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=d3535d2fadcb2edad6756bc8915c93f76e09473b;hp=04564d989e4f23621016d3697d2889e500d1b4af;hb=92261977d81fd6a5bfb1418eddd86582d8a57fcd;hpb=4a92baf3fa97c73f04438b5814db58783f920b32

diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index 04564d989..d3535d2fa 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -26,21 +26,22 @@
 #include <sys/statvfs.h>
 #include <fcntl.h>
 #include <stddef.h>
+#include <sys/xattr.h>
 
 #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*sizeof(HashItem))
 #define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem))
 
-#define DEFAULT_WINDOW_SIZE (8ULL*1024ULL*1024ULL)
-
 #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 */
@@ -61,59 +62,125 @@
 /* n_data was the first entry we added after the initial file format design */
 #define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
 
-#define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
+/* How many entries to keep in the entry array chain cache at max */
+#define CHAIN_CACHE_MAX 20
 
-#define JOURNAL_HEADER_CONTAINS(h, field) \
-        (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field))
+/* 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 const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
+static int journal_file_set_online(JournalFile *f) {
+        assert(f);
 
-void journal_file_close(JournalFile *f) {
-        int t;
+        if (!f->writable)
+                return -EPERM;
 
-        assert(f);
+        if (!(f->fd >= 0 && f->header))
+                return -EINVAL;
 
-        if (f->header) {
-                if (f->writable)
-                        f->header->state = STATE_OFFLINE;
+        switch(f->header->state) {
+                case STATE_ONLINE:
+                        return 0;
 
-                munmap(f->header, PAGE_ALIGN(sizeof(Header)));
+                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);
+
+#ifdef HAVE_GCRYPT
+        /* Write the final tag */
+        if (f->seal && f->writable)
+                journal_file_append_tag(f);
+#endif
+
+        /* Sync everything to disk, before we mark the file offline */
+        if (f->mmap && f->fd >= 0)
+                mmap_cache_close_fd(f->mmap, f->fd);
 
-        for (t = 0; t < _WINDOW_MAX; t++)
-                if (f->windows[t].ptr)
-                        munmap(f->windows[t].ptr, f->windows[t].size);
+        journal_file_set_offline(f);
 
-        if (f->fd >= 0)
-                close_nointr_nofail(f->fd);
+        if (f->header)
+                munmap(f->header, PAGE_ALIGN(sizeof(Header)));
 
+        safe_close(f->fd);
         free(f->path);
 
-#ifdef HAVE_XZ
+        if (f->mmap)
+                mmap_cache_unref(f->mmap);
+
+        hashmap_free_free(f->chain_cache);
+
+#if defined(HAVE_XZ) || defined(HAVE_LZ4)
         free(f->compress_buffer);
 #endif
 
+#ifdef HAVE_GCRYPT
+        if (f->fss_file)
+                munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size));
+        else if (f->fsprg_state)
+                free(f->fsprg_state);
+
+        free(f->fsprg_seed);
+
+        if (f->hmac)
+                gcry_md_close(f->hmac);
+#endif
+
         free(f);
 }
 
 static int journal_file_init_header(JournalFile *f, JournalFile *template) {
-        Header h;
+        Header h = {};
         ssize_t k;
         int r;
 
         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_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ |
+                f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4);
+
+        h.compatible_flags = htole32(
+                f->seal * HEADER_COMPATIBLE_SEALED);
+
         r = sd_id128_randomize(&h.file_id);
         if (r < 0)
                 return r;
 
         if (template) {
                 h.seqnum_id = template->header->seqnum_id;
-                h.tail_seqnum = template->header->tail_seqnum;
+                h.tail_entry_seqnum = template->header->tail_entry_seqnum;
         } else
                 h.seqnum_id = h.file_id;
 
@@ -146,32 +213,76 @@ 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_synchronize();
+        /* Sync the online state to disk */
+        fsync(f->fd);
 
         return 0;
 }
 
 static int journal_file_verify_header(JournalFile *f) {
+        uint32_t flags;
+
         assert(f);
 
-        if (memcmp(f->header, signature, 8))
+        if (memcmp(f->header->signature, HEADER_SIGNATURE, 8))
                 return -EBADMSG;
 
-#ifdef HAVE_XZ
-        if ((le64toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
+        /* In both read and write mode we refuse to open files with
+         * incompatible flags we don't know */
+        flags = le32toh(f->header->incompatible_flags);
+        if (flags & ~HEADER_INCOMPATIBLE_SUPPORTED) {
+                if (flags & ~HEADER_INCOMPATIBLE_ANY)
+                        log_debug("Journal file %s has unknown incompatible flags %"PRIx32,
+                                  f->path, flags & ~HEADER_INCOMPATIBLE_ANY);
+                flags = (flags & HEADER_INCOMPATIBLE_ANY) & ~HEADER_INCOMPATIBLE_SUPPORTED;
+                if (flags)
+                        log_debug("Journal file %s uses incompatible flags %"PRIx32
+                                  " disabled at compilation time.", f->path, flags);
                 return -EPROTONOSUPPORT;
-#else
-        if (f->header->incompatible_flags != 0)
+        }
+
+        /* When open for writing we refuse to open files with
+         * compatible flags, too */
+        flags = le32toh(f->header->compatible_flags);
+        if (f->writable && (flags & ~HEADER_COMPATIBLE_SUPPORTED)) {
+                if (flags & ~HEADER_COMPATIBLE_ANY)
+                        log_debug("Journal file %s has unknown compatible flags %"PRIx32,
+                                  f->path, flags & ~HEADER_COMPATIBLE_ANY);
+                flags = (flags & HEADER_COMPATIBLE_ANY) & ~HEADER_COMPATIBLE_SUPPORTED;
+                if (flags)
+                        log_debug("Journal file %s uses compatible flags %"PRIx32
+                                  " disabled at compilation time.", f->path, flags);
                 return -EPROTONOSUPPORT;
-#endif
+        }
+
+        if (f->header->state >= _STATE_MAX)
+                return -EBADMSG;
 
         /* 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 ((uint64_t) f->last_stat.st_size < (le64toh(f->header->header_size) + le64toh(f->header->arena_size)))
+        if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
+                return -EBADMSG;
+
+        if ((le64toh(f->header->header_size) + le64toh(f->header->arena_size)) > (uint64_t) f->last_stat.st_size)
+                return -ENODATA;
+
+        if (le64toh(f->header->tail_object_offset) > (le64toh(f->header->header_size) + le64toh(f->header->arena_size)))
+                return -ENODATA;
+
+        if (!VALID64(le64toh(f->header->data_hash_table_offset)) ||
+            !VALID64(le64toh(f->header->field_hash_table_offset)) ||
+            !VALID64(le64toh(f->header->tail_object_offset)) ||
+            !VALID64(le64toh(f->header->entry_array_offset)))
+                return -ENODATA;
+
+        if (le64toh(f->header->data_hash_table_offset) < le64toh(f->header->header_size) ||
+            le64toh(f->header->field_hash_table_offset) < le64toh(f->header->header_size) ||
+            le64toh(f->header->tail_object_offset) < le64toh(f->header->header_size) ||
+            le64toh(f->header->entry_array_offset) < le64toh(f->header->header_size))
                 return -ENODATA;
 
         if (f->writable) {
@@ -199,6 +310,11 @@ static int journal_file_verify_header(JournalFile *f) {
                 }
         }
 
+        f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header);
+        f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header);
+
+        f->seal = JOURNAL_HEADER_SEALED(f->header);
+
         return 0;
 }
 
@@ -223,12 +339,10 @@ 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) {
@@ -246,6 +360,11 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
                 }
         }
 
+        /* Increase by larger blocks at once */
+        new_size = ((new_size+FILE_SIZE_INCREASE-1) / FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE;
+        if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
+                new_size = f->metrics.max_size;
+
         /* Note that the glibc fallocate() fallback is very
            inefficient, hence we try to minimize the allocation area
            as we can. */
@@ -261,59 +380,14 @@ 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, bool keep_always, 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);
+        if (size <= 0)
+                return -EINVAL;
 
         /* 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. */
@@ -323,74 +397,25 @@ 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, keep_always, offset, size, &f->last_stat, ret);
 }
 
-static bool verify_hash(Object *o) {
-        uint64_t h1, h2;
+static uint64_t minimum_header_size(Object *o) {
 
-        assert(o);
+        static const 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 == OBJECT_DATA && !(o->object.flags & OBJECT_COMPRESSED)) {
-                h1 = le64toh(o->data.hash);
-                h2 = hash64(o->data.payload, le64toh(o->object.size) - offsetof(Object, data.payload));
-        } else if (o->object.type == OBJECT_FIELD) {
-                h1 = le64toh(o->field.hash);
-                h2 = hash64(o->field.payload, le64toh(o->object.size) - offsetof(Object, field.payload));
-        } else
-                return true;
+        if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0)
+                return sizeof(ObjectHeader);
 
-        return h1 == h2;
+        return table[o->object.type];
 }
 
 int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) {
@@ -401,9 +426,13 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec
 
         assert(f);
         assert(ret);
-        assert(type < _OBJECT_TYPE_MAX);
 
-        r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t);
+        /* Objects may only be located at multiple of 64 bit */
+        if (!VALID64(offset))
+                return -EFAULT;
+
+
+        r = journal_file_move_to(f, type_to_context(type), false, offset, sizeof(ObjectHeader), &t);
         if (r < 0)
                 return r;
 
@@ -413,30 +442,33 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec
         if (s < sizeof(ObjectHeader))
                 return -EBADMSG;
 
-        if (type >= 0 && o->object.type != type)
+        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;
 
         if (s > sizeof(ObjectHeader)) {
-                r = journal_file_move_to(f, o->object.type, offset, s, &t);
+                r = journal_file_move_to(f, o->object.type, false, offset, s, &t);
                 if (r < 0)
                         return r;
 
                 o = (Object*) t;
         }
 
-        if (!verify_hash(o))
-                return -EBADMSG;
-
         *ret = o;
         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->tail_seqnum) + 1;
+        r = le64toh(f->header->tail_entry_seqnum) + 1;
 
         if (seqnum) {
                 /* If an external seqnum counter was passed, we update
@@ -449,25 +481,30 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
                 *seqnum = r;
         }
 
-        f->header->tail_seqnum = htole64(r);
+        f->header->tail_entry_seqnum = htole64(r);
 
-        if (f->header->head_seqnum == 0)
-                f->header->head_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);
 
+        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);
@@ -483,7 +520,7 @@ static int journal_file_append_object(JournalFile *f, int type, uint64_t size, O
         if (r < 0)
                 return r;
 
-        r = journal_file_move_to(f, type, p, size, &t);
+        r = journal_file_move_to(f, type, false, p, size, &t);
         if (r < 0)
                 return r;
 
@@ -509,16 +546,16 @@ static int journal_file_setup_data_hash_table(JournalFile *f) {
 
         assert(f);
 
-        /* We estimate that we need 1 hash table entry per 2K of
+        /* 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 / 2048 / 3) * sizeof(HashItem);
+        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);
+        log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem));
 
         r = journal_file_append_object(f,
                                        OBJECT_DATA_HASH_TABLE,
@@ -527,7 +564,7 @@ static int journal_file_setup_data_hash_table(JournalFile *f) {
         if (r < 0)
                 return r;
 
-        memset(o->hash_table.items, 0, s);
+        memzero(o->hash_table.items, s);
 
         f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
         f->header->data_hash_table_size = htole64(s);
@@ -542,6 +579,9 @@ static int journal_file_setup_field_hash_table(JournalFile *f) {
 
         assert(f);
 
+        /* We use a fixed size hash table for the fields as this
+         * number should grow very slowly only */
+
         s = DEFAULT_FIELD_HASH_TABLE_SIZE;
         r = journal_file_append_object(f,
                                        OBJECT_FIELD_HASH_TABLE,
@@ -550,7 +590,7 @@ static int journal_file_setup_field_hash_table(JournalFile *f) {
         if (r < 0)
                 return r;
 
-        memset(o->hash_table.items, 0, s);
+        memzero(o->hash_table.items, s);
 
         f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
         f->header->field_hash_table_size = htole64(s);
@@ -569,7 +609,8 @@ 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,
+                                 true,
                                  p, s,
                                  &t);
         if (r < 0)
@@ -590,7 +631,8 @@ 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,
+                                 true,
                                  p, s,
                                  &t);
         if (r < 0)
@@ -600,14 +642,61 @@ static int journal_file_map_field_hash_table(JournalFile *f) {
         return 0;
 }
 
-static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash) {
+static int journal_file_link_field(
+                JournalFile *f,
+                Object *o,
+                uint64_t offset,
+                uint64_t hash) {
+
+        uint64_t p, h;
+        int r;
+
+        assert(f);
+        assert(o);
+        assert(offset > 0);
+
+        if (o->object.type != OBJECT_FIELD)
+                return -EINVAL;
+
+        /* This might alter the window we are looking at */
+
+        o->field.next_hash_offset = o->field.head_data_offset = 0;
+
+        h = hash % (le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
+        p = le64toh(f->field_hash_table[h].tail_hash_offset);
+        if (p == 0)
+                f->field_hash_table[h].head_hash_offset = htole64(offset);
+        else {
+                r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
+                if (r < 0)
+                        return r;
+
+                o->field.next_hash_offset = htole64(offset);
+        }
+
+        f->field_hash_table[h].tail_hash_offset = htole64(offset);
+
+        if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
+                f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1);
+
+        return 0;
+}
+
+static int journal_file_link_data(
+                JournalFile *f,
+                Object *o,
+                uint64_t offset,
+                uint64_t hash) {
+
         uint64_t p, h;
         int r;
 
         assert(f);
         assert(o);
         assert(offset > 0);
-        assert(o->object.type == OBJECT_DATA);
+
+        if (o->object.type != OBJECT_DATA)
+                return -EINVAL;
 
         /* This might alter the window we are looking at */
 
@@ -617,10 +706,10 @@ static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, ui
 
         h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
         p = le64toh(f->data_hash_table[h].tail_hash_offset);
-        if (p == 0) {
+        if (p == 0)
                 /* Only entry in the hash table is easy */
                 f->data_hash_table[h].head_hash_offset = htole64(offset);
-        } else {
+        else {
                 /* Move back to the previous data object, to patch in
                  * pointer */
 
@@ -639,6 +728,67 @@ static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, ui
         return 0;
 }
 
+int journal_file_find_field_object_with_hash(
+                JournalFile *f,
+                const void *field, uint64_t size, uint64_t hash,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t p, osize, h;
+        int r;
+
+        assert(f);
+        assert(field && size > 0);
+
+        osize = offsetof(Object, field.payload) + size;
+
+        if (f->header->field_hash_table_size == 0)
+                return -EBADMSG;
+
+        h = hash % (le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
+        p = le64toh(f->field_hash_table[h].head_hash_offset);
+
+        while (p > 0) {
+                Object *o;
+
+                r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o);
+                if (r < 0)
+                        return r;
+
+                if (le64toh(o->field.hash) == hash &&
+                    le64toh(o->object.size) == osize &&
+                    memcmp(o->field.payload, field, size) == 0) {
+
+                        if (ret)
+                                *ret = o;
+                        if (offset)
+                                *offset = p;
+
+                        return 1;
+                }
+
+                p = le64toh(o->field.next_hash_offset);
+        }
+
+        return 0;
+}
+
+int journal_file_find_field_object(
+                JournalFile *f,
+                const void *field, uint64_t size,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t hash;
+
+        assert(f);
+        assert(field && size > 0);
+
+        hash = hash64(field, size);
+
+        return journal_file_find_field_object_with_hash(f,
+                                                        field, size, hash,
+                                                        ret, offset);
+}
+
 int journal_file_find_data_object_with_hash(
                 JournalFile *f,
                 const void *data, uint64_t size, uint64_t hash,
@@ -668,8 +818,7 @@ int journal_file_find_data_object_with_hash(
                 if (le64toh(o->data.hash) != hash)
                         goto next;
 
-                if (o->object.flags & OBJECT_COMPRESSED) {
-#ifdef HAVE_XZ
+                if (o->object.flags & OBJECT_COMPRESSION_MASK) {
                         uint64_t l, rsize;
 
                         l = le64toh(o->object.size);
@@ -678,8 +827,10 @@ 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))
-                                return -EBADMSG;
+                        r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK,
+                                            o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0);
+                        if (r < 0)
+                                return r;
 
                         if (rsize == size &&
                             memcmp(f->compress_buffer, data, size) == 0) {
@@ -692,9 +843,6 @@ int journal_file_find_data_object_with_hash(
 
                                 return 1;
                         }
-#else
-                        return -EPROTONOSUPPORT;
-#endif
 
                 } else if (le64toh(o->object.size) == osize &&
                            memcmp(o->data.payload, data, size) == 0) {
@@ -732,6 +880,68 @@ int journal_file_find_data_object(
                                                        ret, offset);
 }
 
+static int journal_file_append_field(
+                JournalFile *f,
+                const void *field, uint64_t size,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t hash, p;
+        uint64_t osize;
+        Object *o;
+        int r;
+
+        assert(f);
+        assert(field && size > 0);
+
+        hash = hash64(field, size);
+
+        r = journal_file_find_field_object_with_hash(f, field, size, hash, &o, &p);
+        if (r < 0)
+                return r;
+        else if (r > 0) {
+
+                if (ret)
+                        *ret = o;
+
+                if (offset)
+                        *offset = p;
+
+                return 0;
+        }
+
+        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);
+
+        r = journal_file_link_field(f, o, p, hash);
+        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_FIELD, p, &o);
+        if (r < 0)
+                return r;
+
+#ifdef HAVE_GCRYPT
+        r = journal_file_hmac_put_object(f, OBJECT_FIELD, o, p);
+        if (r < 0)
+                return r;
+#endif
+
+        if (ret)
+                *ret = o;
+
+        if (offset)
+                *offset = p;
+
+        return 0;
+}
+
 static int journal_file_append_data(
                 JournalFile *f,
                 const void *data, uint64_t size,
@@ -740,8 +950,8 @@ static int journal_file_append_data(
         uint64_t hash, p;
         uint64_t osize;
         Object *o;
-        int r;
-        bool compressed = false;
+        int r, compression = 0;
+        const void *eq;
 
         assert(f);
         assert(data || size == 0);
@@ -769,25 +979,24 @@ static int journal_file_append_data(
 
         o->data.hash = htole64(hash);
 
-#ifdef HAVE_XZ
-        if (f->compress &&
+#if defined(HAVE_XZ) || defined(HAVE_LZ4)
+        if (f->compress_xz &&
             size >= COMPRESSION_SIZE_THRESHOLD) {
                 uint64_t rsize;
 
-                compressed = compress_blob(data, size, o->data.payload, &rsize);
+                compression = compress_blob(data, size, o->data.payload, &rsize);
 
-                if (compressed) {
+                if (compression) {
                         o->object.size = htole64(offsetof(Object, data.payload) + rsize);
-                        o->object.flags |= OBJECT_COMPRESSED;
+                        o->object.flags |= compression;
 
-                        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);
+                        log_debug("Compressed data object %"PRIu64" -> %"PRIu64" using %s",
+                                  size, rsize, object_compressed_to_string(compression));
                 }
         }
 #endif
 
-        if (!compressed)
+        if (!compression && size > 0)
                 memcpy(o->data.payload, data, size);
 
         r = journal_file_link_data(f, o, p, hash);
@@ -800,6 +1009,30 @@ static int journal_file_append_data(
         if (r < 0)
                 return r;
 
+        if (!data)
+                eq = NULL;
+        else
+                eq = memchr(data, '=', size);
+        if (eq && eq > data) {
+                Object *fo = NULL;
+                uint64_t fp;
+
+                /* Create field object ... */
+                r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp);
+                if (r < 0)
+                        return r;
+
+                /* ... and link it in. */
+                o->data.next_field_offset = fo->field.head_data_offset;
+                fo->field.head_data_offset = le64toh(p);
+        }
+
+#ifdef HAVE_GCRYPT
+        r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p);
+        if (r < 0)
+                return r;
+#endif
+
         if (ret)
                 *ret = o;
 
@@ -811,18 +1044,32 @@ static int journal_file_append_data(
 
 uint64_t journal_file_entry_n_items(Object *o) {
         assert(o);
-        assert(o->object.type == OBJECT_ENTRY);
+
+        if (o->object.type != OBJECT_ENTRY)
+                return 0;
 
         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);
+
+        if (o->object.type != OBJECT_ENTRY_ARRAY)
+                return 0;
 
         return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t);
 }
 
+uint64_t journal_file_hash_table_n_items(Object *o) {
+        assert(o);
+
+        if (o->object.type != OBJECT_DATA_HASH_TABLE &&
+            o->object.type != OBJECT_FIELD_HASH_TABLE)
+                return 0;
+
+        return (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem);
+}
+
 static int link_entry_into_array(JournalFile *f,
                                  le64_t *first,
                                  le64_t *idx,
@@ -870,6 +1117,12 @@ static int link_entry_into_array(JournalFile *f,
         if (r < 0)
                 return r;
 
+#ifdef HAVE_GCRYPT
+        r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q);
+        if (r < 0)
+                return r;
+#endif
+
         o->entry_array.items[i] = htole64(p);
 
         if (ap == 0)
@@ -882,6 +1135,9 @@ static int link_entry_into_array(JournalFile *f,
                 o->entry_array.next_entry_array_offset = htole64(q);
         }
 
+        if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
+                f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1);
+
         *idx = htole64(hidx + 1);
 
         return 0;
@@ -945,7 +1201,9 @@ static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
         assert(f);
         assert(o);
         assert(offset > 0);
-        assert(o->object.type == OBJECT_ENTRY);
+
+        if (o->object.type != OBJECT_ENTRY)
+                return -EINVAL;
 
         __sync_synchronize();
 
@@ -957,7 +1215,7 @@ 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;
@@ -1000,13 +1258,19 @@ 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;
 
+#ifdef HAVE_GCRYPT
+        r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np);
+        if (r < 0)
+                return r;
+#endif
+
         r = journal_file_link_entry(f, o, np);
         if (r < 0)
                 return r;
@@ -1031,7 +1295,17 @@ void journal_file_post_change(JournalFile *f) {
         __sync_synchronize();
 
         if (ftruncate(f->fd, f->last_stat.st_size) < 0)
-                log_error("Failed to to truncate file to its own size: %m");
+                log_error("Failed to truncate file to its own size: %m");
+}
+
+static int entry_item_cmp(const void *_a, const void *_b) {
+        const EntryItem *a = _a, *b = _b;
+
+        if (le64toh(a->object_offset) < le64toh(b->object_offset))
+                return -1;
+        if (le64toh(a->object_offset) > le64toh(b->object_offset))
+                return 1;
+        return 0;
 }
 
 int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqnum, Object **ret, uint64_t *offset) {
@@ -1044,9 +1318,6 @@ 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;
@@ -1056,7 +1327,14 @@ 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);
+#ifdef HAVE_GCRYPT
+        r = journal_file_maybe_append_tag(f, ts->realtime);
+        if (r < 0)
+                return r;
+#endif
+
+        /* alloca() can't take 0, hence let's allocate at least one */
+        items = alloca(sizeof(EntryItem) * MAX(1u, n_iovec));
 
         for (i = 0; i < n_iovec; i++) {
                 uint64_t p;
@@ -1071,6 +1349,10 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
                 items[i].hash = o->data.hash;
         }
 
+        /* Order by the position on disk, in order to improve seek
+         * times for rotating media. */
+        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);
 
         journal_file_post_change(f);
@@ -1078,37 +1360,98 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
         return r;
 }
 
-static int generic_array_get(JournalFile *f,
-                             uint64_t first,
-                             uint64_t i,
-                             Object **ret, uint64_t *offset) {
+typedef struct ChainCacheItem {
+        uint64_t first; /* the array at the beginning 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 */
+        uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */
+} ChainCacheItem;
+
+static void chain_cache_put(
+                Hashmap *h,
+                ChainCacheItem *ci,
+                uint64_t first,
+                uint64_t array,
+                uint64_t begin,
+                uint64_t total,
+                uint64_t last_index) {
+
+        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;
+        ci->last_index = last_index;
+}
+
+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, le64toh(o->entry_array.items[0]), t, i);
 
         r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
         if (r < 0)
@@ -1123,11 +1466,12 @@ static int generic_array_get(JournalFile *f,
         return 1;
 }
 
-static int generic_array_get_plus_one(JournalFile *f,
-                                      uint64_t extra,
-                                      uint64_t first,
-                                      uint64_t i,
-                                      Object **ret, uint64_t *offset) {
+static int generic_array_get_plus_one(
+                JournalFile *f,
+                uint64_t extra,
+                uint64_t first,
+                uint64_t i,
+                Object **ret, uint64_t *offset) {
 
         Object *o;
 
@@ -1158,25 +1502,54 @@ enum {
         TEST_RIGHT
 };
 
-static int generic_array_bisect(JournalFile *f,
-                                uint64_t first,
-                                uint64_t n,
-                                uint64_t needle,
-                                int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
-                                direction_t direction,
-                                Object **ret,
-                                uint64_t *offset,
-                                uint64_t *idx) {
-
-        uint64_t a, p, t = 0, i = 0, last_p = 0;
+static int generic_array_bisect(
+                JournalFile *f,
+                uint64_t first,
+                uint64_t n,
+                uint64_t needle,
+                int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset,
+                uint64_t *idx) {
+
+        uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = (uint64_t) -1;
         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 the
+                         * 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;
+                        last_index = ci->last_index;
+                }
+        }
+
         while (a > 0) {
                 uint64_t left, right, k, lp;
 
@@ -1204,6 +1577,58 @@ static int generic_array_bisect(JournalFile *f,
                 if (r == TEST_RIGHT) {
                         left = 0;
                         right -= 1;
+
+                        if (last_index != (uint64_t) -1) {
+                                assert(last_index <= right);
+
+                                /* If we cached the last index we
+                                 * looked at, let's try to not to jump
+                                 * too wildly around and see if we can
+                                 * limit the range to look at early to
+                                 * the immediate neighbors of the last
+                                 * index we looked at. */
+
+                                if (last_index > 0) {
+                                        uint64_t x = last_index - 1;
+
+                                        p = le64toh(array->entry_array.items[x]);
+                                        if (p <= 0)
+                                                return -EBADMSG;
+
+                                        r = test_object(f, p, needle);
+                                        if (r < 0)
+                                                return r;
+
+                                        if (r == TEST_FOUND)
+                                                r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
+
+                                        if (r == TEST_RIGHT)
+                                                right = x;
+                                        else
+                                                left = x + 1;
+                                }
+
+                                if (last_index < right) {
+                                        uint64_t y = last_index + 1;
+
+                                        p = le64toh(array->entry_array.items[y]);
+                                        if (p <= 0)
+                                                return -EBADMSG;
+
+                                        r = test_object(f, p, needle);
+                                        if (r < 0)
+                                                return r;
+
+                                        if (r == TEST_FOUND)
+                                                r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
+
+                                        if (r == TEST_RIGHT)
+                                                right = y;
+                                        else
+                                                left = y + 1;
+                                }
+                        }
+
                         for (;;) {
                                 if (left == right) {
                                         if (direction == DIRECTION_UP)
@@ -1214,8 +1639,8 @@ static int generic_array_bisect(JournalFile *f,
                                 }
 
                                 assert(left < right);
-
                                 i = (left + right) / 2;
+
                                 p = le64toh(array->entry_array.items[i]);
                                 if (p <= 0)
                                         return -EBADMSG;
@@ -1248,6 +1673,7 @@ static int generic_array_bisect(JournalFile *f,
 
                 n -= k;
                 t += k;
+                last_index = (uint64_t) -1;
                 a = le64toh(array->entry_array.next_entry_array_offset);
         }
 
@@ -1257,6 +1683,9 @@ 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, le64toh(array->entry_array.items[0]), t, subtract_one ? (i > 0 ? i-1 : (uint64_t) -1) : i);
+
         if (subtract_one && i == 0)
                 p = last_p;
         else if (subtract_one)
@@ -1280,16 +1709,18 @@ found:
         return 1;
 }
 
-static int generic_array_bisect_plus_one(JournalFile *f,
-                                         uint64_t extra,
-                                         uint64_t first,
-                                         uint64_t n,
-                                         uint64_t needle,
-                                         int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
-                                         direction_t direction,
-                                         Object **ret,
-                                         uint64_t *offset,
-                                         uint64_t *idx) {
+
+static int generic_array_bisect_plus_one(
+                JournalFile *f,
+                uint64_t extra,
+                uint64_t first,
+                uint64_t n,
+                uint64_t needle,
+                int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset,
+                uint64_t *idx) {
 
         int r;
         bool step_back = false;
@@ -1352,7 +1783,7 @@ 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);
 
@@ -1470,6 +1901,17 @@ 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,
@@ -1478,14 +1920,12 @@ 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)
@@ -1507,7 +1947,7 @@ int journal_file_next_entry(
                 direction_t direction,
                 Object **ret, uint64_t *offset) {
 
-        uint64_t i, n;
+        uint64_t i, n, ofs;
         int r;
 
         assert(f);
@@ -1548,10 +1988,24 @@ int journal_file_next_entry(
         }
 
         /* And jump to it */
-        return generic_array_get(f,
-                                 le64toh(f->header->entry_array_offset),
-                                 i,
-                                 ret, offset);
+        r = generic_array_get(f,
+                              le64toh(f->header->entry_array_offset),
+                              i,
+                              ret, &ofs);
+        if (r <= 0)
+                return r;
+
+        if (p > 0 &&
+            (direction == DIRECTION_DOWN ? ofs <= p : ofs >= p)) {
+                log_debug("%s: entry array corrupted at entry %"PRIu64,
+                          f->path, i);
+                return -EBADMSG;
+        }
+
+        if (offset)
+                *offset = ofs;
+
+        return 1;
 }
 
 int journal_file_skip_entry(
@@ -1699,7 +2153,6 @@ 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;
@@ -1707,8 +2160,7 @@ 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)
@@ -1774,8 +2226,6 @@ int journal_file_move_to_entry_by_monotonic_for_data(
 
                 z = q;
         }
-
-        return 0;
 }
 
 int journal_file_move_to_entry_by_seqnum_for_data(
@@ -1855,11 +2305,15 @@ void journal_file_dump(JournalFile *f) {
                         printf("Type: OBJECT_DATA\n");
                         break;
 
+                case OBJECT_FIELD:
+                        printf("Type: OBJECT_FIELD\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));
+                        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:
@@ -1874,13 +2328,20 @@ 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 seqnum=%"PRIu64" epoch=%"PRIu64"\n",
+                               le64toh(o->tag.seqnum),
+                               le64toh(o->tag.epoch));
+                        break;
+
+                default:
+                        printf("Type: unknown (%u)\n", o->object.type);
                         break;
                 }
 
-                if (o->object.flags & OBJECT_COMPRESSED)
-                        printf("Flags: COMPRESSED\n");
+                if (o->object.flags & OBJECT_COMPRESSION_MASK)
+                        printf("Flags: %s\n",
+                               object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK));
 
                 if (p == le64toh(f->header->tail_object_offset))
                         p = 0;
@@ -1893,9 +2354,20 @@ 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];
 
         assert(f);
 
@@ -1904,52 +2376,78 @@ void journal_file_print_header(JournalFile *f) {
                "Machine ID: %s\n"
                "Boot ID: %s\n"
                "Sequential Number ID: %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"
+               "State: %s\n"
+               "Compatible Flags:%s%s\n"
+               "Incompatible Flags:%s%s%s\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",
+               "Tail Realtime Timestamp: %s\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),
-               (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)));
+               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",
+               JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "",
+               (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "",
+               JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "",
+               JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "",
+               (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "",
+               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)),
+               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: %"PRIu64"\n",
+                       le64toh(f->header->n_tags));
+        if (JOURNAL_HEADER_CONTAINS(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));
 }
 
 int journal_file_open(
                 const char *fname,
                 int flags,
                 mode_t mode,
+                bool compress,
+                bool seal,
                 JournalMetrics *metrics,
+                MMapCache *mmap_cache,
                 JournalFile *template,
                 JournalFile **ret) {
 
@@ -1958,12 +2456,14 @@ int journal_file_open(
         bool newly_created = false;
 
         assert(fname);
+        assert(ret);
 
         if ((flags & O_ACCMODE) != O_RDONLY &&
             (flags & O_ACCMODE) != O_RDWR)
                 return -EINVAL;
 
-        if (!endswith(fname, ".journal"))
+        if (!endswith(fname, ".journal") &&
+            !endswith(fname, ".journal~"))
                 return -EINVAL;
 
         f = new0(JournalFile, 1);
@@ -1971,13 +2471,29 @@ 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;
+#if defined(HAVE_LZ4)
+        f->compress_lz4 = compress;
+#elif defined(HAVE_XZ)
+        f->compress_xz = compress;
+#endif
+#ifdef HAVE_GCRYPT
+        f->seal = seal;
+#endif
 
-        if (template)
-                f->compress = template->compress;
+        if (mmap_cache)
+                f->mmap = mmap_cache_ref(mmap_cache);
+        else {
+                f->mmap = mmap_cache_new();
+                if (!f->mmap) {
+                        r = -ENOMEM;
+                        goto fail;
+                }
+        }
 
         f->path = strdup(fname);
         if (!f->path) {
@@ -1985,6 +2501,12 @@ 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;
@@ -1997,7 +2519,30 @@ int journal_file_open(
         }
 
         if (f->last_stat.st_size == 0 && f->writable) {
-                newly_created = true;
+                uint64_t crtime;
+
+                /* Let's attach the creation time to the journal file,
+                 * so that the vacuuming code knows the age of this
+                 * file even if the file might end up corrupted one
+                 * day... Ideally we'd just use the creation time many
+                 * file systems maintain for each file, but there is
+                 * currently no usable API to query this, hence let's
+                 * emulate this via extended attributes. If extended
+                 * attributes are not supported we'll just skip this,
+                 * and rely solely on mtime/atime/ctime of the file.*/
+
+                crtime = htole64((uint64_t) now(CLOCK_REALTIME));
+                fsetxattr(f->fd, "user.crtime_usec", &crtime, sizeof(crtime), XATTR_CREATE);
+
+#ifdef HAVE_GCRYPT
+                /* Try to load the FSPRG state, and if we can't, then
+                 * just don't do sealing */
+                if (f->seal) {
+                        r = journal_file_fss_load(f);
+                        if (r < 0)
+                                f->seal = false;
+                }
+#endif
 
                 r = journal_file_init_header(f, template);
                 if (r < 0)
@@ -2007,6 +2552,8 @@ int journal_file_open(
                         r = -errno;
                         goto fail;
                 }
+
+                newly_created = true;
         }
 
         if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) {
@@ -2027,6 +2574,14 @@ int journal_file_open(
                         goto fail;
         }
 
+#ifdef HAVE_GCRYPT
+        if (!newly_created && f->writable) {
+                r = journal_file_fss_load(f);
+                if (r < 0)
+                        goto fail;
+        }
+#endif
+
         if (f->writable) {
                 if (metrics) {
                         journal_default_metrics(metrics, f->fd);
@@ -2039,8 +2594,13 @@ int journal_file_open(
                         goto fail;
         }
 
-        if (newly_created) {
+#ifdef HAVE_GCRYPT
+        r = journal_file_hmac_setup(f);
+        if (r < 0)
+                goto fail;
+#endif
 
+        if (newly_created) {
                 r = journal_file_setup_field_hash_table(f);
                 if (r < 0)
                         goto fail;
@@ -2048,6 +2608,12 @@ int journal_file_open(
                 r = journal_file_setup_data_hash_table(f);
                 if (r < 0)
                         goto fail;
+
+#ifdef HAVE_GCRYPT
+                r = journal_file_append_first_tag(f);
+                if (r < 0)
+                        goto fail;
+#endif
         }
 
         r = journal_file_map_field_hash_table(f);
@@ -2058,9 +2624,7 @@ int journal_file_open(
         if (r < 0)
                 goto fail;
 
-        if (ret)
-                *ret = f;
-
+        *ret = f;
         return 0;
 
 fail:
@@ -2069,8 +2633,8 @@ fail:
         return r;
 }
 
-int journal_file_rotate(JournalFile **f) {
-        char *p;
+int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
+        _cleanup_free_ char *p = NULL;
         size_t l;
         JournalFile *old_file, *new_file = NULL;
         int r;
@@ -2087,28 +2651,21 @@ int journal_file_rotate(JournalFile **f) {
                 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->tail_seqnum),
-                 (unsigned long long) le64toh((*f)->header->tail_entry_realtime));
-
         r = rename(old_file->path, p);
-        free(p);
-
         if (r < 0)
                 return -errno;
 
         old_file->header->state = STATE_ARCHIVED;
 
-        r = journal_file_open(old_file->path, old_file->flags, old_file->mode, NULL, old_file, &new_file);
+        r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, seal, NULL, old_file->mmap, old_file, &new_file);
         journal_file_close(old_file);
 
         *f = new_file;
@@ -2119,19 +2676,25 @@ int journal_file_open_reliably(
                 const char *fname,
                 int flags,
                 mode_t mode,
+                bool compress,
+                bool seal,
                 JournalMetrics *metrics,
+                MMapCache *mmap_cache,
                 JournalFile *template,
                 JournalFile **ret) {
 
         int r;
         size_t l;
-        char *p;
+        _cleanup_free_ char *p = NULL;
 
-        r = journal_file_open(fname, flags, mode, metrics, template, ret);
+        r = journal_file_open(fname, flags, mode, compress, seal,
+                              metrics, mmap_cache, 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)
@@ -2140,220 +2703,26 @@ 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);
-        if (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
-                     (int) (l-8), fname,
+        if (asprintf(&p, "%.*s@%016llx-%016" PRIx64 ".journal~",
+                     (int) l - 8, fname,
                      (unsigned long long) now(CLOCK_REALTIME),
-                     random_ull()) < 0)
+                     random_u64()) < 0)
                 return -ENOMEM;
 
         r = rename(fname, p);
-        free(p);
         if (r < 0)
                 return -errno;
 
-        log_warning("File %s corrupted, renaming and replacing.", fname);
-
-        return journal_file_open(fname, flags, mode, metrics, 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;
+        log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
 
-                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;
+        return journal_file_open(fname, flags, mode, compress, seal,
+                                 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) {
@@ -2374,12 +2743,9 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
         ts.monotonic = le64toh(o->entry.monotonic);
         ts.realtime = le64toh(o->entry.realtime);
 
-        if (to->tail_entry_monotonic_valid &&
-            ts.monotonic < le64toh(to->header->tail_entry_monotonic))
-                return -EINVAL;
-
         n = journal_file_entry_n_items(o);
-        items = alloca(sizeof(EntryItem) * n);
+        /* alloca() can't take 0, hence let's allocate at least one */
+        items = alloca(sizeof(EntryItem) * MAX(1u, n));
 
         for (i = 0; i < n; i++) {
                 uint64_t l, h;
@@ -2405,18 +2771,16 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
                 if ((uint64_t) t != l)
                         return -E2BIG;
 
-                if (o->object.flags & OBJECT_COMPRESSED) {
-#ifdef HAVE_XZ
+                if (o->object.flags & OBJECT_COMPRESSION_MASK) {
                         uint64_t rsize;
 
-                        if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize))
-                                return -EBADMSG;
+                        r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK,
+                                            o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0);
+                        if (r < 0)
+                                return r;
 
                         data = from->compress_buffer;
                         l = rsize;
-#else
-                        return -EPROTONOSUPPORT;
-#endif
                 } else
                         data = o->data.payload;
 
@@ -2495,7 +2859,7 @@ 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;
@@ -2504,11 +2868,11 @@ void journal_default_metrics(JournalMetrics *m, int fd) {
                         m->keep_free = DEFAULT_KEEP_FREE;
         }
 
-        log_info("Fixed max_use=%s max_size=%s min_size=%s keep_free=%s",
-                 format_bytes(a, sizeof(a), m->max_use),
-                 format_bytes(b, sizeof(b), m->max_size),
-                 format_bytes(c, sizeof(c), m->min_size),
-                 format_bytes(d, sizeof(d), m->keep_free));
+        log_debug("Fixed max_use=%s max_size=%s min_size=%s keep_free=%s",
+                  format_bytes(a, sizeof(a), m->max_use),
+                  format_bytes(b, sizeof(b), m->max_size),
+                  format_bytes(c, sizeof(c), m->min_size),
+                  format_bytes(d, sizeof(d), m->keep_free));
 }
 
 int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *to) {
@@ -2533,7 +2897,6 @@ 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;
@@ -2541,9 +2904,7 @@ 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;
 
@@ -2577,13 +2938,15 @@ int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, u
         return 1;
 }
 
-bool journal_file_rotate_suggested(JournalFile *f) {
+bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) {
         assert(f);
 
         /* If we gained new header fields we gained new features,
          * hence suggest a rotation */
-        if (le64toh(f->header->header_size) < sizeof(Header))
+        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
@@ -2592,12 +2955,43 @@ bool journal_file_rotate_suggested(JournalFile *f) {
          * 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)
+                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 (%"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))),
+                                  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)
+                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 (%"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))),
+                                  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;
+
+                h = le64toh(f->header->head_entry_realtime);
+                t = now(CLOCK_REALTIME);
+
+                if (h > 0 && t > h + max_file_usec)
+                        return true;
+        }
 
         return false;
 }