X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournal-file.c;h=ff439f24743ea134c6aee802839ed56d3ae51a0a;hp=cd5ab266f004a3d76ae08bb9db778139e70ae5f8;hb=0284adc6a60ce0af1107cb0b50041a65d731f39e;hpb=74d29da52409ab149738ea76824679d29598e0ac

diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index cd5ab266f..ff439f247 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -29,13 +29,13 @@
 
 #include "journal-def.h"
 #include "journal-file.h"
+#include "journal-authenticate.h"
 #include "lookup3.h"
 #include "compress.h"
+#include "fsprg.h"
 
-#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL)
-#define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL)
-
-#define DEFAULT_WINDOW_SIZE (8ULL*1024ULL*1024ULL)
+#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem))
+#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem))
 
 #define COMPRESSION_SIZE_THRESHOLD (512ULL)
 
@@ -58,35 +58,51 @@
  * size */
 #define DEFAULT_KEEP_FREE (1024ULL*1024ULL)                    /* 1 MB */
 
-static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
-
-#define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
+/* n_data was the first entry we added after the initial file format design */
+#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
 
 void journal_file_close(JournalFile *f) {
-        int t;
-
         assert(f);
 
+        /* Write the final tag */
+        if (f->authenticate)
+                journal_file_append_tag(f);
+
+        /* Sync everything to disk, before we mark the file offline */
+        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) {
-                if (f->writable)
+                /* 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;
 
                 munmap(f->header, PAGE_ALIGN(sizeof(Header)));
         }
 
-        for (t = 0; t < _WINDOW_MAX; t++)
-                if (f->windows[t].ptr)
-                        munmap(f->windows[t].ptr, f->windows[t].size);
-
         if (f->fd >= 0)
                 close_nointr_nofail(f->fd);
 
         free(f->path);
 
+        if (f->mmap)
+                mmap_cache_unref(f->mmap);
+
 #ifdef HAVE_XZ
         free(f->compress_buffer);
 #endif
 
+#ifdef HAVE_GCRYPT
+        if (f->fsprg_header)
+                munmap(f->fsprg_header, PAGE_ALIGN(f->fsprg_size));
+
+        if (f->hmac)
+                gcry_md_close(f->hmac);
+#endif
+
         free(f);
 }
 
@@ -98,16 +114,22 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) {
         assert(f);
 
         zero(h);
-        memcpy(h.signature, signature, 8);
+        memcpy(h.signature, HEADER_SIGNATURE, 8);
         h.header_size = htole64(ALIGN64(sizeof(h)));
 
+        h.incompatible_flags =
+                htole32(f->compress ? HEADER_INCOMPATIBLE_COMPRESSED : 0);
+
+        h.compatible_flags =
+                htole32(f->authenticate ? HEADER_COMPATIBLE_AUTHENTICATED : 0);
+
         r = sd_id128_randomize(&h.file_id);
         if (r < 0)
                 return r;
 
         if (template) {
                 h.seqnum_id = template->header->seqnum_id;
-                h.seqnum = template->header->seqnum;
+                h.tail_entry_seqnum = template->header->tail_entry_seqnum;
         } else
                 h.seqnum_id = h.file_id;
 
@@ -142,7 +164,9 @@ static int journal_file_refresh_header(JournalFile *f) {
 
         f->header->state = STATE_ONLINE;
 
-        __sync_synchronize();
+        /* Sync the online state to disk */
+        msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC);
+        fdatasync(f->fd);
 
         return 0;
 }
@@ -150,18 +174,37 @@ static int journal_file_refresh_header(JournalFile *f) {
 static int journal_file_verify_header(JournalFile *f) {
         assert(f);
 
-        if (memcmp(f->header, signature, 8))
+        if (memcmp(f->header->signature, HEADER_SIGNATURE, 8))
                 return -EBADMSG;
 
+        /* In both read and write mode we refuse to open files with
+         * incompatible flags we don't know */
 #ifdef HAVE_XZ
-        if ((le64toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
+        if ((le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
                 return -EPROTONOSUPPORT;
 #else
         if (f->header->incompatible_flags != 0)
                 return -EPROTONOSUPPORT;
 #endif
 
-        if (f->header->header_size != htole64(ALIGN64(sizeof(*(f->header)))))
+        /* When open for writing we refuse to open files with
+         * compatible flags, too */
+        if (f->writable) {
+#ifdef HAVE_GCRYPT
+                if ((le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_AUTHENTICATED) != 0)
+                        return -EPROTONOSUPPORT;
+#else
+                if (f->header->compatible_flags != 0)
+                        return -EPROTONOSUPPORT;
+#endif
+        }
+
+        /* The first addition was n_data, so check that we are at least this large */
+        if (le64toh(f->header->header_size) < HEADER_SIZE_MIN)
+                return -EBADMSG;
+
+        if ((le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED) &&
+                !JOURNAL_HEADER_CONTAINS(f->header, n_tags))
                 return -EBADMSG;
 
         if ((uint64_t) f->last_stat.st_size < (le64toh(f->header->header_size) + le64toh(f->header->arena_size)))
@@ -181,15 +224,20 @@ static int journal_file_verify_header(JournalFile *f) {
 
                 state = f->header->state;
 
-                if (state == STATE_ONLINE)
-                        log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path);
-                        /* FIXME: immediately rotate */
-                else if (state == STATE_ARCHIVED)
+                if (state == STATE_ONLINE) {
+                        log_debug("Journal file %s is already online. Assuming unclean closing.", f->path);
+                        return -EBUSY;
+                } else if (state == STATE_ARCHIVED)
                         return -ESHUTDOWN;
-                else if (state != STATE_OFFLINE)
-                        log_debug("Journal file %s has unknown state %u. Ignoring.", f->path, state);
+                else if (state != STATE_OFFLINE) {
+                        log_debug("Journal file %s has unknown state %u.", f->path, state);
+                        return -EBUSY;
+                }
         }
 
+        f->compress = !!(le32toh(f->header->incompatible_flags) & HEADER_INCOMPATIBLE_COMPRESSED);
+        f->authenticate = !!(le32toh(f->header->compatible_flags) & HEADER_COMPATIBLE_AUTHENTICATED);
+
         return 0;
 }
 
@@ -244,6 +292,8 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
         if (r != 0)
                 return -r;
 
+        mmap_cache_close_fd_range(f->mmap, f->fd, old_size);
+
         if (fstat(f->fd, &f->last_stat) < 0)
                 return -errno;
 
@@ -252,59 +302,11 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
         return 0;
 }
 
-static int journal_file_map(
-                JournalFile *f,
-                uint64_t offset,
-                uint64_t size,
-                void **_window,
-                uint64_t *_woffset,
-                uint64_t *_wsize,
-                void **ret) {
-
-        uint64_t woffset, wsize;
-        void *window;
-
+static int journal_file_move_to(JournalFile *f, int context, uint64_t offset, uint64_t size, void **ret) {
         assert(f);
-        assert(size > 0);
         assert(ret);
 
-        woffset = offset & ~((uint64_t) page_size() - 1ULL);
-        wsize = size + (offset - woffset);
-        wsize = PAGE_ALIGN(wsize);
-
         /* Avoid SIGBUS on invalid accesses */
-        if (woffset + wsize > (uint64_t) PAGE_ALIGN(f->last_stat.st_size))
-                return -EADDRNOTAVAIL;
-
-        window = mmap(NULL, wsize, f->prot, MAP_SHARED, f->fd, woffset);
-        if (window == MAP_FAILED)
-                return -errno;
-
-        if (_window)
-                *_window = window;
-
-        if (_woffset)
-                *_woffset = woffset;
-
-        if (_wsize)
-                *_wsize = wsize;
-
-        *ret = (uint8_t*) window + (offset - woffset);
-
-        return 0;
-}
-
-static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_t size, void **ret) {
-        void *p = NULL;
-        uint64_t delta;
-        int r;
-        Window *w;
-
-        assert(f);
-        assert(ret);
-        assert(wt >= 0);
-        assert(wt < _WINDOW_MAX);
-
         if (offset + size > (uint64_t) f->last_stat.st_size) {
                 /* Hmm, out of range? Let's refresh the fstat() data
                  * first, before we trust that check. */
@@ -314,57 +316,7 @@ static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_
                         return -EADDRNOTAVAIL;
         }
 
-        w = f->windows + wt;
-
-        if (_likely_(w->ptr &&
-                     w->offset <= offset &&
-                     w->offset + w->size >= offset + size)) {
-
-                *ret = (uint8_t*) w->ptr + (offset - w->offset);
-                return 0;
-        }
-
-        if (w->ptr) {
-                if (munmap(w->ptr, w->size) < 0)
-                        return -errno;
-
-                w->ptr = NULL;
-                w->size = w->offset = 0;
-        }
-
-        if (size < DEFAULT_WINDOW_SIZE) {
-                /* If the default window size is larger then what was
-                 * asked for extend the mapping a bit in the hope to
-                 * minimize needed remappings later on. We add half
-                 * the window space before and half behind the
-                 * requested mapping */
-
-                delta = (DEFAULT_WINDOW_SIZE - size) / 2;
-
-                if (delta > offset)
-                        delta = offset;
-
-                offset -= delta;
-                size = DEFAULT_WINDOW_SIZE;
-        } else
-                delta = 0;
-
-        if (offset + size > (uint64_t) f->last_stat.st_size)
-                size = (uint64_t) f->last_stat.st_size - offset;
-
-        if (size <= 0)
-                return -EADDRNOTAVAIL;
-
-        r = journal_file_map(f,
-                             offset, size,
-                             &w->ptr, &w->offset, &w->size,
-                             &p);
-
-        if (r < 0)
-                return r;
-
-        *ret = (uint8_t*) p + delta;
-        return 0;
+        return mmap_cache_get(f->mmap, f->fd, f->prot, context, offset, size, ret);
 }
 
 static bool verify_hash(Object *o) {
@@ -384,17 +336,38 @@ static bool verify_hash(Object *o) {
         return h1 == h2;
 }
 
+static uint64_t minimum_header_size(Object *o) {
+
+        static uint64_t table[] = {
+                [OBJECT_DATA] = sizeof(DataObject),
+                [OBJECT_FIELD] = sizeof(FieldObject),
+                [OBJECT_ENTRY] = sizeof(EntryObject),
+                [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject),
+                [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject),
+                [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject),
+                [OBJECT_TAG] = sizeof(TagObject),
+        };
+
+        if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0)
+                return sizeof(ObjectHeader);
+
+        return table[o->object.type];
+}
+
 int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) {
         int r;
         void *t;
         Object *o;
         uint64_t s;
+        unsigned context;
 
         assert(f);
         assert(ret);
-        assert(type < _OBJECT_TYPE_MAX);
 
-        r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t);
+        /* One context for each type, plus one catch-all for the rest */
+        context = type > 0 && type < _OBJECT_TYPE_MAX ? type : 0;
+
+        r = journal_file_move_to(f, context, offset, sizeof(ObjectHeader), &t);
         if (r < 0)
                 return r;
 
@@ -404,6 +377,12 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec
         if (s < sizeof(ObjectHeader))
                 return -EBADMSG;
 
+        if (o->object.type <= OBJECT_UNUSED)
+                return -EBADMSG;
+
+        if (s < minimum_header_size(o))
+                return -EBADMSG;
+
         if (type >= 0 && o->object.type != type)
                 return -EBADMSG;
 
@@ -422,12 +401,12 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec
         return 0;
 }
 
-static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
+static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) {
         uint64_t r;
 
         assert(f);
 
-        r = le64toh(f->header->seqnum) + 1;
+        r = le64toh(f->header->tail_entry_seqnum) + 1;
 
         if (seqnum) {
                 /* If an external seqnum counter was passed, we update
@@ -440,21 +419,22 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
                 *seqnum = r;
         }
 
-        f->header->seqnum = htole64(r);
+        f->header->tail_entry_seqnum = htole64(r);
 
-        if (f->header->first_seqnum == 0)
-                f->header->first_seqnum = htole64(r);
+        if (f->header->head_entry_seqnum == 0)
+                f->header->head_entry_seqnum = htole64(r);
 
         return r;
 }
 
-static int journal_file_append_object(JournalFile *f, int type, uint64_t size, Object **ret, uint64_t *offset) {
+int journal_file_append_object(JournalFile *f, int type, uint64_t size, Object **ret, uint64_t *offset) {
         int r;
         uint64_t p;
         Object *tail, *o;
         void *t;
 
         assert(f);
+        assert(type > 0 && type < _OBJECT_TYPE_MAX);
         assert(size >= sizeof(ObjectHeader));
         assert(offset);
         assert(ret);
@@ -500,7 +480,17 @@ static int journal_file_setup_data_hash_table(JournalFile *f) {
 
         assert(f);
 
-        s = DEFAULT_DATA_HASH_TABLE_SIZE;
+        /* We estimate that we need 1 hash table entry per 768 of
+           journal file and we want to make sure we never get beyond
+           75% fill level. Calculate the hash table size for the
+           maximum file size based on these metrics. */
+
+        s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem);
+        if (s < DEFAULT_DATA_HASH_TABLE_SIZE)
+                s = DEFAULT_DATA_HASH_TABLE_SIZE;
+
+        log_info("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem)));
+
         r = journal_file_append_object(f,
                                        OBJECT_DATA_HASH_TABLE,
                                        offsetof(Object, hash_table.items) + s,
@@ -550,7 +540,7 @@ static int journal_file_map_data_hash_table(JournalFile *f) {
         s = le64toh(f->header->data_hash_table_size);
 
         r = journal_file_move_to(f,
-                                 WINDOW_DATA_HASH_TABLE,
+                                 OBJECT_DATA_HASH_TABLE,
                                  p, s,
                                  &t);
         if (r < 0)
@@ -571,7 +561,7 @@ static int journal_file_map_field_hash_table(JournalFile *f) {
         s = le64toh(f->header->field_hash_table_size);
 
         r = journal_file_move_to(f,
-                                 WINDOW_FIELD_HASH_TABLE,
+                                 OBJECT_FIELD_HASH_TABLE,
                                  p, s,
                                  &t);
         if (r < 0)
@@ -614,6 +604,9 @@ static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, ui
 
         f->data_hash_table[h].tail_hash_offset = htole64(offset);
 
+        if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
+                f->header->n_data = htole64(le64toh(f->header->n_data) + 1);
+
         return 0;
 }
 
@@ -758,20 +751,22 @@ static int journal_file_append_data(
                         o->object.size = htole64(offsetof(Object, data.payload) + rsize);
                         o->object.flags |= OBJECT_COMPRESSED;
 
-                        f->header->incompatible_flags = htole32(le32toh(f->header->incompatible_flags) | HEADER_INCOMPATIBLE_COMPRESSED);
-
                         log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize);
                 }
         }
 #endif
 
-        if (!compressed)
+        if (!compressed && size > 0)
                 memcpy(o->data.payload, data, size);
 
         r = journal_file_link_data(f, o, p, hash);
         if (r < 0)
                 return r;
 
+        r = journal_file_hmac_put_object(f, OBJECT_DATA, p);
+        if (r < 0)
+                return r;
+
         /* The linking might have altered the window, so let's
          * refresh our pointer */
         r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
@@ -794,7 +789,7 @@ uint64_t journal_file_entry_n_items(Object *o) {
         return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
 }
 
-static uint64_t journal_file_entry_array_n_items(Object *o) {
+uint64_t journal_file_entry_array_n_items(Object *o) {
         assert(o);
         assert(o->object.type == OBJECT_ENTRY_ARRAY);
 
@@ -848,6 +843,10 @@ static int link_entry_into_array(JournalFile *f,
         if (r < 0)
                 return r;
 
+        r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, q);
+        if (r < 0)
+                return r;
+
         o->entry_array.items[i] = htole64(p);
 
         if (ap == 0)
@@ -978,13 +977,17 @@ static int journal_file_append_entry_internal(
         if (r < 0)
                 return r;
 
-        o->entry.seqnum = htole64(journal_file_seqnum(f, seqnum));
+        o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum));
         memcpy(o->entry.items, items, n_items * sizeof(EntryItem));
         o->entry.realtime = htole64(ts->realtime);
         o->entry.monotonic = htole64(ts->monotonic);
         o->entry.xor_hash = htole64(xor_hash);
         o->entry.boot_id = f->header->boot_id;
 
+        r = journal_file_hmac_put_object(f, OBJECT_ENTRY, np);
+        if (r < 0)
+                return r;
+
         r = journal_file_link_entry(f, o, np);
         if (r < 0)
                 return r;
@@ -1034,7 +1037,12 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
             ts->monotonic < le64toh(f->header->tail_entry_monotonic))
                 return -EINVAL;
 
-        items = alloca(sizeof(EntryItem) * n_iovec);
+        r = journal_file_maybe_append_tag(f, ts->realtime);
+        if (r < 0)
+                return r;
+
+        /* alloca() can't take 0, hence let's allocate at least one */
+        items = alloca(sizeof(EntryItem) * MAX(1, n_iovec));
 
         for (i = 0; i < n_iovec; i++) {
                 uint64_t p;
@@ -1809,27 +1817,13 @@ int journal_file_move_to_entry_by_realtime_for_data(
 }
 
 void journal_file_dump(JournalFile *f) {
-        char a[33], b[33], c[33];
         Object *o;
         int r;
         uint64_t p;
 
         assert(f);
 
-        printf("File Path: %s\n"
-               "File ID: %s\n"
-               "Machine ID: %s\n"
-               "Boot ID: %s\n"
-               "Arena size: %llu\n"
-               "Objects: %lu\n"
-               "Entries: %lu\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),
-               (unsigned long long) le64toh(f->header->arena_size),
-               (unsigned long) le64toh(f->header->n_objects),
-               (unsigned long) le64toh(f->header->n_entries));
+        journal_file_print_header(f);
 
         p = le64toh(f->header->header_size);
         while (p != 0) {
@@ -1866,8 +1860,9 @@ void journal_file_dump(JournalFile *f) {
                         printf("Type: OBJECT_ENTRY_ARRAY\n");
                         break;
 
-                case OBJECT_SIGNATURE:
-                        printf("Type: OBJECT_SIGNATURE\n");
+                case OBJECT_TAG:
+                        printf("Type: OBJECT_TAG %llu\n",
+                               (unsigned long long) le64toh(o->tag.seqnum));
                         break;
                 }
 
@@ -1885,10 +1880,76 @@ 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_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)));
+
+        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,
+                MMapCache *mmap_cache,
                 JournalFile *template,
                 JournalFile **ret) {
 
@@ -1910,14 +1971,25 @@ int journal_file_open(
                 return -ENOMEM;
 
         f->fd = -1;
-        f->flags = flags;
         f->mode = mode;
-        f->writable = (flags & O_ACCMODE) != O_RDONLY;
+
+        f->flags = flags;
         f->prot = prot_from_flags(flags);
+        f->writable = (flags & O_ACCMODE) != O_RDONLY;
+        f->compress = compress;
+        f->authenticate = authenticate;
 
-        if (template) {
-                f->metrics = template->metrics;
-                f->compress = template->compress;
+        if (mmap_cache)
+                f->mmap = mmap_cache_ref(mmap_cache);
+        else {
+                /* One context for each type, plus the zeroth catchall
+                 * context. One fd for the file plus one for each type
+                 * (which we need during verification */
+                f->mmap = mmap_cache_new(_OBJECT_TYPE_MAX, 1 + _OBJECT_TYPE_MAX);
+                if (!f->mmap) {
+                        r = -ENOMEM;
+                        goto fail;
+                }
         }
 
         f->path = strdup(fname);
@@ -1940,6 +2012,12 @@ int journal_file_open(
         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;
@@ -1950,7 +2028,7 @@ int journal_file_open(
                 }
         }
 
-        if (f->last_stat.st_size < (off_t) sizeof(Header)) {
+        if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) {
                 r = -EIO;
                 goto fail;
         }
@@ -1968,14 +2046,29 @@ int journal_file_open(
                         goto fail;
         }
 
+        if (!newly_created && f->writable) {
+                r = journal_file_load_fsprg(f);
+                if (r < 0)
+                        goto fail;
+        }
+
         if (f->writable) {
+                if (metrics) {
+                        journal_default_metrics(metrics, f->fd);
+                        f->metrics = *metrics;
+                } else if (template)
+                        f->metrics = template->metrics;
+
                 r = journal_file_refresh_header(f);
                 if (r < 0)
                         goto fail;
+
+                r = journal_file_setup_hmac(f);
+                if (r < 0)
+                        goto fail;
         }
 
         if (newly_created) {
-
                 r = journal_file_setup_field_hash_table(f);
                 if (r < 0)
                         goto fail;
@@ -1983,6 +2076,10 @@ int journal_file_open(
                 r = journal_file_setup_data_hash_table(f);
                 if (r < 0)
                         goto fail;
+
+                r = journal_file_append_first_tag(f);
+                if (r < 0)
+                        goto fail;
         }
 
         r = journal_file_map_field_hash_table(f);
@@ -2004,7 +2101,7 @@ fail:
         return r;
 }
 
-int journal_file_rotate(JournalFile **f) {
+int journal_file_rotate(JournalFile **f, bool compress, bool authenticate) {
         char *p;
         size_t l;
         JournalFile *old_file, *new_file = NULL;
@@ -2032,7 +2129,7 @@ int journal_file_rotate(JournalFile **f) {
         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->seqnum),
+                 (unsigned long long) le64toh((*f)->header->tail_entry_seqnum),
                  (unsigned long long) le64toh((*f)->header->tail_entry_realtime));
 
         r = rename(old_file->path, p);
@@ -2043,7 +2140,7 @@ int journal_file_rotate(JournalFile **f) {
 
         old_file->header->state = STATE_ARCHIVED;
 
-        r = journal_file_open(old_file->path, old_file->flags, old_file->mode, old_file, &new_file);
+        r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, authenticate, NULL, old_file->mmap, old_file, &new_file);
         journal_file_close(old_file);
 
         *f = new_file;
@@ -2054,6 +2151,10 @@ int journal_file_open_reliably(
                 const char *fname,
                 int flags,
                 mode_t mode,
+                bool compress,
+                bool authenticate,
+                JournalMetrics *metrics,
+                MMapCache *mmap,
                 JournalFile *template,
                 JournalFile **ret) {
 
@@ -2061,11 +2162,13 @@ int journal_file_open_reliably(
         size_t l;
         char *p;
 
-        r = journal_file_open(fname, flags, mode, template, ret);
+        r = journal_file_open(fname, flags, mode, compress, authenticate, metrics, mmap, template, ret);
         if (r != -EBADMSG && /* corrupted */
             r != -ENODATA && /* truncated */
             r != -EHOSTDOWN && /* other machine */
-            r != -EPROTONOSUPPORT) /* incompatible feature */
+            r != -EPROTONOSUPPORT && /* incompatible feature */
+            r != -EBUSY && /* unclean shutdown */
+            r != -ESHUTDOWN /* already archived */)
                 return r;
 
         if ((flags & O_ACCMODE) == O_RDONLY)
@@ -2074,6 +2177,9 @@ int journal_file_open_reliably(
         if (!(flags & O_CREAT))
                 return r;
 
+        if (!endswith(fname, ".journal"))
+                return r;
+
         /* The file is corrupted. Rotate it away and try it again (but only once) */
 
         l = strlen(fname);
@@ -2088,207 +2194,11 @@ int journal_file_open_reliably(
         if (r < 0)
                 return -errno;
 
-        log_warning("File %s corrupted, renaming and replacing.", fname);
+        log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
 
-        return journal_file_open(fname, flags, mode, template, ret);
+        return journal_file_open(fname, flags, mode, compress, authenticate, metrics, mmap, template, ret);
 }
 
-struct vacuum_info {
-        off_t usage;
-        char *filename;
-
-        uint64_t realtime;
-        sd_id128_t seqnum_id;
-        uint64_t seqnum;
-
-        bool have_seqnum;
-};
-
-static int vacuum_compare(const void *_a, const void *_b) {
-        const struct vacuum_info *a, *b;
-
-        a = _a;
-        b = _b;
-
-        if (a->have_seqnum && b->have_seqnum &&
-            sd_id128_equal(a->seqnum_id, b->seqnum_id)) {
-                if (a->seqnum < b->seqnum)
-                        return -1;
-                else if (a->seqnum > b->seqnum)
-                        return 1;
-                else
-                        return 0;
-        }
-
-        if (a->realtime < b->realtime)
-                return -1;
-        else if (a->realtime > b->realtime)
-                return 1;
-        else if (a->have_seqnum && b->have_seqnum)
-                return memcmp(&a->seqnum_id, &b->seqnum_id, 16);
-        else
-                return strcmp(a->filename, b->filename);
-}
-
-int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t min_free) {
-        DIR *d;
-        int r = 0;
-        struct vacuum_info *list = NULL;
-        unsigned n_list = 0, n_allocated = 0, i;
-        uint64_t sum = 0;
-
-        assert(directory);
-
-        if (max_use <= 0)
-                return 0;
-
-        d = opendir(directory);
-        if (!d)
-                return -errno;
-
-        for (;;) {
-                int k;
-                struct dirent buf, *de;
-                size_t q;
-                struct stat st;
-                char *p;
-                unsigned long long seqnum = 0, realtime;
-                sd_id128_t seqnum_id;
-                bool have_seqnum;
-
-                k = readdir_r(d, &buf, &de);
-                if (k != 0) {
-                        r = -k;
-                        goto finish;
-                }
-
-                if (!de)
-                        break;
-
-                if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0)
-                        continue;
-
-                if (!S_ISREG(st.st_mode))
-                        continue;
-
-                q = strlen(de->d_name);
-
-                if (endswith(de->d_name, ".journal")) {
-
-                        /* Vacuum archived files */
-
-                        if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8)
-                                continue;
-
-                        if (de->d_name[q-8-16-1] != '-' ||
-                            de->d_name[q-8-16-1-16-1] != '-' ||
-                            de->d_name[q-8-16-1-16-1-32-1] != '@')
-                                continue;
-
-                        p = strdup(de->d_name);
-                        if (!p) {
-                                r = -ENOMEM;
-                                goto finish;
-                        }
-
-                        de->d_name[q-8-16-1-16-1] = 0;
-                        if (sd_id128_from_string(de->d_name + q-8-16-1-16-1-32, &seqnum_id) < 0) {
-                                free(p);
-                                continue;
-                        }
-
-                        if (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) {
-                                free(p);
-                                continue;
-                        }
-
-                        have_seqnum = true;
-
-                } else if (endswith(de->d_name, ".journal~")) {
-                        unsigned long long tmp;
-
-                        /* Vacuum corrupted files */
-
-                        if (q < 1 + 16 + 1 + 16 + 8 + 1)
-                                continue;
-
-                        if (de->d_name[q-1-8-16-1] != '-' ||
-                            de->d_name[q-1-8-16-1-16-1] != '@')
-                                continue;
-
-                        p = strdup(de->d_name);
-                        if (!p) {
-                                r = -ENOMEM;
-                                goto finish;
-                        }
-
-                        if (sscanf(de->d_name + q-1-8-16-1-16, "%16llx-%16llx.journal~", &realtime, &tmp) != 2) {
-                                free(p);
-                                continue;
-                        }
-
-                        have_seqnum = false;
-                } else
-                        continue;
-
-                if (n_list >= n_allocated) {
-                        struct vacuum_info *j;
-
-                        n_allocated = MAX(n_allocated * 2U, 8U);
-                        j = realloc(list, n_allocated * sizeof(struct vacuum_info));
-                        if (!j) {
-                                free(p);
-                                r = -ENOMEM;
-                                goto finish;
-                        }
-
-                        list = j;
-                }
-
-                list[n_list].filename = p;
-                list[n_list].usage = 512UL * (uint64_t) st.st_blocks;
-                list[n_list].seqnum = seqnum;
-                list[n_list].realtime = realtime;
-                list[n_list].seqnum_id = seqnum_id;
-                list[n_list].have_seqnum = have_seqnum;
-
-                sum += list[n_list].usage;
-
-                n_list ++;
-        }
-
-        qsort(list, n_list, sizeof(struct vacuum_info), vacuum_compare);
-
-        for(i = 0; i < n_list; i++) {
-                struct statvfs ss;
-
-                if (fstatvfs(dirfd(d), &ss) < 0) {
-                        r = -errno;
-                        goto finish;
-                }
-
-                if (sum <= max_use &&
-                    (uint64_t) ss.f_bavail * (uint64_t) ss.f_bsize >= min_free)
-                        break;
-
-                if (unlinkat(dirfd(d), list[i].filename, 0) >= 0) {
-                        log_info("Deleted archived journal %s/%s.", directory, list[i].filename);
-                        sum -= list[i].usage;
-                } else if (errno != ENOENT)
-                        log_warning("Failed to delete %s/%s: %m", directory, list[i].filename);
-        }
-
-finish:
-        for (i = 0; i < n_list; i++)
-                free(list[i].filename);
-
-        free(list);
-
-        if (d)
-                closedir(d);
-
-        return r;
-}
 
 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;
@@ -2446,26 +2356,21 @@ void journal_default_metrics(JournalMetrics *m, int fd) {
 }
 
 int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) {
-        Object *o;
-        int r;
-
         assert(f);
         assert(from || to);
 
         if (from) {
-                r = journal_file_next_entry(f, NULL, 0, DIRECTION_DOWN, &o, NULL);
-                if (r <= 0)
-                        return r;
+                if (f->header->head_entry_realtime == 0)
+                        return -ENOENT;
 
-                *from = le64toh(o->entry.realtime);
+                *from = le64toh(f->header->head_entry_realtime);
         }
 
         if (to) {
-                r = journal_file_next_entry(f, NULL, 0, DIRECTION_UP, &o, NULL);
-                if (r <= 0)
-                        return r;
+                if (f->header->tail_entry_realtime == 0)
+                        return -ENOENT;
 
-                *to = le64toh(o->entry.realtime);
+                *to = le64toh(f->header->tail_entry_realtime);
         }
 
         return 1;
@@ -2515,3 +2420,44 @@ int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, u
 
         return 1;
 }
+
+bool journal_file_rotate_suggested(JournalFile *f) {
+        assert(f);
+
+        /* If we gained new header fields we gained new features,
+         * hence suggest a rotation */
+        if (le64toh(f->header->header_size) < sizeof(Header)) {
+                log_debug("%s uses an outdated header, suggesting rotation.", f->path);
+                return true;
+        }
+
+        /* Let's check if the hash tables grew over a certain fill
+         * level (75%, borrowing this value from Java's hash table
+         * implementation), and if so suggest a rotation. To calculate
+         * the fill level we need the n_data field, which only exists
+         * in newer versions. */
+
+        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.",
+                                  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)));
+                        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.",
+                                  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)));
+                        return true;
+                }
+
+        return false;
+}