chiark / gitweb /
journal: implement multiple field matches
authorLennart Poettering <lennart@poettering.net>
Tue, 8 Nov 2011 17:20:03 +0000 (18:20 +0100)
committerLennart Poettering <lennart@poettering.net>
Tue, 8 Nov 2011 17:20:03 +0000 (18:20 +0100)
src/journal/journal-def.h
src/journal/journal-file.c
src/journal/journal-file.h
src/journal/journalctl.c
src/journal/journald.c
src/journal/sd-journal.c
src/journal/sd-journal.h
src/journal/test-journal.c
src/journal/wjournal.c [deleted file]
src/journal/wjournal.h [deleted file]

index 5cb1e6d9c849fce693ef281889f1eab9b7e347c7..1a63ca1f8d20eb078af22615e764ad1925b6e05c 100644 (file)
@@ -31,9 +31,10 @@ typedef struct Header Header;
 typedef struct ObjectHeader ObjectHeader;
 typedef union Object Object;
 typedef struct DataObject DataObject;
+typedef struct FieldObject FieldObject;
 typedef struct EntryObject EntryObject;
 typedef struct HashTableObject HashTableObject;
-typedef struct BisectTableObject BisectTableObject;
+typedef struct EntryArrayObject EntryArrayObject;
 typedef struct EntryItem EntryItem;
 typedef struct HashItem HashItem;
 
@@ -41,9 +42,12 @@ typedef struct HashItem HashItem;
 enum {
         OBJECT_UNUSED,
         OBJECT_DATA,
+        OBJECT_FIELD,
         OBJECT_ENTRY,
-        OBJECT_HASH_TABLE,
-        OBJECT_BISECT_TABLE
+        OBJECT_DATA_HASH_TABLE,
+        OBJECT_FIELD_HASH_TABLE,
+        OBJECT_ENTRY_ARRAY,
+        _OBJECT_TYPE_MAX
 };
 
 _packed_ struct ObjectHeader {
@@ -56,18 +60,26 @@ _packed_ struct ObjectHeader {
 _packed_ struct DataObject {
         ObjectHeader object;
         uint64_t hash;
-        uint64_t head_entry_offset;
-        uint64_t tail_entry_offset;
-        uint64_t prev_hash_offset;
         uint64_t next_hash_offset;
+        uint64_t next_field_offset;
+        uint64_t entry_offset; /* the first array entry we store inline */
+        uint64_t entry_array_offset;
+        uint64_t n_entries;
+        uint8_t payload[];
+};
+
+_packed_ struct FieldObject {
+        ObjectHeader object;
+        uint64_t hash;
+        uint64_t next_hash_offset;
+        uint64_t head_data_offset;
+        uint64_t tail_data_offset;
         uint8_t payload[];
 };
 
 _packed_ struct EntryItem {
         uint64_t object_offset;
         uint64_t hash;
-        uint64_t prev_entry_offset;
-        uint64_t next_entry_offset;
 };
 
 _packed_ struct EntryObject {
@@ -77,8 +89,6 @@ _packed_ struct EntryObject {
         uint64_t monotonic;
         sd_id128_t boot_id;
         uint64_t xor_hash;
-        uint64_t prev_entry_offset;
-        uint64_t next_entry_offset;
         EntryItem items[];
 };
 
@@ -89,20 +99,22 @@ _packed_ struct HashItem {
 
 _packed_ struct HashTableObject {
         ObjectHeader object;
-        HashItem table[];
+        HashItem items[];
 };
 
-_packed_ struct BisectTableObject {
+_packed_ struct EntryArrayObject {
         ObjectHeader object;
-        uint64_t table[];
+        uint64_t next_entry_array_offset;
+        uint64_t items[];
 };
 
 union Object {
         ObjectHeader object;
         DataObject data;
+        FieldObject field;
         EntryObject entry;
         HashTableObject hash_table;
-        BisectTableObject bisect_table;
+        EntryArrayObject entry_array;
 };
 
 enum {
@@ -115,30 +127,30 @@ _packed_ struct Header {
         uint8_t signature[8]; /* "LPKSHHRH" */
         uint32_t compatible_flags;
         uint32_t incompatible_flags;
-        uint32_t state;
-        uint8_t reserved[4];
+        uint8_t state;
+        uint8_t reserved[7];
         sd_id128_t file_id;
         sd_id128_t machine_id;
         sd_id128_t boot_id;
         sd_id128_t seqnum_id;
         uint64_t arena_offset;
         uint64_t arena_size;
-        uint64_t arena_max_size;
-        uint64_t arena_min_size;
-        uint64_t arena_keep_free;
-        uint64_t hash_table_offset;     /* for looking up data objects */
-        uint64_t hash_table_size;
-        uint64_t bisect_table_offset;   /* for looking up entry objects */
-        uint64_t bisect_table_size;
-        uint64_t head_object_offset;
+        uint64_t arena_max_size;  /* obsolete */
+        uint64_t arena_min_size;  /* obsolete */
+        uint64_t arena_keep_free; /* obsolete */
+        uint64_t data_hash_table_offset;     /* for looking up data objects */
+        uint64_t data_hash_table_size;
+        uint64_t field_hash_table_offset;     /* for looking up field objects */
+        uint64_t field_hash_table_size;
         uint64_t tail_object_offset;
-        uint64_t head_entry_offset;
-        uint64_t tail_entry_offset;
-        uint64_t last_bisect_offset;
         uint64_t n_objects;
+        uint64_t n_entries;
         uint64_t seqnum;
+        uint64_t first_seqnum;
+        uint64_t entry_array_offset;
         uint64_t head_entry_realtime;
         uint64_t tail_entry_realtime;
+        uint64_t tail_entry_monotonic;
 };
 
 #endif
index ab45132f0f1a142d87dafa9c8b1cc91633e21156..427631d30a4730b49770b34106acb6218d5f3f5f 100644 (file)
@@ -37,8 +37,8 @@
 
 #define DEFAULT_MAX_USE (16ULL*1024ULL*1024ULL*16ULL)
 
-#define DEFAULT_HASH_TABLE_SIZE (2047ULL*16ULL)
-#define DEFAULT_BISECT_TABLE_SIZE ((DEFAULT_ARENA_MAX_SIZE/(64ULL*1024ULL))*8ULL)
+#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*16ULL)
+#define DEFAULT_FIELD_HASH_TABLE_SIZE (2047ULL*16ULL)
 
 #define DEFAULT_WINDOW_SIZE (128ULL*1024ULL*1024ULL)
 
@@ -47,23 +47,17 @@ static const char signature[] = { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' };
 #define ALIGN64(x) (((x) + 7ULL) & ~7ULL)
 
 void journal_file_close(JournalFile *f) {
-        assert(f);
-
-        if (f->header) {
-                if (f->writable && f->header->state == htole32(STATE_ONLINE))
-                        f->header->state = htole32(STATE_OFFLINE);
+        int t;
 
-                munmap(f->header, PAGE_ALIGN(sizeof(Header)));
-        }
+        assert(f);
 
-        if (f->hash_table_window)
-                munmap(f->hash_table_window, f->hash_table_window_size);
+        if (f->header && f->writable)
+                f->header->state = STATE_OFFLINE;
 
-        if (f->bisect_table_window)
-                munmap(f->bisect_table_window, f->bisect_table_window_size);
 
-        if (f->window)
-                munmap(f->window, f->window_size);
+        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);
@@ -108,6 +102,7 @@ static int journal_file_init_header(JournalFile *f, JournalFile *template) {
 
 static int journal_file_refresh_header(JournalFile *f) {
         int r;
+        sd_id128_t boot_id;
 
         assert(f);
 
@@ -115,11 +110,16 @@ static int journal_file_refresh_header(JournalFile *f) {
         if (r < 0)
                 return r;
 
-        r = sd_id128_get_boot(&f->header->boot_id);
+        r = sd_id128_get_boot(&boot_id);
         if (r < 0)
                 return r;
 
-        f->header->state = htole32(STATE_ONLINE);
+        if (sd_id128_equal(boot_id, f->header->boot_id))
+                f->tail_entry_monotonic_valid = true;
+
+        f->header->boot_id = boot_id;
+
+        f->header->state = STATE_ONLINE;
         return 0;
 }
 
@@ -147,7 +147,7 @@ static int journal_file_verify_header(JournalFile *f) {
                 if (!sd_id128_equal(machine_id, f->header->machine_id))
                         return -EHOSTDOWN;
 
-                state = le32toh(f->header->state);
+                state = f->header->state;
 
                 if (state == STATE_ONLINE)
                         log_debug("Journal file %s is already online. Assuming unclean closing. Ignoring.", f->path);
@@ -254,28 +254,33 @@ static int journal_file_map(
         return 0;
 }
 
-static int journal_file_move_to(JournalFile *f, uint64_t offset, uint64_t size, void **ret) {
+static int journal_file_move_to(JournalFile *f, int wt, uint64_t offset, uint64_t size, void **ret) {
         void *p;
         uint64_t delta;
         int r;
+        Window *w;
 
         assert(f);
         assert(ret);
+        assert(wt >= 0);
+        assert(wt < _WINDOW_MAX);
 
-        if (_likely_(f->window &&
-                     f->window_offset <= offset &&
-                     f->window_offset+f->window_size >= offset + size)) {
+        w = f->windows + wt;
 
-                *ret = (uint8_t*) f->window + (offset - f->window_offset);
+        if (_likely_(w->ptr &&
+                     w->offset <= offset &&
+                     w->offset + w->size >= offset + size)) {
+
+                *ret = (uint8_t*) w->ptr + (offset - w->offset);
                 return 0;
         }
 
-        if (f->window) {
-                if (munmap(f->window, f->window_size) < 0)
+        if (w->ptr) {
+                if (munmap(w->ptr, w->size) < 0)
                         return -errno;
 
-                f->window = NULL;
-                f->window_size = f->window_offset = 0;
+                w->ptr = NULL;
+                w->size = w->offset = 0;
         }
 
         if (size < DEFAULT_WINDOW_SIZE) {
@@ -297,8 +302,8 @@ static int journal_file_move_to(JournalFile *f, uint64_t offset, uint64_t size,
 
         r = journal_file_map(f,
                              offset, size,
-                             &f->window, &f->window_offset, &f->window_size,
-                             & p);
+                             &w->ptr, &w->offset, &w->size,
+                             &p);
 
         if (r < 0)
                 return r;
@@ -308,26 +313,23 @@ static int journal_file_move_to(JournalFile *f, uint64_t offset, uint64_t size,
 }
 
 static bool verify_hash(Object *o) {
-        uint64_t t;
+        uint64_t h1, h2;
 
         assert(o);
 
-        t = le64toh(o->object.type);
-        if (t == OBJECT_DATA) {
-                uint64_t s, h1, h2;
-
-                s = le64toh(o->object.size);
-
+        if (o->object.type == OBJECT_DATA) {
                 h1 = le64toh(o->data.hash);
-                h2 = hash64(o->data.payload, s - offsetof(Object, data.payload));
-
-                return h1 == h2;
-        }
+                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;
 
-        return true;
+        return h1 == h2;
 }
 
-int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Object **ret) {
+int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) {
         int r;
         void *t;
         Object *o;
@@ -335,8 +337,9 @@ int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Objec
 
         assert(f);
         assert(ret);
+        assert(type < _OBJECT_TYPE_MAX);
 
-        r = journal_file_move_to(f, offset, sizeof(ObjectHeader), &t);
+        r = journal_file_move_to(f, type >= 0 ? type : WINDOW_UNKNOWN, offset, sizeof(ObjectHeader), &t);
         if (r < 0)
                 return r;
 
@@ -346,11 +349,11 @@ int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Objec
         if (s < sizeof(ObjectHeader))
                 return -EBADMSG;
 
-        if (type >= 0 && le64toh(o->object.type) != type)
+        if (type >= 0 && o->object.type != type)
                 return -EBADMSG;
 
         if (s > sizeof(ObjectHeader)) {
-                r = journal_file_move_to(f, offset, s, &t);
+                r = journal_file_move_to(f, o->object.type, offset, s, &t);
                 if (r < 0)
                         return r;
 
@@ -372,7 +375,7 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
         r = le64toh(f->header->seqnum) + 1;
 
         if (seqnum) {
-                /* If an external seqno counter was passed, we update
+                /* If an external seqnum counter was passed, we update
                  * both the local and the external one, and set it to
                  * the maximum of both */
 
@@ -384,10 +387,13 @@ static uint64_t journal_file_seqnum(JournalFile *f, uint64_t *seqnum) {
 
         f->header->seqnum = htole64(r);
 
+        if (f->header->first_seqnum == 0)
+                f->header->first_seqnum = htole64(r);
+
         return r;
 }
 
-static int journal_file_append_object(JournalFile *f, uint64_t size, Object **ret, uint64_t *offset) {
+static 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;
@@ -399,11 +405,10 @@ static int journal_file_append_object(JournalFile *f, uint64_t size, Object **re
         assert(ret);
 
         p = le64toh(f->header->tail_object_offset);
-
         if (p == 0)
                 p = le64toh(f->header->arena_offset);
         else {
-                r = journal_file_move_to_object(f, p, -1, &tail);
+                r = journal_file_move_to_object(f, -1, p, &tail);
                 if (r < 0)
                         return r;
 
@@ -414,21 +419,17 @@ static int journal_file_append_object(JournalFile *f, uint64_t size, Object **re
         if (r < 0)
                 return r;
 
-        r = journal_file_move_to(f, p, size, &t);
+        r = journal_file_move_to(f, type, p, size, &t);
         if (r < 0)
                 return r;
 
         o = (Object*) t;
 
         zero(o->object);
-        o->object.type = htole64(OBJECT_UNUSED);
-        zero(o->object.reserved);
+        o->object.type = type;
         o->object.size = htole64(size);
 
         f->header->tail_object_offset = htole64(p);
-        if (f->header->head_object_offset == 0)
-                f->header->head_object_offset = htole64(p);
-
         f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1);
 
         *ret = o;
@@ -437,135 +438,137 @@ static int journal_file_append_object(JournalFile *f, uint64_t size, Object **re
         return 0;
 }
 
-static int journal_file_setup_hash_table(JournalFile *f) {
+static int journal_file_setup_data_hash_table(JournalFile *f) {
         uint64_t s, p;
         Object *o;
         int r;
 
         assert(f);
 
-        s = DEFAULT_HASH_TABLE_SIZE;
-        r = journal_file_append_object(f, offsetof(Object, hash_table.table) + s, &o, &p);
+        s = DEFAULT_DATA_HASH_TABLE_SIZE;
+        r = journal_file_append_object(f,
+                                       OBJECT_DATA_HASH_TABLE,
+                                       offsetof(Object, hash_table.items) + s,
+                                       &o, &p);
         if (r < 0)
                 return r;
 
-        o->object.type = htole64(OBJECT_HASH_TABLE);
-        memset(o->hash_table.table, 0, s);
+        memset(o->hash_table.items, 0, s);
 
-        f->header->hash_table_offset = htole64(p + offsetof(Object, hash_table.table));
-        f->header->hash_table_size = htole64(s);
+        f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
+        f->header->data_hash_table_size = htole64(s);
 
         return 0;
 }
 
-static int journal_file_setup_bisect_table(JournalFile *f) {
+static int journal_file_setup_field_hash_table(JournalFile *f) {
         uint64_t s, p;
         Object *o;
         int r;
 
         assert(f);
 
-        s = DEFAULT_BISECT_TABLE_SIZE;
-        r = journal_file_append_object(f, offsetof(Object, bisect_table.table) + s, &o, &p);
+        s = DEFAULT_FIELD_HASH_TABLE_SIZE;
+        r = journal_file_append_object(f,
+                                       OBJECT_FIELD_HASH_TABLE,
+                                       offsetof(Object, hash_table.items) + s,
+                                       &o, &p);
         if (r < 0)
                 return r;
 
-        o->object.type = htole64(OBJECT_BISECT_TABLE);
-        memset(o->bisect_table.table, 0, s);
+        memset(o->hash_table.items, 0, s);
 
-        f->header->bisect_table_offset = htole64(p + offsetof(Object, bisect_table.table));
-        f->header->bisect_table_size = htole64(s);
+        f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
+        f->header->field_hash_table_size = htole64(s);
 
         return 0;
 }
 
-static int journal_file_map_hash_table(JournalFile *f) {
+static int journal_file_map_data_hash_table(JournalFile *f) {
         uint64_t s, p;
         void *t;
         int r;
 
         assert(f);
 
-        p = le64toh(f->header->hash_table_offset);
-        s = le64toh(f->header->hash_table_size);
+        p = le64toh(f->header->data_hash_table_offset);
+        s = le64toh(f->header->data_hash_table_size);
 
-        r = journal_file_map(f,
-                             p, s,
-                             &f->hash_table_window, NULL, &f->hash_table_window_size,
-                             &t);
+        r = journal_file_move_to(f,
+                                 WINDOW_DATA_HASH_TABLE,
+                                 p, s,
+                                 &t);
         if (r < 0)
                 return r;
 
-        f->hash_table = t;
+        f->data_hash_table = t;
         return 0;
 }
 
-static int journal_file_map_bisect_table(JournalFile *f) {
+static int journal_file_map_field_hash_table(JournalFile *f) {
         uint64_t s, p;
         void *t;
         int r;
 
         assert(f);
 
-        p = le64toh(f->header->bisect_table_offset);
-        s = le64toh(f->header->bisect_table_size);
-
-        r = journal_file_map(f,
-                             p, s,
-                             &f->bisect_table_window, NULL, &f->bisect_table_window_size,
-                             &t);
+        p = le64toh(f->header->field_hash_table_offset);
+        s = le64toh(f->header->field_hash_table_size);
 
+        r = journal_file_move_to(f,
+                                 WINDOW_FIELD_HASH_TABLE,
+                                 p, s,
+                                 &t);
         if (r < 0)
                 return r;
 
-        f->bisect_table = t;
+        f->field_hash_table = t;
         return 0;
 }
 
-static int journal_file_link_data(JournalFile *f, Object *o, uint64_t offset, uint64_t hash_index) {
-        uint64_t p;
+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 == htole64(OBJECT_DATA));
+        assert(o->object.type == OBJECT_DATA);
 
-        o->data.head_entry_offset = o->data.tail_entry_offset = 0;
-        o->data.next_hash_offset = 0;
+        o->data.next_hash_offset = o->data.next_field_offset = 0;
+        o->data.entry_offset = o->data.entry_array_offset = 0;
+        o->data.n_entries = 0;
 
-        p = le64toh(f->hash_table[hash_index].tail_hash_offset);
+        h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
+        p = le64toh(f->data_hash_table[h].head_hash_offset);
         if (p == 0) {
                 /* Only entry in the hash table is easy */
-
-                o->data.prev_hash_offset = 0;
-                f->hash_table[hash_index].head_hash_offset = htole64(offset);
+                f->data_hash_table[h].head_hash_offset = htole64(offset);
         } else {
-                o->data.prev_hash_offset = htole64(p);
-
                 /* Temporarily move back to the previous data object,
                  * to patch in pointer */
 
-                r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+                r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
                 if (r < 0)
                         return r;
 
-                o->data.next_hash_offset = offset;
+                o->data.next_hash_offset = htole64(offset);
 
-                r = journal_file_move_to_object(f, offset, OBJECT_DATA, &o);
+                r = journal_file_move_to_object(f, OBJECT_DATA, offset, &o);
                 if (r < 0)
                         return r;
         }
 
-        f->hash_table[hash_index].tail_hash_offset = htole64(offset);
+        f->data_hash_table[h].tail_hash_offset = htole64(offset);
 
         return 0;
 }
 
-static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
-        uint64_t hash, h, p, np;
-        uint64_t osize;
-        Object *o;
+int journal_file_find_data_object_with_hash(
+                JournalFile *f,
+                const void *data, uint64_t size, uint64_t hash,
+                Object **ret, uint64_t *offset) {
+        uint64_t p, osize, h;
         int r;
 
         assert(f);
@@ -573,14 +576,13 @@ static int journal_file_append_data(JournalFile *f, const void *data, uint64_t s
 
         osize = offsetof(Object, data.payload) + size;
 
-        hash = hash64(data, size);
-        h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
-        p = le64toh(f->hash_table[h].head_hash_offset);
+        h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
+        p = le64toh(f->data_hash_table[h].head_hash_offset);
 
-        while (p != 0) {
-                /* Look for this data object in the hash table */
+        while (p > 0) {
+                Object *o;
 
-                r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+                r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
                 if (r < 0)
                         return r;
 
@@ -596,21 +598,66 @@ static int journal_file_append_data(JournalFile *f, const void *data, uint64_t s
                         if (offset)
                                 *offset = p;
 
-                        return 0;
+                        return 1;
                 }
 
                 p = le64toh(o->data.next_hash_offset);
         }
 
-        r = journal_file_append_object(f, osize, &o, &np);
+        return 0;
+}
+
+int journal_file_find_data_object(
+                JournalFile *f,
+                const void *data, uint64_t size,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t hash;
+
+        assert(f);
+        assert(data || size == 0);
+
+        hash = hash64(data, size);
+
+        return journal_file_find_data_object_with_hash(f,
+                                                       data, size, hash,
+                                                       ret, offset);
+}
+
+static int journal_file_append_data(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset) {
+        uint64_t hash, p;
+        uint64_t osize;
+        Object *o;
+        int r;
+
+        assert(f);
+        assert(data || size == 0);
+
+        hash = hash64(data, size);
+
+        r = journal_file_find_data_object_with_hash(f, data, 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, data.payload) + size;
+        r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p);
         if (r < 0)
                 return r;
 
-        o->object.type = htole64(OBJECT_DATA);
         o->data.hash = htole64(hash);
         memcpy(o->data.payload, data, size);
 
-        r = journal_file_link_data(f, o, np, h);
+        r = journal_file_link_data(f, o, p, hash);
         if (r < 0)
                 return r;
 
@@ -618,7 +665,7 @@ static int journal_file_append_data(JournalFile *f, const void *data, uint64_t s
                 *ret = o;
 
         if (offset)
-                *offset = np;
+                *offset = p;
 
         return 0;
 }
@@ -630,92 +677,154 @@ uint64_t journal_file_entry_n_items(Object *o) {
         return (le64toh(o->object.size) - offsetof(Object, entry.items)) / sizeof(EntryItem);
 }
 
-static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
-        uint64_t p, q;
+static uint64_t journal_file_entry_array_n_items(Object *o) {
+        assert(o);
+        assert(o->object.type == htole64(OBJECT_ENTRY_ARRAY));
+
+        return (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / sizeof(uint64_t);
+}
+
+static int link_entry_into_array(JournalFile *f,
+                                 uint64_t *first,
+                                 uint64_t *idx,
+                                 uint64_t p) {
         int r;
+        uint64_t n = 0, ap = 0, q, i, a, hidx;
+        Object *o;
+
         assert(f);
-        assert(o);
-        assert(offset > 0);
+        assert(first);
+        assert(idx);
+        assert(p > 0);
 
-        p = le64toh(o->entry.items[i].object_offset);
-        if (p == 0)
-                return -EINVAL;
+        a = le64toh(*first);
+        i = hidx = le64toh(*idx);
+        while (a > 0) {
+
+                r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
+                if (r < 0)
+                        return r;
 
-        o->entry.items[i].next_entry_offset = 0;
+                n = journal_file_entry_array_n_items(o);
+                if (i < n) {
+                        o->entry_array.items[i] = htole64(p);
+                        *idx = htole64(hidx + 1);
+                        return 0;
+                }
 
-        /* Move to the data object */
-        r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+                i -= n;
+                ap = a;
+                a = le64toh(o->entry_array.next_entry_array_offset);
+        }
+
+        if (hidx > n)
+                n = (hidx+1) * 2;
+        else
+                n = n * 2;
+
+        if (n < 4)
+                n = 4;
+
+        r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY,
+                                       offsetof(Object, entry_array.items) + n * sizeof(uint64_t),
+                                       &o, &q);
         if (r < 0)
                 return r;
 
-        q = le64toh(o->data.tail_entry_offset);
-        o->data.tail_entry_offset = htole64(offset);
+        o->entry_array.items[i] = htole64(p);
 
-        if (q == 0)
-                o->data.head_entry_offset = htole64(offset);
+        if (ap == 0)
+                *first = q;
         else {
-                uint64_t n, j;
-
-                /* Move to previous entry */
-                r = journal_file_move_to_object(f, q, OBJECT_ENTRY, &o);
+                r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o);
                 if (r < 0)
                         return r;
 
-                n = journal_file_entry_n_items(o);
-                for (j = 0; j < n; j++)
-                        if (le64toh(o->entry.items[j].object_offset) == p)
-                                break;
+                o->entry_array.next_entry_array_offset = htole64(q);
+        }
 
-                if (j >= n)
-                        return -EBADMSG;
+        *idx = htole64(hidx + 1);
+
+        return 0;
+}
 
-                o->entry.items[j].next_entry_offset = offset;
+static int link_entry_into_array_plus_one(JournalFile *f,
+                                          uint64_t *extra,
+                                          uint64_t *first,
+                                          uint64_t *idx,
+                                          uint64_t p) {
+
+        int r;
+
+        assert(f);
+        assert(extra);
+        assert(first);
+        assert(idx);
+        assert(p > 0);
+
+        if (*idx == 0)
+                *extra = htole64(p);
+        else {
+                uint64_t i;
+
+                i = le64toh(*idx) - 1;
+                r = link_entry_into_array(f, first, &i, p);
+                if (r < 0)
+                        return r;
         }
 
-        /* Move back to original entry */
-        r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o);
+        *idx = htole64(le64toh(*idx) + 1);
+        return 0;
+}
+
+static int journal_file_link_entry_item(JournalFile *f, Object *o, uint64_t offset, uint64_t i) {
+        uint64_t p;
+        int r;
+        assert(f);
+        assert(o);
+        assert(offset > 0);
+
+        p = le64toh(o->entry.items[i].object_offset);
+        if (p == 0)
+                return -EINVAL;
+
+        r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
         if (r < 0)
                 return r;
 
-        o->entry.items[i].prev_entry_offset = q;
-        return 0;
+        return link_entry_into_array_plus_one(f,
+                                              &o->data.entry_offset,
+                                              &o->data.entry_array_offset,
+                                              &o->data.n_entries,
+                                              offset);
 }
 
 static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
-        uint64_t p, i, n, k, a, b;
+        uint64_t n, i;
         int r;
 
         assert(f);
         assert(o);
         assert(offset > 0);
-        assert(o->object.type == htole64(OBJECT_ENTRY));
+        assert(o->object.type == OBJECT_ENTRY);
 
         /* Link up the entry itself */
-        p = le64toh(f->header->tail_entry_offset);
+        r = link_entry_into_array(f,
+                                  &f->header->entry_array_offset,
+                                  &f->header->n_entries,
+                                  offset);
+        if (r < 0)
+                return r;
 
-        o->entry.prev_entry_offset = f->header->tail_entry_offset;
-        o->entry.next_entry_offset = 0;
+        log_error("%s %lu", f->path, (unsigned long) f->header->n_entries);
 
-        if (p == 0) {
-                f->header->head_entry_offset = htole64(offset);
+        if (f->header->head_entry_realtime == 0)
                 f->header->head_entry_realtime = o->entry.realtime;
-        } else {
-                /* Temporarily move back to the previous entry, to
-                 * patch in pointer */
-
-                r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
-                if (r < 0)
-                        return r;
-
-                o->entry.next_entry_offset = htole64(offset);
 
-                r = journal_file_move_to_object(f, offset, OBJECT_ENTRY, &o);
-                if (r < 0)
-                        return r;
-        }
-
-        f->header->tail_entry_offset = htole64(offset);
         f->header->tail_entry_realtime = o->entry.realtime;
+        f->header->tail_entry_monotonic = o->entry.monotonic;
+
+        f->tail_entry_monotonic_valid = true;
 
         /* Link up the items */
         n = journal_file_entry_n_items(o);
@@ -725,18 +834,6 @@ static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
                         return r;
         }
 
-        /* Link up the entry in the bisect table */
-        n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t);
-        k = le64toh(f->header->arena_max_size) / n;
-
-        a = (le64toh(f->header->last_bisect_offset) + k - 1) / k;
-        b = offset / k;
-
-        for (; a <= b; a++)
-                f->bisect_table[a] = htole64(offset);
-
-        f->header->last_bisect_offset = htole64(offset + le64toh(o->object.size));
-
         return 0;
 }
 
@@ -745,7 +842,7 @@ static int journal_file_append_entry_internal(
                 const dual_timestamp *ts,
                 uint64_t xor_hash,
                 const EntryItem items[], unsigned n_items,
-                uint64_t *seqno,
+                uint64_t *seqnum,
                 Object **ret, uint64_t *offset) {
         uint64_t np;
         uint64_t osize;
@@ -754,18 +851,18 @@ static int journal_file_append_entry_internal(
 
         assert(f);
         assert(items || n_items == 0);
+        assert(ts);
 
         osize = offsetof(Object, entry.items) + (n_items * sizeof(EntryItem));
 
-        r = journal_file_append_object(f, osize, &o, &np);
+        r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np);
         if (r < 0)
                 return r;
 
-        o->object.type = htole64(OBJECT_ENTRY);
-        o->entry.seqnum = htole64(journal_file_seqnum(f, seqno));
+        o->entry.seqnum = htole64(journal_file_seqnum(f, seqnum));
         memcpy(o->entry.items, items, n_items * sizeof(EntryItem));
-        o->entry.realtime = htole64(ts ? ts->realtime : now(CLOCK_REALTIME));
-        o->entry.monotonic = htole64(ts ? ts->monotonic : now(CLOCK_MONOTONIC));
+        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;
 
@@ -782,15 +879,31 @@ static int journal_file_append_entry_internal(
         return 0;
 }
 
-int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqno, Object **ret, uint64_t *offset) {
+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) {
         unsigned i;
         EntryItem *items;
         int r;
         uint64_t xor_hash = 0;
+        struct dual_timestamp _ts;
 
         assert(f);
         assert(iovec || n_iovec == 0);
 
+        if (!f->writable)
+                return -EPERM;
+
+        if (!ts) {
+                dual_timestamp_get(&_ts);
+                ts = &_ts;
+        }
+
+        if (f->tail_entry_monotonic_valid &&
+            ts->monotonic < le64toh(f->header->tail_entry_monotonic))
+                return -EINVAL;
+
+        if (ts->realtime < le64toh(f->header->tail_entry_realtime))
+                return -EINVAL;
+
         items = new(EntryItem, n_iovec);
         if (!items)
                 return -ENOMEM;
@@ -808,7 +921,7 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
                 items[i].hash = o->data.hash;
         }
 
-        r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqno, ret, offset);
+        r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
 
 finish:
         free(items);
@@ -816,198 +929,590 @@ finish:
         return r;
 }
 
-int journal_file_move_to_entry(JournalFile *f, uint64_t seqnum, Object **ret, uint64_t *offset) {
+static int generic_array_get(JournalFile *f,
+                             uint64_t first,
+                             uint64_t i,
+                             Object **ret, uint64_t *offset) {
+
         Object *o;
-        uint64_t lower, upper, p, n, k;
+        uint64_t p, a;
         int r;
 
         assert(f);
 
-        n = le64toh(f->header->bisect_table_size) / sizeof(uint64_t);
-        k = le64toh(f->header->arena_max_size) / n;
+        a = first;
+        while (a > 0) {
+                uint64_t n;
 
-        lower = 0;
-        upper = le64toh(f->header->last_bisect_offset)/k+1;
-
-        while (lower < upper) {
-                k = (upper + lower) / 2;
-                p = le64toh(f->bisect_table[k]);
+                r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
+                if (r < 0)
+                        return r;
 
-                if (p == 0) {
-                        upper = k;
-                        continue;
+                n = journal_file_entry_array_n_items(o);
+                if (i < n) {
+                        p = le64toh(o->entry_array.items[i]);
+                        break;
                 }
 
-                r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
+                i -= n;
+                a = le64toh(o->entry_array.next_entry_array_offset);
+        }
+
+        if (a <= 0 || p <= 0)
+                return 0;
+
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+        if (r < 0)
+                return r;
+
+        if (ret)
+                *ret = o;
+
+        if (offset)
+                *offset = p;
+
+        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) {
+
+        Object *o;
+
+        assert(f);
+
+        if (i == 0) {
+                int r;
+
+                r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
                 if (r < 0)
                         return r;
 
-                if (o->entry.seqnum == seqnum) {
-                        if (ret)
-                                *ret = o;
+                if (ret)
+                        *ret = o;
 
-                        if (offset)
-                                *offset = p;
+                if (offset)
+                        *offset = extra;
 
-                        return 1;
-                } else if (seqnum < o->entry.seqnum)
-                        upper = k;
-                else if (seqnum > o->entry.seqnum)
-                        lower = k+1;
+                return 1;
         }
 
-        assert(lower == upper);
+        return generic_array_get(f, first, i-1, ret, offset);
+}
 
-        if (lower <= 0)
-                return 0;
+enum {
+        TEST_FOUND,
+        TEST_LEFT,
+        TEST_RIGHT
+};
 
-        /* The object we are looking for is between
-         * bisect_table[lower-1] and bisect_table[lower] */
+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;
+        bool subtract_one = false;
+        Object *o, *array = NULL;
+        int r;
 
-        p = le64toh(f->bisect_table[lower-1]);
+        assert(f);
+        assert(test_object);
 
-        for (;;) {
-                r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
+        a = first;
+        while (a > 0) {
+                uint64_t left, right, k, lp;
+
+                r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array);
                 if (r < 0)
                         return r;
 
-                if (o->entry.seqnum == seqnum) {
-                        if (ret)
-                                *ret = o;
+                k = journal_file_entry_array_n_items(array);
+                right = MIN(k, n);
+                if (right <= 0)
+                        return 0;
 
-                        if (offset)
-                                *offset = p;
+                i = right - 1;
+                lp = p = le64toh(array->entry_array.items[i]);
+                if (p <= 0)
+                        return -EBADMSG;
 
-                        return 1;
+                r = test_object(f, p, needle);
+                if (r < 0)
+                        return r;
 
-                } if (seqnum < o->entry.seqnum)
-                        return 0;
+                if (r == TEST_FOUND)
+                        r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
+
+                if (r == TEST_RIGHT) {
+                        left = 0;
+                        right -= 1;
+                        for (;;) {
+                                if (left == right) {
+                                        if (direction == DIRECTION_UP)
+                                                subtract_one = true;
+
+                                        i = left;
+                                        goto found;
+                                }
+
+                                assert(left < right);
+
+                                i = (left + right) / 2;
+                                p = le64toh(array->entry_array.items[i]);
+                                if (p <= 0)
+                                        return -EBADMSG;
+
+                                r = test_object(f, p, needle);
+                                if (r < 0)
+                                        return r;
 
-                if (o->entry.next_entry_offset == 0)
+                                if (r == TEST_FOUND)
+                                        r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
+
+                                if (r == TEST_RIGHT)
+                                        right = i;
+                                else
+                                        left = i + 1;
+                        }
+                }
+
+                if (k > n)
                         return 0;
 
-                p = le64toh(o->entry.next_entry_offset);
+                last_p = lp;
+
+                n -= k;
+                t += k;
+                a = le64toh(array->entry_array.next_entry_array_offset);
         }
 
         return 0;
+
+found:
+        if (subtract_one && t == 0 && i == 0)
+                return 0;
+
+        if (subtract_one && i == 0)
+                p = last_p;
+        else if (subtract_one)
+                p = le64toh(array->entry_array.items[i-1]);
+        else
+                p = le64toh(array->entry_array.items[i]);
+
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+        if (r < 0)
+                return r;
+
+        if (ret)
+                *ret = o;
+
+        if (offset)
+                *offset = p;
+
+        if (idx)
+                *idx = t + i - (subtract_one ? 1 : 0);
+
+        return 1;
 }
 
-int journal_file_next_entry(JournalFile *f, Object *o, direction_t direction, Object **ret, uint64_t *offset) {
-        uint64_t np;
+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;
 
         assert(f);
+        assert(test_object);
 
-        if (!o)
-                np = le64toh(direction == DIRECTION_DOWN ?
-                             f->header->head_entry_offset :
-                             f->header->tail_entry_offset);
-        else {
-                if (le64toh(o->object.type) != OBJECT_ENTRY)
-                        return -EINVAL;
+        if (n <= 0)
+                return 0;
 
-                np = le64toh(direction == DIRECTION_DOWN ?
-                             o->entry.next_entry_offset :
-                             o->entry.prev_entry_offset);
-        }
+        /* This bisects the array in object 'first', but first checks
+         * an extra  */
+
+        r = test_object(f, extra, needle);
+        if (r < 0)
+                return r;
+        else if (r == TEST_FOUND) {
+                Object *o;
+
+                r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, &o);
+                if (r < 0)
+                        return r;
+
+                if (ret)
+                        *ret = o;
 
-        if (np == 0)
+                if (offset)
+                        *offset = extra;
+        } else if (r == TEST_RIGHT)
                 return 0;
 
-        r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o);
+        r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret, offset, idx);
+
+        if (r > 0)
+                (*idx) ++;
+
+        return r;
+}
+
+static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) {
+        Object *o;
+        int r;
+
+        assert(f);
+        assert(p > 0);
+
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
         if (r < 0)
                 return r;
 
-        if (ret)
-                *ret = o;
+        if (le64toh(o->entry.seqnum) == needle)
+                return TEST_FOUND;
+        else if (le64toh(o->entry.seqnum) < needle)
+                return TEST_LEFT;
+        else
+                return TEST_RIGHT;
+}
 
-        if (offset)
-                *offset = np;
+int journal_file_move_to_entry_by_seqnum(
+                JournalFile *f,
+                uint64_t seqnum,
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset) {
+
+        return generic_array_bisect(f,
+                                    le64toh(f->header->entry_array_offset),
+                                    le64toh(f->header->n_entries),
+                                    seqnum,
+                                    test_object_seqnum,
+                                    direction,
+                                    ret, offset, NULL);
+}
 
-        return 1;
+static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) {
+        Object *o;
+        int r;
+
+        assert(f);
+        assert(p > 0);
+
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+        if (r < 0)
+                return r;
+
+        if (le64toh(o->entry.realtime) == needle)
+                return TEST_FOUND;
+        else if (le64toh(o->entry.realtime) < needle)
+                return TEST_LEFT;
+        else
+                return TEST_RIGHT;
 }
 
-int journal_file_prev_entry(JournalFile *f, Object *o, Object **ret, uint64_t *offset) {
-        uint64_t np;
+int journal_file_move_to_entry_by_realtime(
+                JournalFile *f,
+                uint64_t realtime,
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset) {
+
+        return generic_array_bisect(f,
+                                    le64toh(f->header->entry_array_offset),
+                                    le64toh(f->header->n_entries),
+                                    realtime,
+                                    test_object_realtime,
+                                    direction,
+                                    ret, offset, NULL);
+}
+
+static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) {
+        Object *o;
+        int r;
+
+        assert(f);
+        assert(p > 0);
+
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
+        if (r < 0)
+                return r;
+
+        if (le64toh(o->entry.monotonic) == needle)
+                return TEST_FOUND;
+        else if (le64toh(o->entry.monotonic) < needle)
+                return TEST_LEFT;
+        else
+                return TEST_RIGHT;
+}
+
+int journal_file_move_to_entry_by_monotonic(
+                JournalFile *f,
+                sd_id128_t boot_id,
+                uint64_t monotonic,
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset) {
+
+        char t[8+32+1] = "_BOOT_ID=";
+        Object *o;
+        int r;
+
+        sd_id128_to_string(boot_id, t + 8);
+
+        r = journal_file_find_data_object(f, t, strlen(t), &o, NULL);
+        if (r < 0)
+                return r;
+        else if (r == 0)
+                return -ENOENT;
+
+        return generic_array_bisect_plus_one(f,
+                                             le64toh(o->data.entry_offset),
+                                             le64toh(o->data.entry_array_offset),
+                                             le64toh(o->data.n_entries),
+                                             monotonic,
+                                             test_object_monotonic,
+                                             direction,
+                                             ret, offset, NULL);
+}
+
+static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
+        assert(f);
+        assert(p > 0);
+
+        if (p == needle)
+                return TEST_FOUND;
+        else if (p < needle)
+                return TEST_LEFT;
+        else
+                return TEST_RIGHT;
+}
+
+int journal_file_next_entry(
+                JournalFile *f,
+                Object *o, uint64_t p,
+                direction_t direction,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t i, n;
         int r;
 
         assert(f);
+        assert(p > 0 || !o);
+
+        n = le64toh(f->header->n_entries);
+        if (n <= 0)
+                return 0;
 
         if (!o)
-                np = le64toh(f->header->tail_entry_offset);
+                i = direction == DIRECTION_DOWN ? 0 : n - 1;
         else {
-                if (le64toh(o->object.type) != OBJECT_ENTRY)
+                if (o->object.type != OBJECT_ENTRY)
                         return -EINVAL;
 
-                np = le64toh(o->entry.prev_entry_offset);
+                r = generic_array_bisect(f,
+                                         le64toh(f->header->entry_array_offset),
+                                         le64toh(f->header->n_entries),
+                                         p,
+                                         test_object_offset,
+                                         DIRECTION_DOWN,
+                                         NULL, NULL,
+                                         &i);
+                if (r <= 0)
+                        return r;
+
+                if (direction == DIRECTION_DOWN) {
+                        if (i >= n - 1)
+                                return 0;
+
+                        i++;
+                } else {
+                        if (i <= 0)
+                                return 0;
+
+                        i--;
+                }
         }
 
-        if (np == 0)
-                return 0;
+        /* And jump to it */
+        return generic_array_get(f,
+                                 le64toh(f->header->entry_array_offset),
+                                 i,
+                                 ret, offset);
+}
 
-        r = journal_file_move_to_object(f, np, OBJECT_ENTRY, &o);
-        if (r < 0)
+int journal_file_skip_entry(
+                JournalFile *f,
+                Object *o, uint64_t p,
+                int64_t skip,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t i, n;
+        int r;
+
+        assert(f);
+        assert(o);
+        assert(p > 0);
+
+        if (o->object.type != OBJECT_ENTRY)
+                return -EINVAL;
+
+        r = generic_array_bisect(f,
+                                 le64toh(f->header->entry_array_offset),
+                                 le64toh(f->header->n_entries),
+                                 p,
+                                 test_object_offset,
+                                 DIRECTION_DOWN,
+                                 NULL, NULL,
+                                 &i);
+        if (r <= 0)
                 return r;
 
-        if (ret)
-                *ret = o;
+        /* Calculate new index */
+        if (skip < 0) {
+                if ((uint64_t) -skip >= i)
+                        i = 0;
+                else
+                        i = i - (uint64_t) -skip;
+        } else
+                i  += (uint64_t) skip;
 
-        if (offset)
-                *offset = np;
+        n = le64toh(f->header->n_entries);
+        if (n <= 0)
+                return -EBADMSG;
 
-        return 1;
+        if (i >= n)
+                i = n-1;
+
+        return generic_array_get(f,
+                                 le64toh(f->header->entry_array_offset),
+                                 i,
+                                 ret, offset);
 }
 
-int journal_file_find_first_entry(JournalFile *f, const void *data, uint64_t size, direction_t direction, Object **ret, uint64_t *offset) {
-        uint64_t p, osize, hash, h;
+int journal_file_next_entry_for_data(
+                JournalFile *f,
+                Object *o, uint64_t p,
+                uint64_t data_offset,
+                direction_t direction,
+                Object **ret, uint64_t *offset) {
+
+        uint64_t n, i;
         int r;
+        Object *d;
 
         assert(f);
-        assert(data || size == 0);
+        assert(p > 0 || !o);
 
-        osize = offsetof(Object, data.payload) + size;
+        r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+        if (r <= 0)
+                return r;
 
-        hash = hash64(data, size);
-        h = hash % (le64toh(f->header->hash_table_size) / sizeof(HashItem));
-        p = le64toh(f->hash_table[h].head_hash_offset);
+        n = le64toh(d->data.n_entries);
+        if (n <= 0)
+                return n;
 
-        while (p != 0) {
-                Object *o;
+        if (!o)
+                i = direction == DIRECTION_DOWN ? 0 : n - 1;
+        else {
+                if (o->object.type != OBJECT_ENTRY)
+                        return -EINVAL;
 
-                r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
-                if (r < 0)
+                r = generic_array_bisect_plus_one(f,
+                                                  le64toh(d->data.entry_offset),
+                                                  le64toh(d->data.entry_array_offset),
+                                                  le64toh(d->data.n_entries),
+                                                  p,
+                                                  test_object_offset,
+                                                  DIRECTION_DOWN,
+                                                  NULL, NULL,
+                                                  &i);
+
+                if (r <= 0)
                         return r;
 
-                if (le64toh(o->object.size) == osize &&
-                    memcmp(o->data.payload, data, size) == 0) {
+                if (direction == DIRECTION_DOWN) {
+                        if (i >= n - 1)
+                                return 0;
 
-                        if (le64toh(o->data.hash) != hash)
-                                return -EBADMSG;
+                        i++;
+                } else {
+                        if (i <= 0)
+                                return 0;
 
-                        p = le64toh(direction == DIRECTION_DOWN ?
-                                    o->data.head_entry_offset :
-                                    o->data.tail_entry_offset);
+                        i--;
+                }
 
-                        if (p == 0)
-                                return 0;
+        }
 
-                        r = journal_file_move_to_object(f, p, OBJECT_ENTRY, &o);
-                        if (r < 0)
-                                return r;
+        return generic_array_get_plus_one(f,
+                                          le64toh(d->data.entry_offset),
+                                          le64toh(d->data.entry_array_offset),
+                                          i,
+                                          ret, offset);
+}
 
-                        if (ret)
-                                *ret = o;
+int journal_file_move_to_entry_by_seqnum_for_data(
+                JournalFile *f,
+                uint64_t data_offset,
+                uint64_t seqnum,
+                direction_t direction,
+                Object **ret, uint64_t *offset) {
 
-                        if (offset)
-                                *offset = p;
+        Object *d;
+        int r;
 
-                        return 1;
-                }
+        r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+        if (r <= 0)
+                return r;
 
-                p = le64toh(o->data.next_hash_offset);
-        }
+        return generic_array_bisect_plus_one(f,
+                                             le64toh(d->data.entry_offset),
+                                             le64toh(d->data.entry_array_offset),
+                                             le64toh(d->data.n_entries),
+                                             seqnum,
+                                             test_object_seqnum,
+                                             direction,
+                                             ret, offset, NULL);
+}
 
-        return 0;
+int journal_file_move_to_entry_by_realtime_for_data(
+                JournalFile *f,
+                uint64_t data_offset,
+                uint64_t realtime,
+                direction_t direction,
+                Object **ret, uint64_t *offset) {
+
+        Object *d;
+        int r;
+
+        r = journal_file_move_to_object(f, OBJECT_DATA, data_offset, &d);
+        if (r <= 0)
+                return r;
+
+        return generic_array_bisect_plus_one(f,
+                                             le64toh(d->data.entry_offset),
+                                             le64toh(d->data.entry_array_offset),
+                                             le64toh(d->data.n_entries),
+                                             realtime,
+                                             test_object_realtime,
+                                             direction,
+                                             ret, offset, NULL);
 }
 
 void journal_file_dump(JournalFile *f) {
@@ -1018,18 +1523,24 @@ void journal_file_dump(JournalFile *f) {
 
         assert(f);
 
-        printf("File ID: %s\n"
+        printf("File Path: %s\n"
+               "File ID: %s\n"
                "Machine ID: %s\n"
                "Boot ID: %s\n"
-               "Arena size: %llu\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 long) le64toh(f->header->arena_size),
+               (unsigned long) le64toh(f->header->n_objects),
+               (unsigned long) le64toh(f->header->n_entries));
 
-        p = le64toh(f->header->head_object_offset);
+        p = le64toh(f->header->arena_offset);
         while (p != 0) {
-                r = journal_file_move_to_object(f, p, -1, &o);
+                r = journal_file_move_to_object(f, -1, p, &o);
                 if (r < 0)
                         goto fail;
 
@@ -1050,12 +1561,16 @@ void journal_file_dump(JournalFile *f) {
                                (unsigned long long) le64toh(o->entry.realtime));
                         break;
 
-                case OBJECT_HASH_TABLE:
-                        printf("Type: OBJECT_HASH_TABLE\n");
+                case OBJECT_FIELD_HASH_TABLE:
+                        printf("Type: OBJECT_FIELD_HASH_TABLE\n");
                         break;
 
-                case OBJECT_BISECT_TABLE:
-                        printf("Type: OBJECT_BISECT_TABLE\n");
+                case OBJECT_DATA_HASH_TABLE:
+                        printf("Type: OBJECT_DATA_HASH_TABLE\n");
+                        break;
+
+                case OBJECT_ENTRY_ARRAY:
+                        printf("Type: OBJECT_ENTRY_ARRAY\n");
                         break;
                 }
 
@@ -1153,20 +1668,20 @@ int journal_file_open(
 
         if (newly_created) {
 
-                r = journal_file_setup_hash_table(f);
+                r = journal_file_setup_field_hash_table(f);
                 if (r < 0)
                         goto fail;
 
-                r = journal_file_setup_bisect_table(f);
+                r = journal_file_setup_data_hash_table(f);
                 if (r < 0)
                         goto fail;
         }
 
-        r = journal_file_map_hash_table(f);
+        r = journal_file_map_field_hash_table(f);
         if (r < 0)
                 goto fail;
 
-        r = journal_file_map_bisect_table(f);
+        r = journal_file_map_data_hash_table(f);
         if (r < 0)
                 goto fail;
 
@@ -1381,5 +1896,8 @@ finish:
 
         free(list);
 
+        if (d)
+                closedir(d);
+
         return r;
 }
index 795a446dc8dc636e7408315a07a5c70cbf5cb194..664f917bb0e6dfbcce90f7f961a050e3ddbef0e4 100644 (file)
 #include "util.h"
 #include "sd-id128.h"
 
+typedef struct Window {
+        void *ptr;
+        uint64_t offset;
+        uint64_t size;
+} Window;
+
+enum {
+        WINDOW_UNKNOWN = OBJECT_UNUSED,
+        WINDOW_DATA = OBJECT_DATA,
+        WINDOW_ENTRY = OBJECT_ENTRY,
+        WINDOW_DATA_HASH_TABLE = OBJECT_DATA_HASH_TABLE,
+        WINDOW_FIELD_HASH_TABLE = OBJECT_FIELD_HASH_TABLE,
+        WINDOW_ENTRY_ARRAY = OBJECT_ENTRY_ARRAY,
+        WINDOW_HEADER,
+        _WINDOW_MAX
+};
+
 typedef struct JournalFile {
         int fd;
         char *path;
@@ -36,20 +53,13 @@ typedef struct JournalFile {
         int flags;
         int prot;
         bool writable;
+        bool tail_entry_monotonic_valid;
 
         Header *header;
+        HashItem *data_hash_table;
+        HashItem *field_hash_table;
 
-        HashItem *hash_table;
-        void *hash_table_window;
-        uint64_t hash_table_window_size;
-
-        uint64_t *bisect_table;
-        void *bisect_table_window;
-        uint64_t bisect_table_window_size;
-
-        void *window;
-        uint64_t window_offset;
-        uint64_t window_size;
+        Window windows[_WINDOW_MAX];
 
         uint64_t current_offset;
 } JournalFile;
@@ -60,20 +70,28 @@ typedef enum direction {
 } direction_t;
 
 int journal_file_open(const char *fname, int flags, mode_t mode, JournalFile *template, JournalFile **ret);
-
 void journal_file_close(JournalFile *j);
 
-int journal_file_move_to_object(JournalFile *f, uint64_t offset, int type, Object **ret);
+int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret);
 
 uint64_t journal_file_entry_n_items(Object *o);
 
 int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const struct iovec iovec[], unsigned n_iovec, uint64_t *seqno, Object **ret, uint64_t *offset);
 
-int journal_file_move_to_entry(JournalFile *f, uint64_t seqnum, Object **ret, uint64_t *offset);
+int journal_file_find_data_object(JournalFile *f, const void *data, uint64_t size, Object **ret, uint64_t *offset);
+int journal_file_find_data_object_with_hash(JournalFile *f, const void *data, uint64_t size, uint64_t hash, Object **ret, uint64_t *offset);
 
-int journal_file_find_first_entry(JournalFile *f, const void *data, uint64_t size, direction_t direction, Object **ret, uint64_t *offset);
+int journal_file_next_entry(JournalFile *f, Object *o, uint64_t p, direction_t direction, Object **ret, uint64_t *offset);
+int journal_file_skip_entry(JournalFile *f, Object *o, uint64_t p, int64_t skip, Object **ret, uint64_t *offset);
 
-int journal_file_next_entry(JournalFile *f, Object *o, direction_t direction, Object **ret, uint64_t *offset);
+int journal_file_next_entry_for_data(JournalFile *f, Object *o, uint64_t p, uint64_t data_offset, direction_t direction, Object **ret, uint64_t *offset);
+
+int journal_file_move_to_entry_by_seqnum(JournalFile *f, uint64_t seqnum, direction_t direction, Object **ret, uint64_t *offset);
+int journal_file_move_to_entry_by_realtime(JournalFile *f, uint64_t realtime, direction_t direction, Object **ret, uint64_t *offset);
+int journal_file_move_to_entry_by_monotonic(JournalFile *f, sd_id128_t boot_id, uint64_t monotonic, direction_t direction, Object **ret, uint64_t *offset);
+
+int journal_file_move_to_entry_by_seqnum_for_data(JournalFile *f, uint64_t data_offset, uint64_t seqnum, direction_t direction, Object **ret, uint64_t *offset);
+int journal_file_move_to_entry_by_realtime_for_data(JournalFile *f, uint64_t data_offset, uint64_t realtime, direction_t direction, Object **ret, uint64_t *offset);
 
 void journal_file_dump(JournalFile *f);
 
@@ -81,5 +99,4 @@ int journal_file_rotate(JournalFile **f);
 
 int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t min_free);
 
-
 #endif
index 3e5ee1af967dd1244ef4747bb9bd0b253cf2b77a..9220efdfec45f95df059a5c8bd55104a1cde5d79 100644 (file)
@@ -54,7 +54,7 @@ int main(int argc, char *argv[]) {
                 }
         }
 
-        SD_JOURNAL_FOREACH_BEGIN(j) {
+        SD_JOURNAL_FOREACH(j) {
 
                 const void *data;
                 size_t length;
@@ -71,7 +71,7 @@ int main(int argc, char *argv[]) {
                 free(cursor);
 
                 sd_journal_get_realtime_usec(j, &realtime);
-                sd_journal_get_monotonic_usec(j, &monotonic);
+                sd_journal_get_monotonic_usec(j, &monotonic, NULL);
                 printf("realtime: %llu\n"
                        "monotonic: %llu\n",
                        (unsigned long long) realtime,
@@ -80,7 +80,7 @@ int main(int argc, char *argv[]) {
                 SD_JOURNAL_FOREACH_DATA(j, data, length)
                         printf("\t%.*s\n", (int) length, (const char*) data);
 
-        } SD_JOURNAL_FOREACH_END(j);
+        }
 
 finish:
         if (j)
index ede314a55f204d7c2fead9bcbbe398522e61be81..89d8bee2a2120a48c1472e3a245df589f111cbd5 100644 (file)
@@ -144,14 +144,14 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
                 *audit_session = NULL, *audit_loginuid = NULL,
                 *syslog_priority = NULL, *syslog_facility = NULL,
                 *exe = NULL, *cgroup = NULL;
-        struct iovec iovec[16];
+        struct iovec iovec[17];
         unsigned n = 0;
         char idbuf[33];
         sd_id128_t id;
         int r;
         char *t;
         int priority = LOG_USER | LOG_INFO;
-        uid_t loginuid = 0;
+        uid_t loginuid = 0, realuid = 0;
         JournalFile *f;
 
         parse_syslog_priority((char**) &buf, &priority);
@@ -171,18 +171,20 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
                 uint32_t session;
                 char *path;
 
-                if (asprintf(&pid, "PID=%lu", (unsigned long) ucred->pid) >= 0)
+                realuid = ucred->uid;
+
+                if (asprintf(&pid, "_PID=%lu", (unsigned long) ucred->pid) >= 0)
                         IOVEC_SET_STRING(iovec[n++], pid);
 
-                if (asprintf(&uid, "UID=%lu", (unsigned long) ucred->uid) >= 0)
+                if (asprintf(&uid, "_UID=%lu", (unsigned long) ucred->uid) >= 0)
                         IOVEC_SET_STRING(iovec[n++], uid);
 
-                if (asprintf(&gid, "GID=%lu", (unsigned long) ucred->gid) >= 0)
+                if (asprintf(&gid, "_GID=%lu", (unsigned long) ucred->gid) >= 0)
                         IOVEC_SET_STRING(iovec[n++], gid);
 
                 r = get_process_comm(ucred->pid, &t);
                 if (r >= 0) {
-                        comm = strappend("COMM=", t);
+                        comm = strappend("_COMM=", t);
                         if (comm)
                                 IOVEC_SET_STRING(iovec[n++], comm);
                         free(t);
@@ -190,7 +192,7 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
 
                 r = get_process_exe(ucred->pid, &t);
                 if (r >= 0) {
-                        exe = strappend("EXE=", t);
+                        exe = strappend("_EXE=", t);
                         if (comm)
                                 IOVEC_SET_STRING(iovec[n++], exe);
                         free(t);
@@ -198,7 +200,7 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
 
                 r = get_process_cmdline(ucred->pid, LINE_MAX, false, &t);
                 if (r >= 0) {
-                        cmdline = strappend("CMDLINE=", t);
+                        cmdline = strappend("_CMDLINE=", t);
                         if (cmdline)
                                 IOVEC_SET_STRING(iovec[n++], cmdline);
                         free(t);
@@ -206,17 +208,17 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
 
                 r = audit_session_from_pid(ucred->pid, &session);
                 if (r >= 0)
-                        if (asprintf(&audit_session, "AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
+                        if (asprintf(&audit_session, "_AUDIT_SESSION=%lu", (unsigned long) session) >= 0)
                                 IOVEC_SET_STRING(iovec[n++], audit_session);
 
                 r = audit_loginuid_from_pid(ucred->pid, &loginuid);
                 if (r >= 0)
-                        if (asprintf(&audit_loginuid, "AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
+                        if (asprintf(&audit_loginuid, "_AUDIT_LOGINUID=%lu", (unsigned long) loginuid) >= 0)
                                 IOVEC_SET_STRING(iovec[n++], audit_loginuid);
 
                 r = cg_get_by_pid(SYSTEMD_CGROUP_CONTROLLER, ucred->pid, &path);
                 if (r >= 0) {
-                        cgroup = strappend("SYSTEMD_CGROUP=", path);
+                        cgroup = strappend("_SYSTEMD_CGROUP=", path);
                         if (cgroup)
                                 IOVEC_SET_STRING(iovec[n++], cgroup);
                         free(path);
@@ -224,7 +226,7 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
         }
 
         if (tv) {
-                if (asprintf(&source_time, "SOURCE_REALTIME_TIMESTAMP=%llu",
+                if (asprintf(&source_time, "_SOURCE_REALTIME_TIMESTAMP=%llu",
                              (unsigned long long) timeval_load(tv)) >= 0)
                         IOVEC_SET_STRING(iovec[n++], source_time);
         }
@@ -234,23 +236,23 @@ static void process_message(Server *s, const char *buf, struct ucred *ucred, str
          * anyway. However, we need this indexed, too. */
         r = sd_id128_get_boot(&id);
         if (r >= 0)
-                if (asprintf(&boot_id, "BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
+                if (asprintf(&boot_id, "_BOOT_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
                         IOVEC_SET_STRING(iovec[n++], boot_id);
 
         r = sd_id128_get_machine(&id);
         if (r >= 0)
-                if (asprintf(&machine_id, "MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
+                if (asprintf(&machine_id, "_MACHINE_ID=%s", sd_id128_to_string(id, idbuf)) >= 0)
                         IOVEC_SET_STRING(iovec[n++], machine_id);
 
         t = gethostname_malloc();
         if (t) {
-                hostname = strappend("HOSTNAME=", t);
+                hostname = strappend("_HOSTNAME=", t);
                 if (hostname)
                         IOVEC_SET_STRING(iovec[n++], hostname);
                 free(t);
         }
 
-        f = find_journal(s, loginuid);
+        f = find_journal(s, realuid == 0 ? 0 : loginuid);
         if (!f)
                 log_warning("Dropping message, as we can't find a place to store the data.");
         else {
index 5d518a38709d07c2d1b9e7903f940a13838c90fb..85c57170d56be46d2e3c55b49ab32f303a8593e2 100644 (file)
@@ -40,9 +40,34 @@ struct Match {
         LIST_FIELDS(Match, matches);
 };
 
+typedef enum location_type {
+        LOCATION_HEAD,
+        LOCATION_TAIL,
+        LOCATION_DISCRETE
+} location_type_t;
+
+typedef struct Location {
+        location_type_t type;
+
+        uint64_t seqnum;
+        sd_id128_t seqnum_id;
+        bool seqnum_set;
+
+        uint64_t realtime;
+        bool realtime_set;
+
+        uint64_t monotonic;
+        sd_id128_t boot_id;
+        bool monotonic_set;
+
+        uint64_t xor_hash;
+        bool xor_hash_set;
+} Location;
+
 struct sd_journal {
         Hashmap *files;
 
+        Location current_location;
         JournalFile *current_file;
         uint64_t current_field;
 
@@ -50,7 +75,7 @@ struct sd_journal {
         unsigned n_matches;
 };
 
-static void reset_location(sd_journal *j) {
+static void detach_location(sd_journal *j) {
         Iterator i;
         JournalFile *f;
 
@@ -63,8 +88,66 @@ static void reset_location(sd_journal *j) {
                 f->current_offset = 0;
 }
 
+static void reset_location(sd_journal *j) {
+        assert(j);
+
+        detach_location(j);
+        zero(j->current_location);
+}
+
+static void init_location(Location *l, JournalFile *f, Object *o) {
+        assert(l);
+        assert(f);
+        assert(o->object.type == OBJECT_ENTRY);
+
+        l->type = LOCATION_DISCRETE;
+        l->seqnum = le64toh(o->entry.seqnum);
+        l->seqnum_id = f->header->seqnum_id;
+        l->realtime = le64toh(o->entry.realtime);
+        l->monotonic = le64toh(o->entry.monotonic);
+        l->boot_id = le64toh(o->entry.boot_id);
+        l->xor_hash = le64toh(o->entry.xor_hash);
+
+        l->seqnum_set = l->realtime_set = l->monotonic_set = l->xor_hash_set = true;
+}
+
+static void set_location(sd_journal *j, JournalFile *f, Object *o, uint64_t offset) {
+        assert(j);
+        assert(f);
+        assert(o);
+
+        init_location(&j->current_location, f, o);
+
+        j->current_file = f;
+        j->current_field = 0;
+
+        f->current_offset = offset;
+}
+
+static int same_field(const void *_a, size_t s, const void *_b, size_t t) {
+        const uint8_t *a = _a, *b = _b;
+        size_t j;
+        bool a_good = false, b_good = false, different = false;
+
+        for (j = 0; j < s && j < t; j++) {
+
+                if (a[j] == '=')
+                        a_good = true;
+                if (b[j] == '=')
+                        b_good = true;
+                if (a[j] != b[j])
+                        different = true;
+
+                if (a_good && b_good)
+                        return different ? 0 : 1;
+        }
+
+        return -EINVAL;
+}
+
 int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
-        Match *m;
+        Match *m, *after = NULL;
+        uint64_t le_hash;
 
         assert(j);
 
@@ -73,6 +156,23 @@ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
 
         assert(data);
 
+        le_hash = htole64(hash64(data, size));
+
+        LIST_FOREACH(matches, m, j->matches) {
+                int r;
+
+                if (m->le_hash == le_hash &&
+                    m->size == size &&
+                    memcmp(m->data, data, size) == 0)
+                        return 0;
+
+                r = same_field(data, size, m->data, m->size);
+                if (r < 0)
+                        return r;
+                else if (r > 0)
+                        after = m;
+        }
+
         m = new0(Match, 1);
         if (!m)
                 return -ENOMEM;
@@ -86,12 +186,14 @@ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) {
         }
 
         memcpy(m->data, data, size);
-        m->le_hash = hash64(m->data, size);
+        m->le_hash = le_hash;
 
-        LIST_PREPEND(Match, matches, j->matches, m);
+        /* Matches for the same fields we order adjacent to each
+         * other */
+        LIST_INSERT_AFTER(Match, matches, j->matches, after, m);
         j->n_matches ++;
 
-        reset_location(j);
+        detach_location(j);
 
         return 0;
 }
@@ -109,14 +211,19 @@ void sd_journal_flush_matches(sd_journal *j) {
 
         j->n_matches = 0;
 
-        reset_location(j);
+        detach_location(j);
 }
 
-static int compare_order(JournalFile *af, Object *ao, uint64_t ap,
-                         JournalFile *bf, Object *bo, uint64_t bp) {
+static int compare_order(JournalFile *af, Object *ao,
+                         JournalFile *bf, Object *bo) {
 
         uint64_t a, b;
 
+        assert(af);
+        assert(ao);
+        assert(bf);
+        assert(bo);
+
         /* We operate on two different files here, hence we can access
          * two objects at the same time, which we normally can't.
          *
@@ -179,99 +286,302 @@ static int compare_order(JournalFile *af, Object *ao, uint64_t ap,
         return 0;
 }
 
-static int move_to_next_with_matches(sd_journal *j, JournalFile *f, direction_t direction, Object **o, uint64_t *p) {
+static int compare_with_location(JournalFile *af, Object *ao, Location *l) {
+        uint64_t a;
+
+        assert(af);
+        assert(ao);
+        assert(l);
+        assert(l->type == LOCATION_DISCRETE);
+
+        if (l->monotonic_set &&
+            sd_id128_equal(ao->entry.boot_id, l->boot_id) &&
+            l->realtime_set &&
+            le64toh(ao->entry.realtime) == l->realtime &&
+            l->xor_hash_set &&
+            le64toh(ao->entry.xor_hash) == l->xor_hash)
+                return 0;
+
+        if (l->seqnum_set &&
+            sd_id128_equal(af->header->seqnum_id, l->seqnum_id)) {
+
+                a = le64toh(ao->entry.seqnum);
+
+                if (a < l->seqnum)
+                        return -1;
+                if (a > l->seqnum)
+                        return 1;
+        }
+
+        if (l->monotonic_set &&
+            sd_id128_equal(ao->entry.boot_id, l->boot_id)) {
+
+                a = le64toh(ao->entry.monotonic);
+
+                if (a < l->monotonic)
+                        return -1;
+                if (a > l->monotonic)
+                        return 1;
+        }
+
+        if (l->realtime_set) {
+
+                a = le64toh(ao->entry.realtime);
+
+                if (a < l->realtime)
+                        return -1;
+                if (a > l->realtime)
+                        return 1;
+        }
+
+        if (l->xor_hash_set) {
+                a = le64toh(ao->entry.xor_hash);
+
+                if (a < l->xor_hash)
+                        return -1;
+                if (a > l->xor_hash)
+                        return 1;
+        }
+
+        return 0;
+}
+
+static int find_location(sd_journal *j, JournalFile *f, direction_t direction, Object **ret, uint64_t *offset) {
+        Object *o = NULL;
+        uint64_t p = 0;
         int r;
-        uint64_t cp;
-        Object *c;
 
         assert(j);
-        assert(f);
-        assert(o);
-        assert(p);
 
         if (!j->matches) {
-                /* No matches is easy, just go on to the next entry */
+                /* No matches is simple */
+
+                if (j->current_location.type == LOCATION_HEAD)
+                        r = journal_file_next_entry(f, NULL, 0, DIRECTION_DOWN, &o, &p);
+                else if (j->current_location.type == LOCATION_TAIL)
+                        r = journal_file_next_entry(f, NULL, 0, DIRECTION_UP, &o, &p);
+                else if (j->current_location.seqnum_set &&
+                         sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
+                        r = journal_file_move_to_entry_by_seqnum(f, j->current_location.seqnum, direction, &o, &p);
+                else if (j->current_location.monotonic_set)
+                        r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, &o, &p);
+                else if (j->current_location.realtime_set)
+                        r = journal_file_move_to_entry_by_realtime(f, j->current_location.realtime, direction, &o, &p);
+                else
+                        r = journal_file_next_entry(f, NULL, 0, direction, &o, &p);
+
+                if (r <= 0)
+                        return r;
+
+        } else  {
+                Match *m, *term_match = NULL;
+                Object *to = NULL;
+                uint64_t tp = 0;
+
+                /* We have matches, first, let's jump to the monotonic
+                 * position if we have any, since it implies a
+                 * match. */
 
-                if (f->current_offset > 0) {
-                        r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &c);
-                        if (r < 0)
+                if (j->current_location.type == LOCATION_DISCRETE &&
+                    j->current_location.monotonic_set) {
+
+                        r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, &o, &p);
+                        if (r <= 0)
                                 return r;
-                } else
-                        c = NULL;
+                }
+
+                LIST_FOREACH(matches, m, j->matches) {
+                        Object *c, *d;
+                        uint64_t cp, dp;
+
+                        r = journal_file_find_data_object_with_hash(f, m->data, m->size, m->le_hash, &d, &dp);
+                        if (r <= 0)
+                                return r;
+
+                        if (j->current_location.type == LOCATION_HEAD)
+                                r = journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_DOWN, &c, &cp);
+                        else if (j->current_location.type == LOCATION_TAIL)
+                                r = journal_file_next_entry_for_data(f, NULL, 0, dp, DIRECTION_UP, &c, &cp);
+                        else if (j->current_location.seqnum_set &&
+                                 sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id))
+                                r = journal_file_move_to_entry_by_seqnum_for_data(f, dp, j->current_location.seqnum, direction, &c, &cp);
+                        else if (j->current_location.realtime_set)
+                                r = journal_file_move_to_entry_by_realtime_for_data(f, dp, j->current_location.realtime, direction, &c, &cp);
+                        else
+                                r = journal_file_next_entry_for_data(f, NULL, 0, dp, direction, &c, &cp);
+
+                        if (!term_match) {
+                                term_match = m;
+
+                                if (r > 0) {
+                                        to = c;
+                                        tp = cp;
+                                }
+                        } else if (same_field(term_match->data, term_match->size, m->data, m->size)) {
+
+                                /* Same field as previous match... */
+                                if (r > 0) {
+
+                                        /* Find the earliest of the OR matches */
+
+                                        if (!to ||
+                                            (direction == DIRECTION_DOWN && cp < tp) ||
+                                            (direction == DIRECTION_UP && cp > tp)) {
+                                                to = c;
+                                                tp = tp;
+                                        }
+
+                                }
 
-                return journal_file_next_entry(f, c, direction, o, p);
+                        } else {
+
+                                /* Previous term is finished, did anything match? */
+                                if (!to)
+                                        return 0;
+
+                                /* Find the last of the AND matches */
+                                if (!o ||
+                                    (direction == DIRECTION_DOWN && tp > p) ||
+                                    (direction == DIRECTION_UP && tp < p)) {
+                                        o = to;
+                                        p = tp;
+                                }
+
+                                term_match = m;
+
+                                if (r > 0) {
+                                        to = c;
+                                        tp = cp;
+                                } else {
+                                        to = NULL;
+                                        tp = 0;
+                                }
+                        }
+                }
+
+                /* Last term is finished, did anything match? */
+                if (!to)
+                        return 0;
+
+                if (!o ||
+                    (direction == DIRECTION_DOWN && tp > p) ||
+                    (direction == DIRECTION_UP && tp < p)) {
+                        o = to;
+                        p = tp;
+                }
+
+                if (!o)
+                        return 0;
         }
 
-        /* So there are matches we have to adhere to, let's find the
-         * first entry that matches all of them */
+        if (ret)
+                *ret = o;
 
-        if (f->current_offset > 0)
-                cp = f->current_offset;
-        else {
-                r = journal_file_find_first_entry(f, j->matches->data, j->matches->size, direction, &c, &cp);
+        if (offset)
+                *offset = p;
+
+        return 1;
+}
+
+static int next_with_matches(sd_journal *j, JournalFile *f, direction_t direction, Object **ret, uint64_t *offset) {
+        int r;
+        uint64_t cp;
+        Object *c;
+
+        assert(j);
+        assert(f);
+        assert(ret);
+        assert(offset);
+
+        c = *ret;
+        cp = *offset;
+
+        if (!j->matches) {
+                /* No matches is easy */
+
+                r = journal_file_next_entry(f, c, cp, direction, &c, &cp);
                 if (r <= 0)
                         return r;
 
-                /* We can shortcut this if there's only one match */
-                if (j->n_matches == 1) {
-                        *o = c;
-                        *p = cp;
-                        return r;
-                }
+                if (ret)
+                        *ret = c;
+                if (offset)
+                        *offset = cp;
+                return 1;
         }
 
+        /* So there are matches we have to adhere to, let's find the
+         * first entry that matches all of them */
+
         for (;;) {
                 uint64_t np, n;
-                bool found;
-                Match *m;
-
-                r = journal_file_move_to_object(f, cp, OBJECT_ENTRY, &c);
-                if (r < 0)
-                        return r;
+                bool found, term_result = false;
+                Match *m, *term_match = NULL;
 
                 n = journal_file_entry_n_items(c);
 
                 /* Make sure we don't match the entry we are starting
                  * from. */
-                found = f->current_offset != cp;
+                found = cp > *offset;
 
                 np = 0;
                 LIST_FOREACH(matches, m, j->matches) {
                         uint64_t q, k;
 
+                        /* Let's check if this is the beginning of a
+                         * new term, i.e. has a different field prefix
+                         * as the preceeding match. */
+                        if (!term_match) {
+                                term_match = m;
+                                term_result = false;
+                        } else if (!same_field(term_match->data, term_match->size, m->data, m->size)) {
+                                if (!term_result)
+                                        found = false;
+
+                                term_match = m;
+                                term_result = false;
+                        }
+
                         for (k = 0; k < n; k++)
                                 if (c->entry.items[k].hash == m->le_hash)
                                         break;
 
                         if (k >= n) {
-                                /* Hmm, didn't find any field that matched, so ignore
-                                 * this match. Go on with next match */
-
-                                found = false;
+                                /* Hmm, didn't find any field that
+                                 * matched this rule, so ignore this
+                                 * match. Go on with next match */
                                 continue;
                         }
 
+                        term_result = true;
+
                         /* Hmm, so, this field matched, let's remember
                          * where we'd have to try next, in case the other
                          * matches are not OK */
 
-                        if (direction == DIRECTION_DOWN) {
-                                q = le64toh(c->entry.items[k].next_entry_offset);
-
-                                if (q > np)
-                                        np = q;
-                        } else {
-                                q = le64toh(c->entry.items[k].prev_entry_offset);
+                        r = journal_file_next_entry_for_data(f, c, cp, le64toh(c->entry.items[k].object_offset), direction, NULL, &q);
+                        if (r > 0) {
 
-                                if (q != 0 && (np == 0 || q < np))
-                                        np = q;
+                                if (direction == DIRECTION_DOWN) {
+                                        if (q > np)
+                                                np = q;
+                                } else {
+                                        if (np == 0 || q < np)
+                                                np = q;
+                                }
                         }
                 }
 
+                /* Check the last term */
+                if (term_match && term_result)
+                        found = true;
+
                 /* Did this entry match against all matches? */
                 if (found) {
-                        *o = c;
-                        *p = cp;
+                        if (ret)
+                                *ret = c;
+                        if (offset)
+                                *offset = cp;
                         return 1;
                 }
 
@@ -285,6 +595,62 @@ static int move_to_next_with_matches(sd_journal *j, JournalFile *f, direction_t
         }
 }
 
+static int next_beyond_location(sd_journal *j, JournalFile *f, direction_t direction, Object **ret, uint64_t *offset) {
+        Object *c;
+        uint64_t cp;
+        int compare_value, r;
+
+        assert(j);
+        assert(f);
+
+        if (f->current_offset > 0) {
+                r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &c);
+                if (r < 0)
+                        return r;
+
+                cp = f->current_offset;
+
+                r = next_with_matches(j, f, direction, &c, &cp);
+                if (r <= 0)
+                        return r;
+
+                compare_value = 1;
+        } else {
+                r = find_location(j, f, direction, &c, &cp);
+                if (r <= 0)
+                        return r;
+
+                compare_value = 0;
+        }
+
+        for (;;) {
+                bool found;
+
+                if (j->current_location.type == LOCATION_DISCRETE) {
+                        int k;
+
+                        k = compare_with_location(f, c, &j->current_location);
+                        if (direction == DIRECTION_DOWN)
+                                found = k >= compare_value;
+                        else
+                                found = k <= -compare_value;
+                } else
+                        found = true;
+
+                if (found) {
+                        if (ret)
+                                *ret = c;
+                        if (offset)
+                                *offset = cp;
+                        return 1;
+                }
+
+                r = next_with_matches(j, f, direction, &c, &cp);
+                if (r <= 0)
+                        return r;
+        }
+}
+
 static int real_journal_next(sd_journal *j, direction_t direction) {
         JournalFile *f, *new_current = NULL;
         Iterator i;
@@ -297,57 +663,88 @@ static int real_journal_next(sd_journal *j, direction_t direction) {
         HASHMAP_FOREACH(f, j->files, i) {
                 Object *o;
                 uint64_t p;
+                bool found;
 
-                r = move_to_next_with_matches(j, f, direction, &o, &p);
+                r = next_beyond_location(j, f, direction, &o, &p);
                 if (r < 0)
                         return r;
                 else if (r == 0)
                         continue;
 
-                if (!new_current ||
-                    compare_order(new_current, new_entry, new_offset, f, o, p) > 0) {
+                if (!new_current)
+                        found = true;
+                else {
+                        int k;
+
+                        k = compare_order(f, o, new_current, new_entry);
+
+                        if (direction == DIRECTION_DOWN)
+                                found = k < 0;
+                        else
+                                found = k > 0;
+                }
+
+                if (found) {
                         new_current = f;
                         new_entry = o;
                         new_offset = p;
                 }
         }
 
-        if (new_current) {
-                j->current_file = new_current;
-                j->current_file->current_offset = new_offset;
-                j->current_field = 0;
+        if (!new_current)
+                return 0;
 
-                /* Skip over any identical entries in the other files too */
+        set_location(j, new_current, new_entry, new_offset);
 
-                HASHMAP_FOREACH(f, j->files, i) {
-                        Object *o;
-                        uint64_t p;
+        return 1;
+}
 
-                        if (j->current_file == f)
-                                continue;
+int sd_journal_next(sd_journal *j) {
+        return real_journal_next(j, DIRECTION_DOWN);
+}
 
-                        r = move_to_next_with_matches(j, f, direction, &o, &p);
-                        if (r < 0)
-                                return r;
-                        else if (r == 0)
-                                continue;
+int sd_journal_previous(sd_journal *j) {
+        return real_journal_next(j, DIRECTION_UP);
+}
 
-                        if (compare_order(new_current, new_entry, new_offset, f, o, p) == 0)
-                                f->current_offset = p;
-                }
+int sd_journal_next_skip(sd_journal *j, uint64_t skip) {
+        int c = 0, r;
 
-                return 1;
+        assert(j);
+
+        while (skip > 0) {
+                r = sd_journal_next(j);
+                if (r < 0)
+                        return r;
+
+                if (r == 0)
+                        return c;
+
+                skip--;
+                c++;
         }
 
-        return 0;
+        return c;
 }
 
-int sd_journal_next(sd_journal *j) {
-        return real_journal_next(j, DIRECTION_DOWN);
-}
+int sd_journal_previous_skip(sd_journal *j, uint64_t skip) {
+        int c = 0, r;
 
-int sd_journal_previous(sd_journal *j) {
-        return real_journal_next(j, DIRECTION_UP);
+        assert(j);
+
+        while (skip > 0) {
+                r = sd_journal_previous(j);
+                if (r < 0)
+                        return r;
+
+                if (r == 0)
+                        return c;
+
+                skip--;
+                c++;
+        }
+
+        return 1;
 }
 
 int sd_journal_get_cursor(sd_journal *j, char **cursor) {
@@ -361,7 +758,7 @@ int sd_journal_get_cursor(sd_journal *j, char **cursor) {
         if (!j->current_file || j->current_file->current_offset <= 0)
                 return -EADDRNOTAVAIL;
 
-        r = journal_file_move_to_object(j->current_file, j->current_file->current_offset, OBJECT_ENTRY, &o);
+        r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o);
         if (r < 0)
                 return r;
 
@@ -380,8 +777,150 @@ int sd_journal_get_cursor(sd_journal *j, char **cursor) {
         return 1;
 }
 
-int sd_journal_set_cursor(sd_journal *j, const char *cursor) {
-        return -EINVAL;
+int sd_journal_seek_cursor(sd_journal *j, const char *cursor) {
+        char *w;
+        size_t l;
+        char *state;
+        unsigned long long seqnum, monotonic, realtime, xor_hash;
+        bool
+                seqnum_id_set = false,
+                seqnum_set = false,
+                boot_id_set = false,
+                monotonic_set = false,
+                realtime_set = false,
+                xor_hash_set = false;
+        sd_id128_t seqnum_id, boot_id;
+
+        assert(j);
+        assert(cursor);
+
+        FOREACH_WORD_SEPARATOR(w, l, cursor, ";", state) {
+                char *item;
+                int k = 0;
+
+                if (l < 2 || w[1] != '=')
+                        return -EINVAL;
+
+                item = strndup(w, l);
+                if (!item)
+                        return -ENOMEM;
+
+                switch (w[0]) {
+
+                case 's':
+                        seqnum_id_set = true;
+                        k = sd_id128_from_string(w+2, &seqnum_id);
+                        break;
+
+                case 'i':
+                        seqnum_set = true;
+                        if (sscanf(w+2, "%llx", &seqnum) != 1)
+                                k = -EINVAL;
+                        break;
+
+                case 'b':
+                        boot_id_set = true;
+                        k = sd_id128_from_string(w+2, &boot_id);
+                        break;
+
+                case 'm':
+                        monotonic_set = true;
+                        if (sscanf(w+2, "%llx", &monotonic) != 1)
+                                k = -EINVAL;
+                        break;
+
+                case 't':
+                        realtime_set = true;
+                        if (sscanf(w+2, "%llx", &realtime) != 1)
+                                k = -EINVAL;
+                        break;
+
+                case 'x':
+                        xor_hash_set = true;
+                        if (sscanf(w+2, "%llx", &xor_hash) != 1)
+                                k = -EINVAL;
+                        break;
+                }
+
+                free(item);
+
+                if (k < 0)
+                        return k;
+        }
+
+        if ((!seqnum_set || !seqnum_id_set) &&
+            (!monotonic_set || !boot_id_set) &&
+            !realtime_set)
+                return -EINVAL;
+
+        reset_location(j);
+
+        j->current_location.type = LOCATION_DISCRETE;
+
+        if (realtime_set) {
+                j->current_location.realtime = (uint64_t) realtime;
+                j->current_location.realtime_set = true;
+        }
+
+        if (seqnum_set && seqnum_id_set) {
+                j->current_location.seqnum = (uint64_t) seqnum;
+                j->current_location.seqnum_id = seqnum_id;
+                j->current_location.seqnum_set = true;
+        }
+
+        if (monotonic_set && boot_id_set) {
+                j->current_location.monotonic = (uint64_t) monotonic;
+                j->current_location.boot_id = boot_id;
+                j->current_location.monotonic_set = true;
+        }
+
+        if (xor_hash_set) {
+                j->current_location.xor_hash = (uint64_t) xor_hash;
+                j->current_location.xor_hash_set = true;
+        }
+
+        return 0;
+}
+
+int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec) {
+        assert(j);
+
+        reset_location(j);
+        j->current_location.type = LOCATION_DISCRETE;
+        j->current_location.boot_id = boot_id;
+        j->current_location.monotonic = usec;
+        j->current_location.monotonic_set = true;
+
+        return 0;
+}
+
+int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec) {
+        assert(j);
+
+        reset_location(j);
+        j->current_location.type = LOCATION_DISCRETE;
+        j->current_location.realtime = usec;
+        j->current_location.realtime_set = true;
+
+        return 0;
+}
+
+int sd_journal_seek_head(sd_journal *j) {
+        assert(j);
+
+        reset_location(j);
+        j->current_location.type = LOCATION_HEAD;
+
+        return 0;
+}
+
+int sd_journal_seek_tail(sd_journal *j) {
+        assert(j);
+
+        reset_location(j);
+        j->current_location.type = LOCATION_TAIL;
+
+        return 0;
 }
 
 static int add_file(sd_journal *j, const char *prefix, const char *dir, const char *filename) {
@@ -411,6 +950,9 @@ static int add_file(sd_journal *j, const char *prefix, const char *dir, const ch
                 return r;
         }
 
+        journal_file_dump(f);
+
+
         r = hashmap_put(j->files, f->path, f);
         if (r < 0) {
                 journal_file_close(f);
@@ -557,20 +1099,20 @@ int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret) {
 
         f = j->current_file;
         if (!f)
-                return 0;
+                return -EADDRNOTAVAIL;
 
         if (f->current_offset <= 0)
-                return 0;
+                return -EADDRNOTAVAIL;
 
-        r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
         if (r < 0)
                 return r;
 
         *ret = le64toh(o->entry.realtime);
-        return 1;
+        return 0;
 }
 
-int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret) {
+int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id) {
         Object *o;
         JournalFile *f;
         int r;
@@ -581,25 +1123,28 @@ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret) {
 
         f = j->current_file;
         if (!f)
-                return 0;
+                return -EADDRNOTAVAIL;
 
         if (f->current_offset <= 0)
-                return 0;
+                return -EADDRNOTAVAIL;
 
-        r = sd_id128_get_boot(&id);
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
         if (r < 0)
                 return r;
 
-        r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
-        if (r < 0)
-                return r;
+        if (ret_boot_id)
+                *ret_boot_id = o->entry.boot_id;
+        else {
+                r = sd_id128_get_boot(&id);
+                if (r < 0)
+                        return r;
 
-        if (!sd_id128_equal(id, o->entry.boot_id))
-                return 0;
+                if (!sd_id128_equal(id, o->entry.boot_id))
+                        return -ENOENT;
+        }
 
         *ret = le64toh(o->entry.monotonic);
-        return 1;
-
+        return 0;
 }
 
 int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *size) {
@@ -619,12 +1164,12 @@ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, siz
 
         f = j->current_file;
         if (!f)
-                return 0;
+                return -EADDRNOTAVAIL;
 
         if (f->current_offset <= 0)
-                return 0;
+                return -EADDRNOTAVAIL;
 
-        r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
         if (r < 0)
                 return r;
 
@@ -632,16 +1177,16 @@ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, siz
 
         n = journal_file_entry_n_items(o);
         for (i = 0; i < n; i++) {
-                uint64_t p, l, h;
+                uint64_t p, l, le_hash;
                 size_t t;
 
                 p = le64toh(o->entry.items[i].object_offset);
-                h = o->entry.items[j->current_field].hash;
-                r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+                le_hash = o->entry.items[j->current_field].hash;
+                r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
                 if (r < 0)
                         return r;
 
-                if (h != o->data.hash)
+                if (le_hash != o->data.hash)
                         return -EBADMSG;
 
                 l = le64toh(o->object.size) - offsetof(Object, data.payload);
@@ -658,23 +1203,23 @@ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, siz
                         *data = o->data.payload;
                         *size = t;
 
-                        return 1;
+                        return 0;
                 }
 
-                r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
+                r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
                 if (r < 0)
                         return r;
         }
 
-        return 0;
+        return -ENOENT;
 }
 
 int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
         JournalFile *f;
-        uint64_t p, l, n, h;
-        size_t t;
+        uint64_t p, l, n, le_hash;
         int r;
         Object *o;
+        size_t t;
 
         assert(j);
         assert(data);
@@ -682,12 +1227,12 @@ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
 
         f = j->current_file;
         if (!f)
-                return 0;
+                return -EADDRNOTAVAIL;
 
         if (f->current_offset <= 0)
-                return 0;
+                return -EADDRNOTAVAIL;
 
-        r = journal_file_move_to_object(f, f->current_offset, OBJECT_ENTRY, &o);
+        r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o);
         if (r < 0)
                 return r;
 
@@ -696,12 +1241,12 @@ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
                 return 0;
 
         p = le64toh(o->entry.items[j->current_field].object_offset);
-        h = o->entry.items[j->current_field].hash;
-        r = journal_file_move_to_object(f, p, OBJECT_DATA, &o);
+        le_hash = o->entry.items[j->current_field].hash;
+        r = journal_file_move_to_object(f, OBJECT_DATA, p, &o);
         if (r < 0)
                 return r;
 
-        if (h != o->data.hash)
+        if (le_hash != o->data.hash)
                 return -EBADMSG;
 
         l = le64toh(o->object.size) - offsetof(Object, data.payload);
@@ -719,24 +1264,8 @@ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) {
         return 1;
 }
 
-void sd_journal_start_data(sd_journal *j) {
+void sd_journal_restart_data(sd_journal *j) {
         assert(j);
 
         j->current_field = 0;
 }
-
-int sd_journal_seek_head(sd_journal *j) {
-        assert(j);
-
-        reset_location(j);
-
-        return real_journal_next(j, DIRECTION_DOWN);
-}
-
-int sd_journal_seek_tail(sd_journal *j) {
-        assert(j);
-
-        reset_location(j);
-
-        return real_journal_next(j, DIRECTION_UP);
-}
index 9978ca9ac0dd2f500fbae691c6e87b843709d72b..8d7e314223c1ce008b2fc23224fc1a16c95c79e9 100644 (file)
 #include <inttypes.h>
 #include <sys/types.h>
 
+#include "sd-id128.h"
+
 /* TODO:
  *
  *   - check LE/BE conversion for 8bit, 16bit, 32bit values
- *   - implement parallel traversal
  *   - implement inotify usage on client
  *   - implement audit gateway
  *   - implement native gateway
  *   - implement stdout gateway
- *   - extend hash table/bisect table as we go
+ *   - extend hash tables table as we go
  *   - accelerate looking for "all hostnames" and suchlike.
  *   - throttling
  *   - enforce limit on open journal files in journald and journalctl
+ *   - cryptographic hash
+ *   - fix space reservation logic
+ *   - comm, argv can be manipulated, should it be _COMM=, _CMDLINE= or COMM=, CMDLINE=?
  */
 
 typedef struct sd_journal sd_journal;
@@ -47,59 +51,52 @@ void sd_journal_close(sd_journal *j);
 int sd_journal_previous(sd_journal *j);
 int sd_journal_next(sd_journal *j);
 
+int sd_journal_previous_skip(sd_journal *j, uint64_t skip);
+int sd_journal_next_skip(sd_journal *j, uint64_t skip);
+
 int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret);
-int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret);
+int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id);
 int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *l);
 int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *l);
-void sd_journal_start_data(sd_journal *j);
+void sd_journal_restart_data(sd_journal *j);
 
 int sd_journal_add_match(sd_journal *j, const void *data, size_t size);
 void sd_journal_flush_matches(sd_journal *j);
 
 int sd_journal_seek_head(sd_journal *j);
 int sd_journal_seek_tail(sd_journal *j);
-
-int sd_journal_seek_monotonic_usec(sd_journal *j, uint64_t usec);  /* missing */
-int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec);   /* missing */
+int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec);
+int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec);
+int sd_journal_seek_cursor(sd_journal *j, const char *cursor);
 
 int sd_journal_get_cursor(sd_journal *j, char **cursor);
-int sd_journal_set_cursor(sd_journal *j, const char *cursor);      /* missing */
 
-int sd_journal_unique_seek(sd_journal *j, const char *field);      /* missing */
-int sd_journal_unique_enumerate(sd_journal *j, const void **data, size_t *l); /* missing */
-
-int sd_journal_get_fd(sd_journal *j);                              /* missing */
+int sd_journal_query_unique(sd_journal *j, const char *field);      /* missing */
+int sd_journal_enumerate_unique(sd_journal *j, const void **data, size_t *l); /* missing */
+void sd_journal_restart_unique(sd_journal *j);                      /* missing */
 
 enum {
         SD_JOURNAL_NOP,
         SD_JOURNAL_APPEND,
-        SD_JOURNAL_DROP
+        SD_JOURNAL_INVALIDATE_ADD,
+        SD_JOURNAL_INVALIDATE_REMOVE
 };
 
+int sd_journal_get_fd(sd_journal *j);                              /* missing */
 int sd_journal_process(sd_journal *j);                             /* missing */
 
-#define SD_JOURNAL_FOREACH_BEGIN(j)             \
-        if (sd_journal_seek_head(j) > 0) do {
-
-#define SD_JOURNAL_FOREACH_END(j)               \
-        } while (sd_journal_next(j) > 0)
-
-#define SD_JOURNAL_FOREACH_CONTINUE(j)          \
-        do {
-
-#define SD_JOURNAL_FOREACH_BACKWARDS_BEGIN(j)   \
-        if (sd_journal_seek_tail(j) > 0) do {
-
-#define SD_JOURNAL_FOREACH_BACKWARDS_END(j)     \
-        } while (sd_journal_previous(j) > 0)
+#define SD_JOURNAL_FOREACH(j)                                           \
+        if (sd_journal_seek_head(j) >= 0)                               \
+                while (sd_journal_next(j) > 0)                          \
 
-#define SD_JOURNAL_FOREACH_BACKWARDS_CONTINUE(j) \
-        do {
+#define SD_JOURNAL_FOREACH_BACKWARDS(j)                                 \
+        if (sd_journal_seek_tail(j) >= 0)                               \
+                while (sd_journal_previous(j) > 0)                      \
 
 #define SD_JOURNAL_FOREACH_DATA(j, data, l)                             \
-        for (sd_journal_start_data(j); sd_journal_enumerate_data((j), &(data), &(l)) > 0; )
+        for (sd_journal_restart_data(j); sd_journal_enumerate_data((j), &(data), &(l)) > 0; )
 
-#define SD_JOURNAL_FOREACH_UNIQUE(j, data, l)                      \
-        while (sd_journal_enumerate_unique_data((j), &(data), &(l)) > 0)
+#define SD_JOURNAL_FOREACH_UNIQUE(j, data, l)                           \
+        for (sd_journal_restart_unique(j); sd_journal_enumerate_data((j), &(data), &(l)) > 0; )
 
 #endif
index 93e2b4dab4039b6fb1484eb33670da8b2bfa0b64..45ced12b46ef3ca89e8278003c10c8fca4399f0f 100644 (file)
@@ -31,6 +31,7 @@ int main(int argc, char *argv[]) {
         struct iovec iovec;
         static const char test[] = "test", test2[] = "test2";
         Object *o;
+        uint64_t p;
 
         log_set_max_level(LOG_DEBUG);
 
@@ -54,41 +55,55 @@ int main(int argc, char *argv[]) {
 
         journal_file_dump(f);
 
-        assert(journal_file_next_entry(f, NULL, DIRECTION_DOWN, &o, NULL) == 1);
+        assert(journal_file_next_entry(f, NULL, 0, DIRECTION_DOWN, &o, &p) == 1);
         assert(le64toh(o->entry.seqnum) == 1);
 
-        assert(journal_file_next_entry(f, o, DIRECTION_DOWN, &o, NULL) == 1);
+        assert(journal_file_next_entry(f, o, p, DIRECTION_DOWN, &o, &p) == 1);
         assert(le64toh(o->entry.seqnum) == 2);
 
-        assert(journal_file_next_entry(f, o, DIRECTION_DOWN, &o, NULL) == 1);
+        assert(journal_file_next_entry(f, o, p, DIRECTION_DOWN, &o, &p) == 1);
         assert(le64toh(o->entry.seqnum) == 3);
 
-        assert(journal_file_next_entry(f, o, DIRECTION_DOWN, &o, NULL) == 0);
+        assert(journal_file_next_entry(f, o, p, DIRECTION_DOWN, &o, &p) == 0);
 
-        assert(journal_file_find_first_entry(f, test, strlen(test), DIRECTION_DOWN, &o, NULL) == 1);
+        assert(journal_file_next_entry(f, NULL, 0, DIRECTION_DOWN, &o, &p) == 1);
         assert(le64toh(o->entry.seqnum) == 1);
 
-        assert(journal_file_find_first_entry(f, test, strlen(test), DIRECTION_UP, &o, NULL) == 1);
+        assert(journal_file_skip_entry(f, o, p, 2, &o, &p) == 1);
         assert(le64toh(o->entry.seqnum) == 3);
 
-        assert(journal_file_find_first_entry(f, test2, strlen(test2), DIRECTION_UP, &o, NULL) == 1);
+        assert(journal_file_skip_entry(f, o, p, -2, &o, &p) == 1);
+        assert(le64toh(o->entry.seqnum) == 1);
+
+        assert(journal_file_skip_entry(f, o, p, -2, &o, &p) == 1);
+        assert(le64toh(o->entry.seqnum) == 1);
+
+        assert(journal_file_find_data_object(f, test, strlen(test), NULL, &p) == 1);
+        assert(journal_file_next_entry_for_data(f, NULL, 0, p, DIRECTION_DOWN, &o, NULL) == 1);
+        assert(le64toh(o->entry.seqnum) == 1);
+
+        assert(journal_file_next_entry_for_data(f, NULL, 0, p, DIRECTION_UP, &o, NULL) == 1);
+        assert(le64toh(o->entry.seqnum) == 3);
+
+        assert(journal_file_find_data_object(f, test2, strlen(test2), NULL, &p) == 1);
+        assert(journal_file_next_entry_for_data(f, NULL, 0, p, DIRECTION_UP, &o, NULL) == 1);
         assert(le64toh(o->entry.seqnum) == 2);
 
-        assert(journal_file_find_first_entry(f, test2, strlen(test2), DIRECTION_DOWN, &o, NULL) == 1);
+        assert(journal_file_next_entry_for_data(f, NULL, 0, p, DIRECTION_DOWN, &o, NULL) == 1);
         assert(le64toh(o->entry.seqnum) == 2);
 
-        assert(journal_file_find_first_entry(f, "quux", 4, DIRECTION_DOWN, &o, NULL) == 0);
+        assert(journal_file_find_data_object(f, "quux", 4, NULL, &p) == 0);
 
-        assert(journal_file_move_to_entry(f, 1, &o, NULL) == 1);
+        assert(journal_file_move_to_entry_by_seqnum(f, 1, DIRECTION_DOWN, &o, NULL) == 1);
         assert(le64toh(o->entry.seqnum) == 1);
 
-        assert(journal_file_move_to_entry(f, 3, &o, NULL) == 1);
+        assert(journal_file_move_to_entry_by_seqnum(f, 3, DIRECTION_DOWN, &o, NULL) == 1);
         assert(le64toh(o->entry.seqnum) == 3);
 
-        assert(journal_file_move_to_entry(f, 2, &o, NULL) == 1);
+        assert(journal_file_move_to_entry_by_seqnum(f, 2, DIRECTION_DOWN, &o, NULL) == 1);
         assert(le64toh(o->entry.seqnum) == 2);
 
-        assert(journal_file_move_to_entry(f, 10, &o, NULL) == 0);
+        assert(journal_file_move_to_entry_by_seqnum(f, 10, DIRECTION_DOWN, &o, NULL) == 0);
 
         journal_file_rotate(&f);
         journal_file_rotate(&f);
diff --git a/src/journal/wjournal.c b/src/journal/wjournal.c
deleted file mode 100644 (file)
index 3122aa0..0000000
+++ /dev/null
@@ -1,57 +0,0 @@
-/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
-
-/***
-  This file is part of systemd.
-
-  Copyright 2011 Lennart Poettering
-
-  systemd is free software; you can redistribute it and/or modify it
-  under the terms of the GNU General Public License as published by
-  the Free Software Foundation; either version 2 of the License, or
-  (at your option) any later version.
-
-  systemd is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with systemd; If not, see <http://www.gnu.org/licenses/>.
-***/
-
-#include "wjournal.h"
-#include "journal-def.h"
-
-struct WJournal {
-        int fd;
-
-        Header *header;
-        HashItem *hash_table;
-        uint64_t *bisect_table;
-};
-
-int wjournal_open(const char *fn, WJournal **ret) {
-        assert(fn);
-        assert(ret);
-}
-
-void wjournal_close(WJournal *j) {
-        assert(j);
-
-        if (j->fd >= 0)
-                close_nointr_nofail(j->fd);
-
-        if (j->header) {
-                munmap(j->header, PAGE_ALIGN(sizeof(Header)));
-
-        }
-
-        free(j);
-}
-
-int wjournal_write_object_begin(WJournal *j, uint64_t type, uint64_t size, Object **ret);
-int wjournal_write_object_finish(WJournal *j, Object *ret);
-
-int wjournal_write_field(WJournal *j, const char *buffer, uint64_t size, Object **ret);
-int wjournal_write_entry(WJournal *j, const Field *fields, unsigned n_fields, Object **ret);
-int wjournal_write_eof(WJournal *j);
diff --git a/src/journal/wjournal.h b/src/journal/wjournal.h
deleted file mode 100644 (file)
index b0250d0..0000000
+++ /dev/null
@@ -1,39 +0,0 @@
-/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
-
-#ifndef foojournalhfoo
-#define foojournalhfoo
-
-/***
-  This file is part of systemd.
-
-  Copyright 2011 Lennart Poettering
-
-  systemd is free software; you can redistribute it and/or modify it
-  under the terms of the GNU General Public License as published by
-  the Free Software Foundation; either version 2 of the License, or
-  (at your option) any later version.
-
-  systemd is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with systemd; If not, see <http://www.gnu.org/licenses/>.
-***/
-
-#include <inttypes.h>
-
-typedef struct WJournal WJournal;
-
-int wjournal_open(const char *fn, WJournal **ret);
-void wjournal_close(WJournal *j);
-
-int wjournal_write_object_begin(WJournal *j, uint64_t type, uint64_t size, Object **ret);
-int wjournal_write_object_finish(WJournal *j, Object *ret);
-
-int wjournal_write_field(WJournal *j, const char *buffer, uint64_t size, Object **ret);
-int wjournal_write_entry(WJournal *j, const Field *fields, unsigned n_fields, Object **ret);
-int wjournal_write_eof(WJournal *j);
-
-#endif