Header, file_id));
+ gcry_md_write(f->hmac, &f->header->seqnum_id, offsetof(Header, arena_size) - offsetof(Header, seqnum_id));
+ gcry_md_write(f->hmac, &f->header->data_hash_table_offset, offsetof(Header, tail_object_offset) - offsetof(Header, data_hash_table_offset));
+ gcry_md_write(f->hmac, &f->header->head_seqnum, offsetof(Header, head_entry_realtime) - offsetof(Header, head_seqnum));
+
+ return 0;
+}
+
+static int journal_file_load_fsprg(JournalFile *f) {
+ int r, fd = -1;
+ char *p = NULL;
+ struct stat st;
+ FSPRGHeader *m = NULL;
+ sd_id128_t machine;
+
+ assert(f);
+
+ if (!f->authenticate)
+ return 0;
+
+ r = sd_id128_get_machine(&machine);
+ if (r < 0)
+ return r;
+
+ if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/fsprg",
+ SD_ID128_FORMAT_VAL(machine)) < 0)
+ return -ENOMEM;
+
+ fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY, 0600);
+ if (fd < 0) {
+ log_error("Failed to open %s: %m", p);
+ r = -errno;
+ goto finish;
+ }
+
+ if (fstat(fd, &st) < 0) {
+ r = -errno;
+ goto finish;
+ }
+
+ if (st.st_size < (off_t) sizeof(FSPRGHeader)) {
+ r = -ENODATA;
+ goto finish;
+ }
+
+ m = mmap(NULL, PAGE_ALIGN(sizeof(FSPRGHeader)), PROT_READ, MAP_SHARED, fd, 0);
+ if (m == MAP_FAILED) {
+ m = NULL;
+ r = -errno;
+ goto finish;
+ }
+
+ if (memcmp(m->signature, FSPRG_HEADER_SIGNATURE, 8) != 0) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ if (m->incompatible_flags != 0) {
+ r = -EPROTONOSUPPORT;
+ goto finish;
+ }
+
+ if (le64toh(m->header_size) < sizeof(FSPRGHeader)) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ if (le64toh(m->state_size) != FSPRG_stateinbytes(m->secpar)) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ f->fsprg_size = le64toh(m->header_size) + le64toh(m->state_size);
+ if ((uint64_t) st.st_size < f->fsprg_size) {
+ r = -ENODATA;
+ goto finish;
+ }
+
+ if (!sd_id128_equal(machine, m->machine_id)) {
+ r = -EHOSTDOWN;
+ goto finish;
+ }
+
+ if (le64toh(m->fsprg_start_usec) <= 0 ||
+ le64toh(m->fsprg_interval_usec) <= 0) {
+ r = -EBADMSG;
+ goto finish;
+ }
+
+ f->fsprg_header = mmap(NULL, PAGE_ALIGN(f->fsprg_size), PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
+ if (f->fsprg_header == MAP_FAILED) {
+ f->fsprg_header = NULL;
+ r = -errno;
+ goto finish;
+ }
+
+ r = 0;
+
+finish:
+ if (m)
+ munmap(m, PAGE_ALIGN(sizeof(FSPRGHeader)));
+
+ if (fd >= 0)
+ close_nointr_nofail(fd);
+
+ free(p);
+ return r;
+}
+
+static int journal_file_setup_hmac(JournalFile *f) {
+ gcry_error_t e;
+
+ if (!f->authenticate)
+ return 0;
+
+ e = gcry_md_open(&f->hmac, GCRY_MD_SHA256, GCRY_MD_FLAG_HMAC);
+ if (e != 0)
+ return -ENOTSUP;
+
+ return 0;
+}
+
+static int journal_file_append_first_tag(JournalFile *f) {
+ int r;
+ uint64_t p;
+
+ if (!f->authenticate)
+ return 0;
+
+ log_debug("Calculating first tag...");
+
+ r = journal_file_hmac_put_header(f);
+ if (r < 0)
+ return r;
+
+ p = le64toh(f->header->field_hash_table_offset);
+ if (p < offsetof(Object, hash_table.items))
+ return -EINVAL;
+ p -= offsetof(Object, hash_table.items);
+
+ r = journal_file_hmac_put_object(f, OBJECT_FIELD_HASH_TABLE, p);
+ if (r < 0)
+ return r;
+
+ p = le64toh(f->header->data_hash_table_offset);
+ if (p < offsetof(Object, hash_table.items))
+ return -EINVAL;
+ p -= offsetof(Object, hash_table.items);
+
+ r = journal_file_hmac_put_object(f, OBJECT_DATA_HASH_TABLE, p);
+ if (r < 0)
+ return r;
+
+ r = journal_file_append_tag(f);
+ if (r < 0)
+ return r;
+
+ return 0;
+}
+
+void journal_file_dump(JournalFile *f) {
+ Object *o;
+ int r;
+ uint64_t p;
+
+ assert(f);
+
+ journal_file_print_header(f);
+
+ p = le64toh(f->header->header_size);
+ while (p != 0) {
+ r = journal_file_move_to_object(f, -1, p, &o);
+ if (r < 0)
+ goto fail;
+
+ switch (o->object.type) {
+
+ case OBJECT_UNUSED:
+ printf("Type: OBJECT_UNUSED\n");
+ break;
+
+ case OBJECT_DATA:
+ printf("Type: OBJECT_DATA\n");
+ break;
+
+ case OBJECT_ENTRY:
+ printf("Type: OBJECT_ENTRY %llu %llu %llu\n",
+ (unsigned long long) le64toh(o->entry.seqnum),
+ (unsigned long long) le64toh(o->entry.monotonic),
+ (unsigned long long) le64toh(o->entry.realtime));
+ break;
+
+ case OBJECT_FIELD_HASH_TABLE:
+ printf("Type: OBJECT_FIELD_HASH_TABLE\n");
+ break;
+
+ case OBJECT_DATA_HASH_TABLE:
+ printf("Type: OBJECT_DATA_HASH_TABLE\n");
+ break;
+
+ case OBJECT_ENTRY_ARRAY:
+ printf("Type: OBJECT_ENTRY_ARRAY\n");
+ break;
+
+ case OBJECT_TAG:
+ printf("Type: OBJECT_TAG %llu\n",
+ (unsigned long long) le64toh(o->tag.seqnum));
+ break;
+ }
+
+ if (o->object.flags & OBJECT_COMPRESSED)
+ printf("Flags: COMPRESSED\n");
+
+ if (p == le64toh(f->header->tail_object_offset))
+ p = 0;
+ else
+ p = p + ALIGN64(le64toh(o->object.size));
+ }
+
+ return;
+fail:
+ log_error("File corrupt");
+}
+
+void journal_file_print_header(JournalFile *f) {
+ char a[33], b[33], c[33];
+ char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX];
+
+ assert(f);
+
+ printf("File Path: %s\n"
+ "File ID: %s\n"
+ "Machine ID: %s\n"
+ "Boot ID: %s\n"
+ "Sequential Number ID: %s\n"
+ "State: %s\n"
+ "Compatible Flags:%s%s\n"
+ "Incompatible Flags:%s%s\n"
+ "Header size: %llu\n"
+ "Arena size: %llu\n"
+ "Data Hash Table Size: %llu\n"
+ "Field Hash Table Size: %llu\n"
+ "Objects: %llu\n"
+ "Entry Objects: %llu\n"
+ "Rotate Suggested: %s\n"
+ "Head Sequential Number: %llu\n"
+ "Tail Sequential Number: %llu\n"
+ "Head Realtime Timestamp: %s\n"
+ "Tail Realtime Timestamp: %s\n",
+ f->path,
+ sd_id128_to_string(f->header->file_id, a),
+ sd_id128_to_string(f->header->machine_id, b),
+ sd_id128_to_string(f->header->boot_id, c),
+ sd_id128_to_string(f->header->seqnum_id, c),
+ f->header->state == STATE_OFFLINE ? "offline" :
+ f->header->state == STATE_ONLINE ? "online" :
+ f->header->state == STATE_ARCHIVED ? "archived" : "unknown",
+ (f->header->compatible_flags & HEADER_COMPATIBLE_AUTHENTICATED) ? " AUTHENTICATED" : "",
+ (f->header->compatible_flags & ~HEADER_COMPATIBLE_AUTHENTICATED) ? " ???" : "",
+ (f->header->incompatible_flags & HEADER_INCOMPATIBLE_COMPRESSED) ? " COMPRESSED" : "",
+ (f->header->incompatible_flags & ~HEADER_INCOMPATIBLE_COMPRESSED) ? " ???" : "",
+ (unsigned long long) le64toh(f->header->header_size),
+ (unsigned long long) le64toh(f->header->arena_size),
+ (unsigned long long) le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+ (unsigned long long) le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
+ (unsigned long long) le64toh(f->header->n_objects),
+ (unsigned long long) le64toh(f->header->n_entries),
+ yes_no(journal_file_rotate_suggested(f)),
+ (unsigned long long) le64toh(f->header->head_seqnum),
+ (unsigned long long) le64toh(f->header->tail_seqnum),
+ format_timestamp(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
+ format_timestamp(y, sizeof(y), le64toh(f->header->tail_entry_realtime)));
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
+ printf("Data Objects: %llu\n"
+ "Data Hash Table Fill: %.1f%%\n",
+ (unsigned long long) le64toh(f->header->n_data),
+ 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))));
+
+ if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
+ printf("Field Objects: %llu\n"
+ "Field Hash Table Fill: %.1f%%\n",
+ (unsigned long long) le64toh(f->header->n_fields),
+ 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))));
+}
+
+int journal_file_open(
+ const char *fname,
+ int flags,
+ mode_t mode,
+ bool compress,
+ bool authenticate,
+ JournalMetrics *metrics,
+ JournalFile *template,
+ JournalFile **ret) {
+
+ JournalFile *f;
+ int r;
+ bool newly_created = false;
+
+ assert(fname);
+
+ if ((flags & O_ACCMODE) != O_RDONLY &&
+ (flags & O_ACCMODE) != O_RDWR)
+ return -EINVAL;
+
+ if (!endswith(fname, ".journal"))
+ return -EINVAL;
+
+ f = new0(JournalFile, 1);
+ if (!f)
+ return -ENOMEM;
+
+ f->fd = -1;
+ f->mode = mode;
+
+ f->flags = flags;
+ f->prot = prot_from_flags(flags);
+ f->writable = (flags & O_ACCMODE) != O_RDONLY;
+ f->compress = compress;
+ f->authenticate = authenticate;
+
+ f->path = strdup(fname);
+ if (!f->path) {
+ r = -ENOMEM;
+ goto fail;
+ }
+
+ f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode);
+ if (f->fd < 0) {
+ r = -errno;
+ goto fail;
+ }
+
+ if (fstat(f->fd, &f->last_stat) < 0) {
+ r = -errno;
+ goto fail;
+ }
+
+ if (f->last_stat.st_size == 0 && f->writable) {
+ newly_created = true;
+
+ /* Try to load the FSPRG state, and if we can't, then
+ * just don't do authentication */
+ r = journal_file_load_fsprg(f);
+ if (r < 0)
+ f->authenticate = false;
+
+ r = journal_file_init_header(f, template);
+ if (r < 0)
+ goto fail;
+
+ if (fstat(f->fd, &f->last_stat) < 0) {
+ r = -errno;