chiark / gitweb /
journal: when appending to journal file, allocate larger blocks at once
[elogind.git] / src / journal / journal-file.c
index 43fd4453a44accfde468c66c37f1e236072f43e2..d606adaf8e38dba7d3b3237c7b01712d88951e74 100644 (file)
@@ -44,7 +44,7 @@
 #define COMPRESSION_SIZE_THRESHOLD (512ULL)
 
 /* This is the minimum journal file size */
-#define JOURNAL_FILE_SIZE_MIN (64ULL*1024ULL)                  /* 64 KiB */
+#define JOURNAL_FILE_SIZE_MIN (4ULL*1024ULL*1024ULL)           /* 4 MiB */
 
 /* These are the lower and upper bounds if we deduce the max_use value
  * from the file system size */
 /* n_data was the first entry we added after the initial file format design */
 #define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
 
+/* How many entries to keep in the entry array chain cache at max */
+#define CHAIN_CACHE_MAX 20
+
+/* How much to increase the journal file size at once each time we allocate something new. */
+#define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL)              /* 8MB */
+
+static int journal_file_set_online(JournalFile *f) {
+        assert(f);
+
+        if (!f->writable)
+                return -EPERM;
+
+        if (!(f->fd >= 0 && f->header))
+                return -EINVAL;
+
+        switch(f->header->state) {
+                case STATE_ONLINE:
+                        return 0;
+
+                case STATE_OFFLINE:
+                        f->header->state = STATE_ONLINE;
+                        fsync(f->fd);
+                        return 0;
+
+                default:
+                        return -EINVAL;
+        }
+}
+
+int journal_file_set_offline(JournalFile *f) {
+        assert(f);
+
+        if (!f->writable)
+                return -EPERM;
+
+        if (!(f->fd >= 0 && f->header))
+                return -EINVAL;
+
+        if (f->header->state != STATE_ONLINE)
+                return 0;
+
+        fsync(f->fd);
+
+        f->header->state = STATE_OFFLINE;
+
+        fsync(f->fd);
+
+        return 0;
+}
+
 void journal_file_close(JournalFile *f) {
         assert(f);
 
@@ -78,16 +128,10 @@ void journal_file_close(JournalFile *f) {
         if (f->mmap && f->fd >= 0)
                 mmap_cache_close_fd(f->mmap, f->fd);
 
-        if (f->writable && f->fd >= 0)
-                fdatasync(f->fd);
-
-        if (f->header) {
-                /* Mark the file offline. Don't override the archived state if it already is set */
-                if (f->writable && f->header->state == STATE_ONLINE)
-                        f->header->state = STATE_OFFLINE;
+        journal_file_set_offline(f);
 
+        if (f->header)
                 munmap(f->header, PAGE_ALIGN(sizeof(Header)));
-        }
 
         if (f->fd >= 0)
                 close_nointr_nofail(f->fd);
@@ -97,6 +141,8 @@ void journal_file_close(JournalFile *f) {
         if (f->mmap)
                 mmap_cache_unref(f->mmap);
 
+        hashmap_free_free(f->chain_cache);
+
 #ifdef HAVE_XZ
         free(f->compress_buffer);
 #endif
@@ -172,11 +218,10 @@ static int journal_file_refresh_header(JournalFile *f) {
 
         f->header->boot_id = boot_id;
 
-        f->header->state = STATE_ONLINE;
+        journal_file_set_online(f);
 
         /* Sync the online state to disk */
-        msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC);
-        fdatasync(f->fd);
+        fsync(f->fd);
 
         return 0;
 }
@@ -270,7 +315,7 @@ static int journal_file_verify_header(JournalFile *f) {
 }
 
 static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
-        uint64_t old_size, new_size;
+        uint64_t old_size, new_size, file_size;
         int r;
 
         assert(f);
@@ -290,12 +335,10 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
         if (new_size <= old_size)
                 return 0;
 
-        if (f->metrics.max_size > 0 &&
-            new_size > f->metrics.max_size)
+        if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
                 return -E2BIG;
 
-        if (new_size > f->metrics.min_size &&
-            f->metrics.keep_free > 0) {
+        if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) {
                 struct statvfs svfs;
 
                 if (fstatvfs(f->fd, &svfs) >= 0) {
@@ -320,8 +363,16 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
         if (r != 0)
                 return -r;
 
-        if (fstat(f->fd, &f->last_stat) < 0)
-                return -errno;
+        /* Increase the file size a bit further than this, so that we
+         * we can create larger memory maps to cache */
+        file_size = ((new_size+FILE_SIZE_INCREASE-1) / FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE;
+        if (file_size > (uint64_t) f->last_stat.st_size) {
+                if (file_size > new_size)
+                        ftruncate(f->fd, file_size);
+
+                if (fstat(f->fd, &f->last_stat) < 0)
+                        return -errno;
+        }
 
         f->header->arena_size = htole64(new_size - le64toh(f->header->header_size));
 
@@ -452,6 +503,10 @@ int journal_file_append_object(JournalFile *f, int type, uint64_t size, Object *
         assert(offset);
         assert(ret);
 
+        r = journal_file_set_online(f);
+        if (r < 0)
+                return r;
+
         p = le64toh(f->header->tail_object_offset);
         if (p == 0)
                 p = le64toh(f->header->header_size);
@@ -502,7 +557,7 @@ static int journal_file_setup_data_hash_table(JournalFile *f) {
         if (s < DEFAULT_DATA_HASH_TABLE_SIZE)
                 s = DEFAULT_DATA_HASH_TABLE_SIZE;
 
-        log_debug("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem)));
+        log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem));
 
         r = journal_file_append_object(f,
                                        OBJECT_DATA_HASH_TABLE,
@@ -775,7 +830,7 @@ int journal_file_find_data_object_with_hash(
 
                         l -= offsetof(Object, data.payload);
 
-                        if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize))
+                        if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0))
                                 return -EBADMSG;
 
                         if (rsize == size &&
@@ -860,6 +915,8 @@ static int journal_file_append_field(
 
         osize = offsetof(Object, field.payload) + size;
         r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p);
+        if (r < 0)
+                return r;
 
         o->field.hash = htole64(hash);
         memcpy(o->field.payload, field, size);
@@ -938,7 +995,7 @@ static int journal_file_append_data(
                         o->object.size = htole64(offsetof(Object, data.payload) + rsize);
                         o->object.flags |= OBJECT_COMPRESSED;
 
-                        log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize);
+                        log_debug("Compressed data object %"PRIu64" -> %"PRIu64, size, rsize);
                 }
         }
 #endif
@@ -1159,7 +1216,7 @@ static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
         if (r < 0)
                 return r;
 
-        /* log_debug("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries); */
+        /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */
 
         if (f->header->head_entry_realtime == 0)
                 f->header->head_entry_realtime = o->entry.realtime;
@@ -1262,9 +1319,6 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
         assert(f);
         assert(iovec || n_iovec == 0);
 
-        if (!f->writable)
-                return -EPERM;
-
         if (!ts) {
                 dual_timestamp_get(&_ts);
                 ts = &_ts;
@@ -1281,7 +1335,7 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
 #endif
 
         /* alloca() can't take 0, hence let's allocate at least one */
-        items = alloca(sizeof(EntryItem) * MAX(1, n_iovec));
+        items = alloca(sizeof(EntryItem) * MAX(1u, n_iovec));
 
         for (i = 0; i < n_iovec; i++) {
                 uint64_t p;
@@ -1298,7 +1352,7 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
 
         /* Order by the position on disk, in order to improve seek
          * times for rotating media. */
-        qsort(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
+        qsort_safe(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
 
         r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
 
@@ -1307,37 +1361,94 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
         return r;
 }
 
+typedef struct ChainCacheItem {
+        uint64_t first; /* the array at the begin of the chain */
+        uint64_t array; /* the cached array */
+        uint64_t begin; /* the first item in the cached array */
+        uint64_t total; /* the total number of items in all arrays before this one in the chain */
+} ChainCacheItem;
+
+static void chain_cache_put(
+                Hashmap *h,
+                ChainCacheItem *ci,
+                uint64_t first,
+                uint64_t array,
+                uint64_t begin,
+                uint64_t total) {
+
+        if (!ci) {
+                /* If the chain item to cache for this chain is the
+                 * first one it's not worth caching anything */
+                if (array == first)
+                        return;
+
+                if (hashmap_size(h) >= CHAIN_CACHE_MAX)
+                        ci = hashmap_steal_first(h);
+                else {
+                        ci = new(ChainCacheItem, 1);
+                        if (!ci)
+                                return;
+                }
+
+                ci->first = first;
+
+                if (hashmap_put(h, &ci->first, ci) < 0) {
+                        free(ci);
+                        return;
+                }
+        } else
+                assert(ci->first == first);
+
+        ci->array = array;
+        ci->begin = begin;
+        ci->total = total;
+}
+
 static int generic_array_get(JournalFile *f,
                              uint64_t first,
                              uint64_t i,
                              Object **ret, uint64_t *offset) {
 
         Object *o;
-        uint64_t p = 0, a;
+        uint64_t p = 0, a, t = 0;
         int r;
+        ChainCacheItem *ci;
 
         assert(f);
 
         a = first;
+
+        /* Try the chain cache first */
+        ci = hashmap_get(f->chain_cache, &first);
+        if (ci && i > ci->total) {
+                a = ci->array;
+                i -= ci->total;
+                t = ci->total;
+        }
+
         while (a > 0) {
-                uint64_t n;
+                uint64_t k;
 
                 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
                 if (r < 0)
                         return r;
 
-                n = journal_file_entry_array_n_items(o);
-                if (i < n) {
+                k = journal_file_entry_array_n_items(o);
+                if (i < k) {
                         p = le64toh(o->entry_array.items[i]);
-                        break;
+                        goto found;
                 }
 
-                i -= n;
+                i -= k;
+                t += k;
                 a = le64toh(o->entry_array.next_entry_array_offset);
         }
 
-        if (a <= 0 || p <= 0)
-                return 0;
+        return 0;
+
+found:
+        /* Let's cache this item for the next invocation */
+        chain_cache_put(f->chain_cache, ci, first, a, o->entry_array.items[0], t);
 
         r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
         if (r < 0)
@@ -1401,11 +1512,38 @@ static int generic_array_bisect(JournalFile *f,
         bool subtract_one = false;
         Object *o, *array = NULL;
         int r;
+        ChainCacheItem *ci;
 
         assert(f);
         assert(test_object);
 
+        /* Start with the first array in the chain */
         a = first;
+
+        ci = hashmap_get(f->chain_cache, &first);
+        if (ci && n > ci->total) {
+                /* Ah, we have iterated this bisection array chain
+                 * previously! Let's see if we can skip ahead in the
+                 * chain, as far as the last time. But we can't jump
+                 * backwards in the chain, so let's check that
+                 * first. */
+
+                r = test_object(f, ci->begin, needle);
+                if (r < 0)
+                        return r;
+
+                if (r == TEST_LEFT) {
+                        /* OK, what we are looking for is right of th
+                         * begin of this EntryArray, so let's jump
+                         * straight to previously cached array in the
+                         * chain */
+
+                        a = ci->array;
+                        n -= ci->total;
+                        t = ci->total;
+                }
+        }
+
         while (a > 0) {
                 uint64_t left, right, k, lp;
 
@@ -1486,6 +1624,9 @@ found:
         if (subtract_one && t == 0 && i == 0)
                 return 0;
 
+        /* Let's cache this item for the next invocation */
+        chain_cache_put(f->chain_cache, ci, first, a, array->entry_array.items[0], t);
+
         if (subtract_one && i == 0)
                 p = last_p;
         else if (subtract_one)
@@ -1581,7 +1722,7 @@ found:
         return 1;
 }
 
-static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
+_pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
         assert(f);
         assert(p > 0);
 
@@ -1699,6 +1840,17 @@ static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) {
                 return TEST_RIGHT;
 }
 
+static inline int find_data_object_by_boot_id(
+                JournalFile *f,
+                sd_id128_t boot_id,
+                Object **o,
+                uint64_t *b) {
+        char t[sizeof("_BOOT_ID=")-1 + 32 + 1] = "_BOOT_ID=";
+
+        sd_id128_to_string(boot_id, t + 9);
+        return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b);
+}
+
 int journal_file_move_to_entry_by_monotonic(
                 JournalFile *f,
                 sd_id128_t boot_id,
@@ -1707,14 +1859,12 @@ int journal_file_move_to_entry_by_monotonic(
                 Object **ret,
                 uint64_t *offset) {
 
-        char t[9+32+1] = "_BOOT_ID=";
         Object *o;
         int r;
 
         assert(f);
 
-        sd_id128_to_string(boot_id, t + 9);
-        r = journal_file_find_data_object(f, t, strlen(t), &o, NULL);
+        r = find_data_object_by_boot_id(f, boot_id, &o, NULL);
         if (r < 0)
                 return r;
         if (r == 0)
@@ -1928,7 +2078,6 @@ int journal_file_move_to_entry_by_monotonic_for_data(
                 direction_t direction,
                 Object **ret, uint64_t *offset) {
 
-        char t[9+32+1] = "_BOOT_ID=";
         Object *o, *d;
         int r;
         uint64_t b, z;
@@ -1936,8 +2085,7 @@ int journal_file_move_to_entry_by_monotonic_for_data(
         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);
+        r = find_data_object_by_boot_id(f, boot_id, &o, &b);
         if (r < 0)
                 return r;
         if (r == 0)
@@ -2089,10 +2237,10 @@ void journal_file_dump(JournalFile *f) {
                         break;
 
                 case OBJECT_ENTRY:
-                        printf("Type: OBJECT_ENTRY seqnum=%llu monotonic=%llu realtime=%llu\n",
-                               (unsigned long long) le64toh(o->entry.seqnum),
-                               (unsigned long long) le64toh(o->entry.monotonic),
-                               (unsigned long long) le64toh(o->entry.realtime));
+                        printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n",
+                               le64toh(o->entry.seqnum),
+                               le64toh(o->entry.monotonic),
+                               le64toh(o->entry.realtime));
                         break;
 
                 case OBJECT_FIELD_HASH_TABLE:
@@ -2108,9 +2256,9 @@ void journal_file_dump(JournalFile *f) {
                         break;
 
                 case OBJECT_TAG:
-                        printf("Type: OBJECT_TAG seqnum=%llu epoch=%llu\n",
-                               (unsigned long long) le64toh(o->tag.seqnum),
-                               (unsigned long long) le64toh(o->tag.epoch));
+                        printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n",
+                               le64toh(o->tag.seqnum),
+                               le64toh(o->tag.epoch));
                         break;
 
                 default:
@@ -2132,9 +2280,18 @@ fail:
         log_error("File corrupt");
 }
 
+static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) {
+        const char *x;
+
+        x = format_timestamp(buf, l, t);
+        if (x)
+                return x;
+        return " --- ";
+}
+
 void journal_file_print_header(JournalFile *f) {
-        char a[33], b[33], c[33];
-        char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX];
+        char a[33], b[33], c[33], d[33];
+        char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX];
         struct stat st;
         char bytes[FORMAT_BYTES_MAX];
 
@@ -2148,22 +2305,23 @@ void journal_file_print_header(JournalFile *f) {
                "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"
+               "Header size: %"PRIu64"\n"
+               "Arena size: %"PRIu64"\n"
+               "Data Hash Table Size: %"PRIu64"\n"
+               "Field Hash Table Size: %"PRIu64"\n"
                "Rotate Suggested: %s\n"
-               "Head Sequential Number: %llu\n"
-               "Tail Sequential Number: %llu\n"
+               "Head Sequential Number: %"PRIu64"\n"
+               "Tail Sequential Number: %"PRIu64"\n"
                "Head Realtime Timestamp: %s\n"
                "Tail Realtime Timestamp: %s\n"
-               "Objects: %llu\n"
-               "Entry Objects: %llu\n",
+               "Tail Monotonic Timestamp: %s\n"
+               "Objects: %"PRIu64"\n"
+               "Entry Objects: %"PRIu64"\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),
+               sd_id128_to_string(f->header->seqnum_id, d),
                f->header->state == STATE_OFFLINE ? "OFFLINE" :
                f->header->state == STATE_ONLINE ? "ONLINE" :
                f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN",
@@ -2171,36 +2329,37 @@ void journal_file_print_header(JournalFile *f) {
                (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SEALED) ? " ???" : "",
                JOURNAL_HEADER_COMPRESSED(f->header) ? " COMPRESSED" : "",
                (le32toh(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),
+               le64toh(f->header->header_size),
+               le64toh(f->header->arena_size),
+               le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+               le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
                yes_no(journal_file_rotate_suggested(f, 0)),
-               (unsigned long long) le64toh(f->header->head_entry_seqnum),
-               (unsigned long long) le64toh(f->header->tail_entry_seqnum),
-               format_timestamp(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
-               format_timestamp(y, sizeof(y), le64toh(f->header->tail_entry_realtime)),
-               (unsigned long long) le64toh(f->header->n_objects),
-               (unsigned long long) le64toh(f->header->n_entries));
+               le64toh(f->header->head_entry_seqnum),
+               le64toh(f->header->tail_entry_seqnum),
+               format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
+               format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)),
+               format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC),
+               le64toh(f->header->n_objects),
+               le64toh(f->header->n_entries));
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
-                printf("Data Objects: %llu\n"
+                printf("Data Objects: %"PRIu64"\n"
                        "Data Hash Table Fill: %.1f%%\n",
-                       (unsigned long long) le64toh(f->header->n_data),
+                       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"
+                printf("Field Objects: %"PRIu64"\n"
                        "Field Hash Table Fill: %.1f%%\n",
-                       (unsigned long long) le64toh(f->header->n_fields),
+                       le64toh(f->header->n_fields),
                        100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))));
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_tags))
-                printf("Tag Objects: %llu\n",
-                       (unsigned long long) le64toh(f->header->n_tags));
+                printf("Tag Objects: %"PRIu64"\n",
+                       le64toh(f->header->n_tags));
         if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
-                printf("Entry Array Objects: %llu\n",
-                       (unsigned long long) le64toh(f->header->n_entry_arrays));
+                printf("Entry Array Objects: %"PRIu64"\n",
+                       le64toh(f->header->n_entry_arrays));
 
         if (fstat(f->fd, &st) >= 0)
                 printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (off_t) st.st_blocks * 512ULL));
@@ -2265,6 +2424,12 @@ int journal_file_open(
                 goto fail;
         }
 
+        f->chain_cache = hashmap_new(uint64_hash_func, uint64_compare_func);
+        if (!f->chain_cache) {
+                r = -ENOMEM;
+                goto fail;
+        }
+
         f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode);
         if (f->fd < 0) {
                 r = -errno;
@@ -2394,7 +2559,7 @@ fail:
 }
 
 int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
-        char *p;
+        _cleanup_free_ char *p = NULL;
         size_t l;
         JournalFile *old_file, *new_file = NULL;
         int r;
@@ -2411,22 +2576,15 @@ int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
                 return -EINVAL;
 
         l = strlen(old_file->path);
-
-        p = new(char, l + 1 + 32 + 1 + 16 + 1 + 16 + 1);
-        if (!p)
+        r = asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal",
+                     (int) l - 8, old_file->path,
+                     SD_ID128_FORMAT_VAL(old_file->header->seqnum_id),
+                     le64toh((*f)->header->head_entry_seqnum),
+                     le64toh((*f)->header->head_entry_realtime));
+        if (r < 0)
                 return -ENOMEM;
 
-        memcpy(p, old_file->path, l - 8);
-        p[l-8] = '@';
-        sd_id128_to_string(old_file->header->seqnum_id, p + l - 8 + 1);
-        snprintf(p + l - 8 + 1 + 32, 1 + 16 + 1 + 16 + 8 + 1,
-                 "-%016llx-%016llx.journal",
-                 (unsigned long long) le64toh((*f)->header->head_entry_seqnum),
-                 (unsigned long long) le64toh((*f)->header->head_entry_realtime));
-
         r = rename(old_file->path, p);
-        free(p);
-
         if (r < 0)
                 return -errno;
 
@@ -2452,7 +2610,7 @@ int journal_file_open_reliably(
 
         int r;
         size_t l;
-        char *p;
+        _cleanup_free_ char *p = NULL;
 
         r = journal_file_open(fname, flags, mode, compress, seal,
                               metrics, mmap_cache, template, ret);
@@ -2477,13 +2635,12 @@ int journal_file_open_reliably(
 
         l = strlen(fname);
         if (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
-                     (int) (l-8), fname,
+                     (int) l - 8, fname,
                      (unsigned long long) now(CLOCK_REALTIME),
                      random_ull()) < 0)
                 return -ENOMEM;
 
         r = rename(fname, p);
-        free(p);
         if (r < 0)
                 return -errno;
 
@@ -2493,7 +2650,6 @@ int journal_file_open_reliably(
                                  metrics, mmap_cache, template, ret);
 }
 
-
 int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p, uint64_t *seqnum, Object **ret, uint64_t *offset) {
         uint64_t i, n;
         uint64_t q, xor_hash = 0;
@@ -2547,7 +2703,7 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
 #ifdef HAVE_XZ
                         uint64_t rsize;
 
-                        if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize))
+                        if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0))
                                 return -EBADMSG;
 
                         data = from->compress_buffer;
@@ -2633,7 +2789,7 @@ void journal_default_metrics(JournalMetrics *m, int fd) {
         if (m->keep_free == (uint64_t) -1) {
 
                 if (fs_size > 0) {
-                        m->keep_free = PAGE_ALIGN(fs_size / 20); /* 5% of file system size */
+                        m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */
 
                         if (m->keep_free > DEFAULT_KEEP_FREE_UPPER)
                                 m->keep_free = DEFAULT_KEEP_FREE_UPPER;
@@ -2671,7 +2827,6 @@ int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *
 }
 
 int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) {
-        char t[9+32+1] = "_BOOT_ID=";
         Object *o;
         uint64_t p;
         int r;
@@ -2679,9 +2834,7 @@ int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, u
         assert(f);
         assert(from || to);
 
-        sd_id128_to_string(boot_id, t + 9);
-
-        r = journal_file_find_data_object(f, t, strlen(t), &o, &p);
+        r = find_data_object_by_boot_id(f, boot_id, &o, &p);
         if (r <= 0)
                 return r;
 
@@ -2733,26 +2886,33 @@ bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) {
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
                 if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) {
-                        log_debug("Data hash table of %s has a fill level at %.1f (%llu of %llu items, %llu file size, %llu bytes per hash table item), suggesting rotation.",
+                        log_debug("Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" bytes per hash table item), suggesting rotation.",
                                   f->path,
                                   100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))),
-                                  (unsigned long long) le64toh(f->header->n_data),
-                                  (unsigned long long) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)),
-                                  (unsigned long long) (f->last_stat.st_size),
-                                  (unsigned long long) (f->last_stat.st_size / le64toh(f->header->n_data)));
+                                  le64toh(f->header->n_data),
+                                  le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+                                  (unsigned long long) f->last_stat.st_size,
+                                  f->last_stat.st_size / le64toh(f->header->n_data));
                         return true;
                 }
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
                 if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) {
-                        log_debug("Field hash table of %s has a fill level at %.1f (%llu of %llu items), suggesting rotation.",
+                        log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.",
                                   f->path,
                                   100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))),
-                                  (unsigned long long) le64toh(f->header->n_fields),
-                                  (unsigned long long) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)));
+                                  le64toh(f->header->n_fields),
+                                  le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
                         return true;
                 }
 
+        /* Are the data objects properly indexed by field objects? */
+        if (JOURNAL_HEADER_CONTAINS(f->header, n_data) &&
+            JOURNAL_HEADER_CONTAINS(f->header, n_fields) &&
+            le64toh(f->header->n_data) > 0 &&
+            le64toh(f->header->n_fields) == 0)
+                return true;
+
         if (max_file_usec > 0) {
                 usec_t t, h;