X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ianmdlvl/git?p=elogind.git;a=blobdiff_plain;f=src%2Fjournal%2Fjournal-file.c;h=2845e05ce0adcb7333e71c3c85cd5433a6727d12;hp=43fd4453a44accfde468c66c37f1e236072f43e2;hb=2a560338c471f47ca0caf6f1ec8c54a61e005d7f;hpb=3c1668da6202f1ead3d4d3981b89e9da1a0e98e3

diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index 43fd4453a..2845e05ce 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -26,11 +26,9 @@
 #include <sys/statvfs.h>
 #include <fcntl.h>
 #include <stddef.h>
+#include <linux/fs.h>
 
-#ifdef HAVE_XATTR
-#include <attr/xattr.h>
-#endif
-
+#include "btrfs-util.h"
 #include "journal-def.h"
 #include "journal-file.h"
 #include "journal-authenticate.h"
@@ -44,7 +42,7 @@
 #define COMPRESSION_SIZE_THRESHOLD (512ULL)
 
 /* This is the minimum journal file size */
-#define JOURNAL_FILE_SIZE_MIN (64ULL*1024ULL)                  /* 64 KiB */
+#define JOURNAL_FILE_SIZE_MIN (4ULL*1024ULL*1024ULL)           /* 4 MiB */
 
 /* These are the lower and upper bounds if we deduce the max_use value
  * from the file system size */
@@ -65,6 +63,71 @@
 /* n_data was the first entry we added after the initial file format design */
 #define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data))
 
+/* How many entries to keep in the entry array chain cache at max */
+#define CHAIN_CACHE_MAX 20
+
+/* How much to increase the journal file size at once each time we allocate something new. */
+#define FILE_SIZE_INCREASE (8ULL*1024ULL*1024ULL)              /* 8MB */
+
+/* Reread fstat() of the file for detecting deletions at least this often */
+#define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC)
+
+/* The mmap context to use for the header we pick as one above the last defined typed */
+#define CONTEXT_HEADER _OBJECT_TYPE_MAX
+
+static int journal_file_set_online(JournalFile *f) {
+        assert(f);
+
+        if (!f->writable)
+                return -EPERM;
+
+        if (!(f->fd >= 0 && f->header))
+                return -EINVAL;
+
+        if (mmap_cache_got_sigbus(f->mmap, f->fd))
+                return -EIO;
+
+        switch(f->header->state) {
+                case STATE_ONLINE:
+                        return 0;
+
+                case STATE_OFFLINE:
+                        f->header->state = STATE_ONLINE;
+                        fsync(f->fd);
+                        return 0;
+
+                default:
+                        return -EINVAL;
+        }
+}
+
+int journal_file_set_offline(JournalFile *f) {
+        assert(f);
+
+        if (!f->writable)
+                return -EPERM;
+
+        if (!(f->fd >= 0 && f->header))
+                return -EINVAL;
+
+        if (f->header->state != STATE_ONLINE)
+                return 0;
+
+        fsync(f->fd);
+
+        if (mmap_cache_got_sigbus(f->mmap, f->fd))
+                return -EIO;
+
+        f->header->state = STATE_OFFLINE;
+
+        if (mmap_cache_got_sigbus(f->mmap, f->fd))
+                return -EIO;
+
+        fsync(f->fd);
+
+        return 0;
+}
+
 void journal_file_close(JournalFile *f) {
         assert(f);
 
@@ -74,30 +137,32 @@ void journal_file_close(JournalFile *f) {
                 journal_file_append_tag(f);
 #endif
 
-        /* Sync everything to disk, before we mark the file offline */
+        journal_file_set_offline(f);
+
         if (f->mmap && f->fd >= 0)
                 mmap_cache_close_fd(f->mmap, f->fd);
 
-        if (f->writable && f->fd >= 0)
-                fdatasync(f->fd);
+        if (f->fd >= 0 && f->defrag_on_close) {
 
-        if (f->header) {
-                /* Mark the file offline. Don't override the archived state if it already is set */
-                if (f->writable && f->header->state == STATE_ONLINE)
-                        f->header->state = STATE_OFFLINE;
+                /* Be friendly to btrfs: turn COW back on again now,
+                 * and defragment the file. We won't write to the file
+                 * ever again, hence remove all fragmentation, and
+                 * reenable all the good bits COW usually provides
+                 * (such as data checksumming). */
 
-                munmap(f->header, PAGE_ALIGN(sizeof(Header)));
+                (void) chattr_fd(f->fd, false, FS_NOCOW_FL);
+                (void) btrfs_defrag_fd(f->fd);
         }
 
-        if (f->fd >= 0)
-                close_nointr_nofail(f->fd);
-
+        safe_close(f->fd);
         free(f->path);
 
         if (f->mmap)
                 mmap_cache_unref(f->mmap);
 
-#ifdef HAVE_XZ
+        ordered_hashmap_free_free(f->chain_cache);
+
+#if defined(HAVE_XZ) || defined(HAVE_LZ4)
         free(f->compress_buffer);
 #endif
 
@@ -117,21 +182,21 @@ void journal_file_close(JournalFile *f) {
 }
 
 static int journal_file_init_header(JournalFile *f, JournalFile *template) {
-        Header h;
+        Header h = {};
         ssize_t k;
         int r;
 
         assert(f);
 
-        zero(h);
         memcpy(h.signature, HEADER_SIGNATURE, 8);
         h.header_size = htole64(ALIGN64(sizeof(h)));
 
-        h.incompatible_flags =
-                htole32(f->compress ? HEADER_INCOMPATIBLE_COMPRESSED : 0);
+        h.incompatible_flags |= htole32(
+                f->compress_xz * HEADER_INCOMPATIBLE_COMPRESSED_XZ |
+                f->compress_lz4 * HEADER_INCOMPATIBLE_COMPRESSED_LZ4);
 
-        h.compatible_flags =
-                htole32(f->seal ? HEADER_COMPATIBLE_SEALED : 0);
+        h.compatible_flags = htole32(
+                f->seal * HEADER_COMPATIBLE_SEALED);
 
         r = sd_id128_randomize(&h.file_id);
         if (r < 0)
@@ -154,8 +219,8 @@ 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;
+        int r;
 
         assert(f);
 
@@ -172,16 +237,17 @@ static int journal_file_refresh_header(JournalFile *f) {
 
         f->header->boot_id = boot_id;
 
-        f->header->state = STATE_ONLINE;
+        r = journal_file_set_online(f);
 
         /* Sync the online state to disk */
-        msync(f->header, PAGE_ALIGN(sizeof(Header)), MS_SYNC);
-        fdatasync(f->fd);
+        fsync(f->fd);
 
-        return 0;
+        return r;
 }
 
 static int journal_file_verify_header(JournalFile *f) {
+        uint32_t flags;
+
         assert(f);
 
         if (memcmp(f->header->signature, HEADER_SIGNATURE, 8))
@@ -189,24 +255,30 @@ static int journal_file_verify_header(JournalFile *f) {
 
         /* In both read and write mode we refuse to open files with
          * incompatible flags we don't know */
-#ifdef HAVE_XZ
-        if ((le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) != 0)
+        flags = le32toh(f->header->incompatible_flags);
+        if (flags & ~HEADER_INCOMPATIBLE_SUPPORTED) {
+                if (flags & ~HEADER_INCOMPATIBLE_ANY)
+                        log_debug("Journal file %s has unknown incompatible flags %"PRIx32,
+                                  f->path, flags & ~HEADER_INCOMPATIBLE_ANY);
+                flags = (flags & HEADER_INCOMPATIBLE_ANY) & ~HEADER_INCOMPATIBLE_SUPPORTED;
+                if (flags)
+                        log_debug("Journal file %s uses incompatible flags %"PRIx32
+                                  " disabled at compilation time.", f->path, flags);
                 return -EPROTONOSUPPORT;
-#else
-        if (f->header->incompatible_flags != 0)
-                return -EPROTONOSUPPORT;
-#endif
+        }
 
         /* When open for writing we refuse to open files with
          * compatible flags, too */
-        if (f->writable) {
-#ifdef HAVE_GCRYPT
-                if ((le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SEALED) != 0)
-                        return -EPROTONOSUPPORT;
-#else
-                if (f->header->compatible_flags != 0)
-                        return -EPROTONOSUPPORT;
-#endif
+        flags = le32toh(f->header->compatible_flags);
+        if (f->writable && (flags & ~HEADER_COMPATIBLE_SUPPORTED)) {
+                if (flags & ~HEADER_COMPATIBLE_ANY)
+                        log_debug("Journal file %s has unknown compatible flags %"PRIx32,
+                                  f->path, flags & ~HEADER_COMPATIBLE_ANY);
+                flags = (flags & HEADER_COMPATIBLE_ANY) & ~HEADER_COMPATIBLE_SUPPORTED;
+                if (flags)
+                        log_debug("Journal file %s uses compatible flags %"PRIx32
+                                  " disabled at compilation time.", f->path, flags);
+                return -EPROTONOSUPPORT;
         }
 
         if (f->header->state >= _STATE_MAX)
@@ -231,12 +303,6 @@ static int journal_file_verify_header(JournalFile *f) {
             !VALID64(le64toh(f->header->entry_array_offset)))
                 return -ENODATA;
 
-        if (le64toh(f->header->data_hash_table_offset) < le64toh(f->header->header_size) ||
-            le64toh(f->header->field_hash_table_offset) < le64toh(f->header->header_size) ||
-            le64toh(f->header->tail_object_offset) < le64toh(f->header->header_size) ||
-            le64toh(f->header->entry_array_offset) < le64toh(f->header->header_size))
-                return -ENODATA;
-
         if (f->writable) {
                 uint8_t state;
                 sd_id128_t machine_id;
@@ -257,18 +323,35 @@ static int journal_file_verify_header(JournalFile *f) {
                 } else if (state == STATE_ARCHIVED)
                         return -ESHUTDOWN;
                 else if (state != STATE_OFFLINE) {
-                        log_debug("Journal file %s has unknown state %u.", f->path, state);
+                        log_debug("Journal file %s has unknown state %i.", f->path, state);
                         return -EBUSY;
                 }
         }
 
-        f->compress = JOURNAL_HEADER_COMPRESSED(f->header);
+        f->compress_xz = JOURNAL_HEADER_COMPRESSED_XZ(f->header);
+        f->compress_lz4 = JOURNAL_HEADER_COMPRESSED_LZ4(f->header);
 
         f->seal = JOURNAL_HEADER_SEALED(f->header);
 
         return 0;
 }
 
+static int journal_file_fstat(JournalFile *f) {
+        assert(f);
+        assert(f->fd >= 0);
+
+        if (fstat(f->fd, &f->last_stat) < 0)
+                return -errno;
+
+        f->last_stat_usec = now(CLOCK_MONOTONIC);
+
+        /* Refuse appending to files that are already deleted */
+        if (f->last_stat.st_nlink <= 0)
+                return -EIDRM;
+
+        return 0;
+}
+
 static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) {
         uint64_t old_size, new_size;
         int r;
@@ -279,6 +362,9 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
          * for sure, since we always call posix_fallocate()
          * ourselves */
 
+        if (mmap_cache_got_sigbus(f->mmap, f->fd))
+                return -EIO;
+
         old_size =
                 le64toh(f->header->header_size) +
                 le64toh(f->header->arena_size);
@@ -287,15 +373,26 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
         if (new_size < le64toh(f->header->header_size))
                 new_size = le64toh(f->header->header_size);
 
-        if (new_size <= old_size)
-                return 0;
+        if (new_size <= old_size) {
+
+                /* We already pre-allocated enough space, but before
+                 * we write to it, let's check with fstat() if the
+                 * file got deleted, in order make sure we don't throw
+                 * away the data immediately. Don't check fstat() for
+                 * all writes though, but only once ever 10s. */
 
-        if (f->metrics.max_size > 0 &&
-            new_size > f->metrics.max_size)
+                if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC))
+                        return 0;
+
+                return journal_file_fstat(f);
+        }
+
+        /* Allocate more space. */
+
+        if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
                 return -E2BIG;
 
-        if (new_size > f->metrics.min_size &&
-            f->metrics.keep_free > 0) {
+        if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) {
                 struct statvfs svfs;
 
                 if (fstatvfs(f->fd, &svfs) >= 0) {
@@ -313,6 +410,11 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
                 }
         }
 
+        /* Increase by larger blocks at once */
+        new_size = ((new_size+FILE_SIZE_INCREASE-1) / FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE;
+        if (f->metrics.max_size > 0 && new_size > f->metrics.max_size)
+                new_size = f->metrics.max_size;
+
         /* Note that the glibc fallocate() fallback is very
            inefficient, hence we try to minimize the allocation area
            as we can. */
@@ -320,15 +422,21 @@ static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size)
         if (r != 0)
                 return -r;
 
-        if (fstat(f->fd, &f->last_stat) < 0)
-                return -errno;
-
         f->header->arena_size = htole64(new_size - le64toh(f->header->header_size));
 
-        return 0;
+        return journal_file_fstat(f);
 }
 
-static int journal_file_move_to(JournalFile *f, int context, bool keep_always, uint64_t offset, uint64_t size, void **ret) {
+static unsigned type_to_context(ObjectType type) {
+        /* One context for each type, plus one catch-all for the rest */
+        assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS);
+        assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS);
+        return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0;
+}
+
+static int journal_file_move_to(JournalFile *f, ObjectType type, bool keep_always, uint64_t offset, uint64_t size, void **ret) {
+        int r;
+
         assert(f);
         assert(ret);
 
@@ -340,17 +448,20 @@ static int journal_file_move_to(JournalFile *f, int context, bool keep_always, u
                 /* Hmm, out of range? Let's refresh the fstat() data
                  * first, before we trust that check. */
 
-                if (fstat(f->fd, &f->last_stat) < 0 ||
-                    offset + size > (uint64_t) f->last_stat.st_size)
+                r = journal_file_fstat(f);
+                if (r < 0)
+                        return r;
+
+                if (offset + size > (uint64_t) f->last_stat.st_size)
                         return -EADDRNOTAVAIL;
         }
 
-        return mmap_cache_get(f->mmap, f->fd, f->prot, context, keep_always, offset, size, &f->last_stat, ret);
+        return mmap_cache_get(f->mmap, f->fd, f->prot, type_to_context(type), keep_always, offset, size, &f->last_stat, ret);
 }
 
 static uint64_t minimum_header_size(Object *o) {
 
-        static uint64_t table[] = {
+        static const uint64_t table[] = {
                 [OBJECT_DATA] = sizeof(DataObject),
                 [OBJECT_FIELD] = sizeof(FieldObject),
                 [OBJECT_ENTRY] = sizeof(EntryObject),
@@ -366,12 +477,11 @@ static uint64_t minimum_header_size(Object *o) {
         return table[o->object.type];
 }
 
-int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Object **ret) {
+int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) {
         int r;
         void *t;
         Object *o;
         uint64_t s;
-        unsigned context;
 
         assert(f);
         assert(ret);
@@ -380,10 +490,7 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec
         if (!VALID64(offset))
                 return -EFAULT;
 
-        /* One context for each type, plus one catch-all for the rest */
-        context = type > 0 && type < _OBJECT_TYPE_MAX ? type : 0;
-
-        r = journal_file_move_to(f, context, false, offset, sizeof(ObjectHeader), &t);
+        r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), &t);
         if (r < 0)
                 return r;
 
@@ -399,11 +506,11 @@ int journal_file_move_to_object(JournalFile *f, int type, uint64_t offset, Objec
         if (s < minimum_header_size(o))
                 return -EBADMSG;
 
-        if (type > 0 && o->object.type != type)
+        if (type > OBJECT_UNUSED && o->object.type != type)
                 return -EBADMSG;
 
         if (s > sizeof(ObjectHeader)) {
-                r = journal_file_move_to(f, o->object.type, false, offset, s, &t);
+                r = journal_file_move_to(f, type, false, offset, s, &t);
                 if (r < 0)
                         return r;
 
@@ -440,23 +547,27 @@ static uint64_t journal_file_entry_seqnum(JournalFile *f, uint64_t *seqnum) {
         return r;
 }
 
-int journal_file_append_object(JournalFile *f, int type, uint64_t size, Object **ret, uint64_t *offset) {
+int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret, uint64_t *offset) {
         int r;
         uint64_t p;
         Object *tail, *o;
         void *t;
 
         assert(f);
-        assert(type > 0 && type < _OBJECT_TYPE_MAX);
+        assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX);
         assert(size >= sizeof(ObjectHeader));
         assert(offset);
         assert(ret);
 
+        r = journal_file_set_online(f);
+        if (r < 0)
+                return r;
+
         p = le64toh(f->header->tail_object_offset);
         if (p == 0)
                 p = le64toh(f->header->header_size);
         else {
-                r = journal_file_move_to_object(f, -1, p, &tail);
+                r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail);
                 if (r < 0)
                         return r;
 
@@ -502,7 +613,7 @@ static int journal_file_setup_data_hash_table(JournalFile *f) {
         if (s < DEFAULT_DATA_HASH_TABLE_SIZE)
                 s = DEFAULT_DATA_HASH_TABLE_SIZE;
 
-        log_debug("Reserving %llu entries in hash table.", (unsigned long long) (s / sizeof(HashItem)));
+        log_debug("Reserving %"PRIu64" entries in hash table.", s / sizeof(HashItem));
 
         r = journal_file_append_object(f,
                                        OBJECT_DATA_HASH_TABLE,
@@ -511,7 +622,7 @@ static int journal_file_setup_data_hash_table(JournalFile *f) {
         if (r < 0)
                 return r;
 
-        memset(o->hash_table.items, 0, s);
+        memzero(o->hash_table.items, s);
 
         f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
         f->header->data_hash_table_size = htole64(s);
@@ -537,7 +648,7 @@ static int journal_file_setup_field_hash_table(JournalFile *f) {
         if (r < 0)
                 return r;
 
-        memset(o->hash_table.items, 0, s);
+        memzero(o->hash_table.items, s);
 
         f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items));
         f->header->field_hash_table_size = htole64(s);
@@ -595,7 +706,7 @@ static int journal_file_link_field(
                 uint64_t offset,
                 uint64_t hash) {
 
-        uint64_t p, h;
+        uint64_t p, h, m;
         int r;
 
         assert(f);
@@ -605,11 +716,14 @@ static int journal_file_link_field(
         if (o->object.type != OBJECT_FIELD)
                 return -EINVAL;
 
-        /* This might alter the window we are looking at */
+        m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem);
+        if (m <= 0)
+                return -EBADMSG;
 
+        /* This might alter the window we are looking at */
         o->field.next_hash_offset = o->field.head_data_offset = 0;
 
-        h = hash % (le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
+        h = hash % m;
         p = le64toh(f->field_hash_table[h].tail_hash_offset);
         if (p == 0)
                 f->field_hash_table[h].head_hash_offset = htole64(offset);
@@ -635,7 +749,7 @@ static int journal_file_link_data(
                 uint64_t offset,
                 uint64_t hash) {
 
-        uint64_t p, h;
+        uint64_t p, h, m;
         int r;
 
         assert(f);
@@ -645,13 +759,16 @@ static int journal_file_link_data(
         if (o->object.type != OBJECT_DATA)
                 return -EINVAL;
 
-        /* This might alter the window we are looking at */
+        m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem);
+        if (m <= 0)
+                return -EBADMSG;
 
+        /* This might alter the window we are looking at */
         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;
 
-        h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
+        h = hash % m;
         p = le64toh(f->data_hash_table[h].tail_hash_offset);
         if (p == 0)
                 /* Only entry in the hash table is easy */
@@ -680,7 +797,7 @@ int journal_file_find_field_object_with_hash(
                 const void *field, uint64_t size, uint64_t hash,
                 Object **ret, uint64_t *offset) {
 
-        uint64_t p, osize, h;
+        uint64_t p, osize, h, m;
         int r;
 
         assert(f);
@@ -688,10 +805,12 @@ int journal_file_find_field_object_with_hash(
 
         osize = offsetof(Object, field.payload) + size;
 
-        if (f->header->field_hash_table_size == 0)
+        m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem);
+
+        if (m <= 0)
                 return -EBADMSG;
 
-        h = hash % (le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
+        h = hash % m;
         p = le64toh(f->field_hash_table[h].head_hash_offset);
 
         while (p > 0) {
@@ -741,7 +860,7 @@ int journal_file_find_data_object_with_hash(
                 const void *data, uint64_t size, uint64_t hash,
                 Object **ret, uint64_t *offset) {
 
-        uint64_t p, osize, h;
+        uint64_t p, osize, h, m;
         int r;
 
         assert(f);
@@ -749,10 +868,11 @@ int journal_file_find_data_object_with_hash(
 
         osize = offsetof(Object, data.payload) + size;
 
-        if (f->header->data_hash_table_size == 0)
+        m = le64toh(f->header->data_hash_table_size) / sizeof(HashItem);
+        if (m <= 0)
                 return -EBADMSG;
 
-        h = hash % (le64toh(f->header->data_hash_table_size) / sizeof(HashItem));
+        h = hash % m;
         p = le64toh(f->data_hash_table[h].head_hash_offset);
 
         while (p > 0) {
@@ -765,9 +885,10 @@ int journal_file_find_data_object_with_hash(
                 if (le64toh(o->data.hash) != hash)
                         goto next;
 
-                if (o->object.flags & OBJECT_COMPRESSED) {
-#ifdef HAVE_XZ
-                        uint64_t l, rsize;
+                if (o->object.flags & OBJECT_COMPRESSION_MASK) {
+#if defined(HAVE_XZ) || defined(HAVE_LZ4)
+                        uint64_t l;
+                        size_t rsize;
 
                         l = le64toh(o->object.size);
                         if (l <= offsetof(Object, data.payload))
@@ -775,8 +896,10 @@ int journal_file_find_data_object_with_hash(
 
                         l -= offsetof(Object, data.payload);
 
-                        if (!uncompress_blob(o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize))
-                                return -EBADMSG;
+                        r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK,
+                                            o->data.payload, l, &f->compress_buffer, &f->compress_buffer_size, &rsize, 0);
+                        if (r < 0)
+                                return r;
 
                         if (rsize == size &&
                             memcmp(f->compress_buffer, data, size) == 0) {
@@ -792,7 +915,6 @@ int journal_file_find_data_object_with_hash(
 #else
                         return -EPROTONOSUPPORT;
 #endif
-
                 } else if (le64toh(o->object.size) == osize &&
                            memcmp(o->data.payload, data, size) == 0) {
 
@@ -860,6 +982,8 @@ static int journal_file_append_field(
 
         osize = offsetof(Object, field.payload) + size;
         r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p);
+        if (r < 0)
+                return r;
 
         o->field.hash = htole64(hash);
         memcpy(o->field.payload, field, size);
@@ -897,8 +1021,7 @@ static int journal_file_append_data(
         uint64_t hash, p;
         uint64_t osize;
         Object *o;
-        int r;
-        bool compressed = false;
+        int r, compression = 0;
         const void *eq;
 
         assert(f);
@@ -927,23 +1050,24 @@ static int journal_file_append_data(
 
         o->data.hash = htole64(hash);
 
-#ifdef HAVE_XZ
-        if (f->compress &&
+#if defined(HAVE_XZ) || defined(HAVE_LZ4)
+        if (f->compress_xz &&
             size >= COMPRESSION_SIZE_THRESHOLD) {
-                uint64_t rsize;
+                size_t rsize;
 
-                compressed = compress_blob(data, size, o->data.payload, &rsize);
+                compression = compress_blob(data, size, o->data.payload, &rsize);
 
-                if (compressed) {
+                if (compression) {
                         o->object.size = htole64(offsetof(Object, data.payload) + rsize);
-                        o->object.flags |= OBJECT_COMPRESSED;
+                        o->object.flags |= compression;
 
-                        log_debug("Compressed data object %lu -> %lu", (unsigned long) size, (unsigned long) rsize);
+                        log_debug("Compressed data object %"PRIu64" -> %zu using %s",
+                                  size, rsize, object_compressed_to_string(compression));
                 }
         }
 #endif
 
-        if (!compressed && size > 0)
+        if (!compression && size > 0)
                 memcpy(o->data.payload, data, size);
 
         r = journal_file_link_data(f, o, p, hash);
@@ -956,10 +1080,13 @@ static int journal_file_append_data(
         if (r < 0)
                 return r;
 
-        eq = memchr(data, '=', size);
+        if (!data)
+                eq = NULL;
+        else
+                eq = memchr(data, '=', size);
         if (eq && eq > data) {
+                Object *fo = NULL;
                 uint64_t fp;
-                Object *fo;
 
                 /* Create field object ... */
                 r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, &fp);
@@ -1159,7 +1286,7 @@ static int journal_file_link_entry(JournalFile *f, Object *o, uint64_t offset) {
         if (r < 0)
                 return r;
 
-        /* log_debug("=> %s seqnr=%lu n_entries=%lu", f->path, (unsigned long) o->entry.seqnum, (unsigned long) f->header->n_entries); */
+        /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */
 
         if (f->header->head_entry_realtime == 0)
                 f->header->head_entry_realtime = o->entry.realtime;
@@ -1239,7 +1366,7 @@ void journal_file_post_change(JournalFile *f) {
         __sync_synchronize();
 
         if (ftruncate(f->fd, f->last_stat.st_size) < 0)
-                log_error("Failed to truncate file to its own size: %m");
+                log_error_errno(errno, "Failed to truncate file to its own size: %m");
 }
 
 static int entry_item_cmp(const void *_a, const void *_b) {
@@ -1262,9 +1389,6 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
         assert(f);
         assert(iovec || n_iovec == 0);
 
-        if (!f->writable)
-                return -EPERM;
-
         if (!ts) {
                 dual_timestamp_get(&_ts);
                 ts = &_ts;
@@ -1281,7 +1405,7 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
 #endif
 
         /* alloca() can't take 0, hence let's allocate at least one */
-        items = alloca(sizeof(EntryItem) * MAX(1, n_iovec));
+        items = alloca(sizeof(EntryItem) * MAX(1u, n_iovec));
 
         for (i = 0; i < n_iovec; i++) {
                 uint64_t p;
@@ -1298,46 +1422,116 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
 
         /* Order by the position on disk, in order to improve seek
          * times for rotating media. */
-        qsort(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
+        qsort_safe(items, n_iovec, sizeof(EntryItem), entry_item_cmp);
 
         r = journal_file_append_entry_internal(f, ts, xor_hash, items, n_iovec, seqnum, ret, offset);
 
+        /* If the memory mapping triggered a SIGBUS then we return an
+         * IO error and ignore the error code passed down to us, since
+         * it is very likely just an effect of a nullified replacement
+         * mapping page */
+
+        if (mmap_cache_got_sigbus(f->mmap, f->fd))
+                r = -EIO;
+
         journal_file_post_change(f);
 
         return r;
 }
 
-static int generic_array_get(JournalFile *f,
-                             uint64_t first,
-                             uint64_t i,
-                             Object **ret, uint64_t *offset) {
+typedef struct ChainCacheItem {
+        uint64_t first; /* the array at the beginning of the chain */
+        uint64_t array; /* the cached array */
+        uint64_t begin; /* the first item in the cached array */
+        uint64_t total; /* the total number of items in all arrays before this one in the chain */
+        uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */
+} ChainCacheItem;
+
+static void chain_cache_put(
+                OrderedHashmap *h,
+                ChainCacheItem *ci,
+                uint64_t first,
+                uint64_t array,
+                uint64_t begin,
+                uint64_t total,
+                uint64_t last_index) {
+
+        if (!ci) {
+                /* If the chain item to cache for this chain is the
+                 * first one it's not worth caching anything */
+                if (array == first)
+                        return;
+
+                if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) {
+                        ci = ordered_hashmap_steal_first(h);
+                        assert(ci);
+                } else {
+                        ci = new(ChainCacheItem, 1);
+                        if (!ci)
+                                return;
+                }
+
+                ci->first = first;
+
+                if (ordered_hashmap_put(h, &ci->first, ci) < 0) {
+                        free(ci);
+                        return;
+                }
+        } else
+                assert(ci->first == first);
+
+        ci->array = array;
+        ci->begin = begin;
+        ci->total = total;
+        ci->last_index = last_index;
+}
+
+static int generic_array_get(
+                JournalFile *f,
+                uint64_t first,
+                uint64_t i,
+                Object **ret, uint64_t *offset) {
 
         Object *o;
-        uint64_t p = 0, a;
+        uint64_t p = 0, a, t = 0;
         int r;
+        ChainCacheItem *ci;
 
         assert(f);
 
         a = first;
+
+        /* Try the chain cache first */
+        ci = ordered_hashmap_get(f->chain_cache, &first);
+        if (ci && i > ci->total) {
+                a = ci->array;
+                i -= ci->total;
+                t = ci->total;
+        }
+
         while (a > 0) {
-                uint64_t n;
+                uint64_t k;
 
                 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
                 if (r < 0)
                         return r;
 
-                n = journal_file_entry_array_n_items(o);
-                if (i < n) {
+                k = journal_file_entry_array_n_items(o);
+                if (i < k) {
                         p = le64toh(o->entry_array.items[i]);
-                        break;
+                        goto found;
                 }
 
-                i -= n;
+                i -= k;
+                t += k;
                 a = le64toh(o->entry_array.next_entry_array_offset);
         }
 
-        if (a <= 0 || p <= 0)
-                return 0;
+        return 0;
+
+found:
+        /* Let's cache this item for the next invocation */
+        chain_cache_put(f->chain_cache, ci, first, a, le64toh(o->entry_array.items[0]), t, i);
 
         r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
         if (r < 0)
@@ -1352,11 +1546,12 @@ static int generic_array_get(JournalFile *f,
         return 1;
 }
 
-static int generic_array_get_plus_one(JournalFile *f,
-                                      uint64_t extra,
-                                      uint64_t first,
-                                      uint64_t i,
-                                      Object **ret, uint64_t *offset) {
+static int generic_array_get_plus_one(
+                JournalFile *f,
+                uint64_t extra,
+                uint64_t first,
+                uint64_t i,
+                Object **ret, uint64_t *offset) {
 
         Object *o;
 
@@ -1387,25 +1582,54 @@ enum {
         TEST_RIGHT
 };
 
-static int generic_array_bisect(JournalFile *f,
-                                uint64_t first,
-                                uint64_t n,
-                                uint64_t needle,
-                                int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
-                                direction_t direction,
-                                Object **ret,
-                                uint64_t *offset,
-                                uint64_t *idx) {
-
-        uint64_t a, p, t = 0, i = 0, last_p = 0;
+static int generic_array_bisect(
+                JournalFile *f,
+                uint64_t first,
+                uint64_t n,
+                uint64_t needle,
+                int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset,
+                uint64_t *idx) {
+
+        uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = (uint64_t) -1;
         bool subtract_one = false;
         Object *o, *array = NULL;
         int r;
+        ChainCacheItem *ci;
 
         assert(f);
         assert(test_object);
 
+        /* Start with the first array in the chain */
         a = first;
+
+        ci = ordered_hashmap_get(f->chain_cache, &first);
+        if (ci && n > ci->total) {
+                /* Ah, we have iterated this bisection array chain
+                 * previously! Let's see if we can skip ahead in the
+                 * chain, as far as the last time. But we can't jump
+                 * backwards in the chain, so let's check that
+                 * first. */
+
+                r = test_object(f, ci->begin, needle);
+                if (r < 0)
+                        return r;
+
+                if (r == TEST_LEFT) {
+                        /* OK, what we are looking for is right of the
+                         * begin of this EntryArray, so let's jump
+                         * straight to previously cached array in the
+                         * chain */
+
+                        a = ci->array;
+                        n -= ci->total;
+                        t = ci->total;
+                        last_index = ci->last_index;
+                }
+        }
+
         while (a > 0) {
                 uint64_t left, right, k, lp;
 
@@ -1433,6 +1657,58 @@ static int generic_array_bisect(JournalFile *f,
                 if (r == TEST_RIGHT) {
                         left = 0;
                         right -= 1;
+
+                        if (last_index != (uint64_t) -1) {
+                                assert(last_index <= right);
+
+                                /* If we cached the last index we
+                                 * looked at, let's try to not to jump
+                                 * too wildly around and see if we can
+                                 * limit the range to look at early to
+                                 * the immediate neighbors of the last
+                                 * index we looked at. */
+
+                                if (last_index > 0) {
+                                        uint64_t x = last_index - 1;
+
+                                        p = le64toh(array->entry_array.items[x]);
+                                        if (p <= 0)
+                                                return -EBADMSG;
+
+                                        r = test_object(f, p, needle);
+                                        if (r < 0)
+                                                return r;
+
+                                        if (r == TEST_FOUND)
+                                                r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
+
+                                        if (r == TEST_RIGHT)
+                                                right = x;
+                                        else
+                                                left = x + 1;
+                                }
+
+                                if (last_index < right) {
+                                        uint64_t y = last_index + 1;
+
+                                        p = le64toh(array->entry_array.items[y]);
+                                        if (p <= 0)
+                                                return -EBADMSG;
+
+                                        r = test_object(f, p, needle);
+                                        if (r < 0)
+                                                return r;
+
+                                        if (r == TEST_FOUND)
+                                                r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT;
+
+                                        if (r == TEST_RIGHT)
+                                                right = y;
+                                        else
+                                                left = y + 1;
+                                }
+                        }
+
                         for (;;) {
                                 if (left == right) {
                                         if (direction == DIRECTION_UP)
@@ -1443,8 +1719,8 @@ static int generic_array_bisect(JournalFile *f,
                                 }
 
                                 assert(left < right);
-
                                 i = (left + right) / 2;
+
                                 p = le64toh(array->entry_array.items[i]);
                                 if (p <= 0)
                                         return -EBADMSG;
@@ -1463,7 +1739,7 @@ static int generic_array_bisect(JournalFile *f,
                         }
                 }
 
-                if (k > n) {
+                if (k >= n) {
                         if (direction == DIRECTION_UP) {
                                 i = n;
                                 subtract_one = true;
@@ -1477,6 +1753,7 @@ static int generic_array_bisect(JournalFile *f,
 
                 n -= k;
                 t += k;
+                last_index = (uint64_t) -1;
                 a = le64toh(array->entry_array.next_entry_array_offset);
         }
 
@@ -1486,6 +1763,9 @@ found:
         if (subtract_one && t == 0 && i == 0)
                 return 0;
 
+        /* Let's cache this item for the next invocation */
+        chain_cache_put(f->chain_cache, ci, first, a, le64toh(array->entry_array.items[0]), t, subtract_one ? (i > 0 ? i-1 : (uint64_t) -1) : i);
+
         if (subtract_one && i == 0)
                 p = last_p;
         else if (subtract_one)
@@ -1509,16 +1789,17 @@ found:
         return 1;
 }
 
-static int generic_array_bisect_plus_one(JournalFile *f,
-                                         uint64_t extra,
-                                         uint64_t first,
-                                         uint64_t n,
-                                         uint64_t needle,
-                                         int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
-                                         direction_t direction,
-                                         Object **ret,
-                                         uint64_t *offset,
-                                         uint64_t *idx) {
+static int generic_array_bisect_plus_one(
+                JournalFile *f,
+                uint64_t extra,
+                uint64_t first,
+                uint64_t n,
+                uint64_t needle,
+                int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle),
+                direction_t direction,
+                Object **ret,
+                uint64_t *offset,
+                uint64_t *idx) {
 
         int r;
         bool step_back = false;
@@ -1581,7 +1862,7 @@ found:
         return 1;
 }
 
-static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
+_pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
         assert(f);
         assert(p > 0);
 
@@ -1593,23 +1874,6 @@ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) {
                 return TEST_RIGHT;
 }
 
-int journal_file_move_to_entry_by_offset(
-                JournalFile *f,
-                uint64_t p,
-                direction_t direction,
-                Object **ret,
-                uint64_t *offset) {
-
-        return generic_array_bisect(f,
-                                    le64toh(f->header->entry_array_offset),
-                                    le64toh(f->header->n_entries),
-                                    p,
-                                    test_object_offset,
-                                    direction,
-                                    ret, offset, NULL);
-}
-
-
 static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) {
         Object *o;
         int r;
@@ -1699,6 +1963,18 @@ static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) {
                 return TEST_RIGHT;
 }
 
+static int find_data_object_by_boot_id(
+                JournalFile *f,
+                sd_id128_t boot_id,
+                Object **o,
+                uint64_t *b) {
+
+        char t[sizeof("_BOOT_ID=")-1 + 32 + 1] = "_BOOT_ID=";
+
+        sd_id128_to_string(boot_id, t + 9);
+        return journal_file_find_data_object(f, t, sizeof(t) - 1, o, b);
+}
+
 int journal_file_move_to_entry_by_monotonic(
                 JournalFile *f,
                 sd_id128_t boot_id,
@@ -1707,14 +1983,12 @@ int journal_file_move_to_entry_by_monotonic(
                 Object **ret,
                 uint64_t *offset) {
 
-        char t[9+32+1] = "_BOOT_ID=";
         Object *o;
         int r;
 
         assert(f);
 
-        sd_id128_to_string(boot_id, t + 9);
-        r = journal_file_find_data_object(f, t, strlen(t), &o, NULL);
+        r = find_data_object_by_boot_id(f, boot_id, &o, NULL);
         if (r < 0)
                 return r;
         if (r == 0)
@@ -1730,28 +2004,97 @@ int journal_file_move_to_entry_by_monotonic(
                                              ret, offset, NULL);
 }
 
+void journal_file_reset_location(JournalFile *f) {
+        f->location_type = LOCATION_HEAD;
+        f->current_offset = 0;
+        f->current_seqnum = 0;
+        f->current_realtime = 0;
+        f->current_monotonic = 0;
+        zero(f->current_boot_id);
+        f->current_xor_hash = 0;
+}
+
+void journal_file_save_location(JournalFile *f, direction_t direction, Object *o, uint64_t offset) {
+        f->last_direction = direction;
+        f->location_type = LOCATION_SEEK;
+        f->current_offset = offset;
+        f->current_seqnum = le64toh(o->entry.seqnum);
+        f->current_realtime = le64toh(o->entry.realtime);
+        f->current_monotonic = le64toh(o->entry.monotonic);
+        f->current_boot_id = o->entry.boot_id;
+        f->current_xor_hash = le64toh(o->entry.xor_hash);
+}
+
+int journal_file_compare_locations(JournalFile *af, JournalFile *bf) {
+        assert(af);
+        assert(bf);
+        assert(af->location_type == LOCATION_SEEK);
+        assert(bf->location_type == LOCATION_SEEK);
+
+        /* If contents and timestamps match, these entries are
+         * identical, even if the seqnum does not match */
+        if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) &&
+            af->current_monotonic == bf->current_monotonic &&
+            af->current_realtime == bf->current_realtime &&
+            af->current_xor_hash == bf->current_xor_hash)
+                return 0;
+
+        if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) {
+
+                /* If this is from the same seqnum source, compare
+                 * seqnums */
+                if (af->current_seqnum < bf->current_seqnum)
+                        return -1;
+                if (af->current_seqnum > bf->current_seqnum)
+                        return 1;
+
+                /* Wow! This is weird, different data but the same
+                 * seqnums? Something is borked, but let's make the
+                 * best of it and compare by time. */
+        }
+
+        if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) {
+
+                /* If the boot id matches, compare monotonic time */
+                if (af->current_monotonic < bf->current_monotonic)
+                        return -1;
+                if (af->current_monotonic > bf->current_monotonic)
+                        return 1;
+        }
+
+        /* Otherwise, compare UTC time */
+        if (af->current_realtime < bf->current_realtime)
+                return -1;
+        if (af->current_realtime > bf->current_realtime)
+                return 1;
+
+        /* Finally, compare by contents */
+        if (af->current_xor_hash < bf->current_xor_hash)
+                return -1;
+        if (af->current_xor_hash > bf->current_xor_hash)
+                return 1;
+
+        return 0;
+}
+
 int journal_file_next_entry(
                 JournalFile *f,
-                Object *o, uint64_t p,
+                uint64_t p,
                 direction_t direction,
                 Object **ret, uint64_t *offset) {
 
-        uint64_t i, n;
+        uint64_t i, n, ofs;
         int r;
 
         assert(f);
-        assert(p > 0 || !o);
 
         n = le64toh(f->header->n_entries);
         if (n <= 0)
                 return 0;
 
-        if (!o)
+        if (p == 0)
                 i = direction == DIRECTION_DOWN ? 0 : n - 1;
         else {
-                if (o->object.type != OBJECT_ENTRY)
-                        return -EINVAL;
-
                 r = generic_array_bisect(f,
                                          le64toh(f->header->entry_array_offset),
                                          le64toh(f->header->n_entries),
@@ -1777,59 +2120,24 @@ int journal_file_next_entry(
         }
 
         /* And jump to it */
-        return generic_array_get(f,
-                                 le64toh(f->header->entry_array_offset),
-                                 i,
-                                 ret, offset);
-}
-
-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);
+        r = generic_array_get(f,
+                              le64toh(f->header->entry_array_offset),
+                              i,
+                              ret, &ofs);
         if (r <= 0)
                 return r;
 
-        /* Calculate new index */
-        if (skip < 0) {
-                if ((uint64_t) -skip >= i)
-                        i = 0;
-                else
-                        i = i - (uint64_t) -skip;
-        } else
-                i  += (uint64_t) skip;
-
-        n = le64toh(f->header->n_entries);
-        if (n <= 0)
+        if (p > 0 &&
+            (direction == DIRECTION_DOWN ? ofs <= p : ofs >= p)) {
+                log_debug("%s: entry array corrupted at entry %"PRIu64,
+                          f->path, i);
                 return -EBADMSG;
+        }
 
-        if (i >= n)
-                i = n-1;
+        if (offset)
+                *offset = ofs;
 
-        return generic_array_get(f,
-                                 le64toh(f->header->entry_array_offset),
-                                 i,
-                                 ret, offset);
+        return 1;
 }
 
 int journal_file_next_entry_for_data(
@@ -1928,7 +2236,6 @@ int journal_file_move_to_entry_by_monotonic_for_data(
                 direction_t direction,
                 Object **ret, uint64_t *offset) {
 
-        char t[9+32+1] = "_BOOT_ID=";
         Object *o, *d;
         int r;
         uint64_t b, z;
@@ -1936,8 +2243,7 @@ int journal_file_move_to_entry_by_monotonic_for_data(
         assert(f);
 
         /* First, seek by time */
-        sd_id128_to_string(boot_id, t + 9);
-        r = journal_file_find_data_object(f, t, strlen(t), &o, &b);
+        r = find_data_object_by_boot_id(f, boot_id, &o, &b);
         if (r < 0)
                 return r;
         if (r == 0)
@@ -2003,8 +2309,6 @@ int journal_file_move_to_entry_by_monotonic_for_data(
 
                 z = q;
         }
-
-        return 0;
 }
 
 int journal_file_move_to_entry_by_seqnum_for_data(
@@ -2070,7 +2374,7 @@ void journal_file_dump(JournalFile *f) {
 
         p = le64toh(f->header->header_size);
         while (p != 0) {
-                r = journal_file_move_to_object(f, -1, p, &o);
+                r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o);
                 if (r < 0)
                         goto fail;
 
@@ -2089,10 +2393,10 @@ void journal_file_dump(JournalFile *f) {
                         break;
 
                 case OBJECT_ENTRY:
-                        printf("Type: OBJECT_ENTRY seqnum=%llu monotonic=%llu realtime=%llu\n",
-                               (unsigned long long) le64toh(o->entry.seqnum),
-                               (unsigned long long) le64toh(o->entry.monotonic),
-                               (unsigned long long) le64toh(o->entry.realtime));
+                        printf("Type: OBJECT_ENTRY seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n",
+                               le64toh(o->entry.seqnum),
+                               le64toh(o->entry.monotonic),
+                               le64toh(o->entry.realtime));
                         break;
 
                 case OBJECT_FIELD_HASH_TABLE:
@@ -2108,18 +2412,19 @@ void journal_file_dump(JournalFile *f) {
                         break;
 
                 case OBJECT_TAG:
-                        printf("Type: OBJECT_TAG seqnum=%llu epoch=%llu\n",
-                               (unsigned long long) le64toh(o->tag.seqnum),
-                               (unsigned long long) le64toh(o->tag.epoch));
+                        printf("Type: OBJECT_TAG seqnum=%"PRIu64" epoch=%"PRIu64"\n",
+                               le64toh(o->tag.seqnum),
+                               le64toh(o->tag.epoch));
                         break;
 
                 default:
-                        printf("Type: unknown (%u)\n", o->object.type);
+                        printf("Type: unknown (%i)\n", o->object.type);
                         break;
                 }
 
-                if (o->object.flags & OBJECT_COMPRESSED)
-                        printf("Flags: COMPRESSED\n");
+                if (o->object.flags & OBJECT_COMPRESSION_MASK)
+                        printf("Flags: %s\n",
+                               object_compressed_to_string(o->object.flags & OBJECT_COMPRESSION_MASK));
 
                 if (p == le64toh(f->header->tail_object_offset))
                         p = 0;
@@ -2132,9 +2437,18 @@ fail:
         log_error("File corrupt");
 }
 
+static const char* format_timestamp_safe(char *buf, size_t l, usec_t t) {
+        const char *x;
+
+        x = format_timestamp(buf, l, t);
+        if (x)
+                return x;
+        return " --- ";
+}
+
 void journal_file_print_header(JournalFile *f) {
-        char a[33], b[33], c[33];
-        char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX];
+        char a[33], b[33], c[33], d[33];
+        char x[FORMAT_TIMESTAMP_MAX], y[FORMAT_TIMESTAMP_MAX], z[FORMAT_TIMESTAMP_MAX];
         struct stat st;
         char bytes[FORMAT_BYTES_MAX];
 
@@ -2147,60 +2461,63 @@ void journal_file_print_header(JournalFile *f) {
                "Sequential Number ID: %s\n"
                "State: %s\n"
                "Compatible Flags:%s%s\n"
-               "Incompatible Flags:%s%s\n"
-               "Header size: %llu\n"
-               "Arena size: %llu\n"
-               "Data Hash Table Size: %llu\n"
-               "Field Hash Table Size: %llu\n"
+               "Incompatible Flags:%s%s%s\n"
+               "Header size: %"PRIu64"\n"
+               "Arena size: %"PRIu64"\n"
+               "Data Hash Table Size: %"PRIu64"\n"
+               "Field Hash Table Size: %"PRIu64"\n"
                "Rotate Suggested: %s\n"
-               "Head Sequential Number: %llu\n"
-               "Tail Sequential Number: %llu\n"
+               "Head Sequential Number: %"PRIu64"\n"
+               "Tail Sequential Number: %"PRIu64"\n"
                "Head Realtime Timestamp: %s\n"
                "Tail Realtime Timestamp: %s\n"
-               "Objects: %llu\n"
-               "Entry Objects: %llu\n",
+               "Tail Monotonic Timestamp: %s\n"
+               "Objects: %"PRIu64"\n"
+               "Entry Objects: %"PRIu64"\n",
                f->path,
                sd_id128_to_string(f->header->file_id, a),
                sd_id128_to_string(f->header->machine_id, b),
                sd_id128_to_string(f->header->boot_id, c),
-               sd_id128_to_string(f->header->seqnum_id, c),
+               sd_id128_to_string(f->header->seqnum_id, d),
                f->header->state == STATE_OFFLINE ? "OFFLINE" :
                f->header->state == STATE_ONLINE ? "ONLINE" :
                f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN",
                JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "",
-               (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SEALED) ? " ???" : "",
-               JOURNAL_HEADER_COMPRESSED(f->header) ? " COMPRESSED" : "",
-               (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_COMPRESSED) ? " ???" : "",
-               (unsigned long long) le64toh(f->header->header_size),
-               (unsigned long long) le64toh(f->header->arena_size),
-               (unsigned long long) le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
-               (unsigned long long) le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
+               (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "",
+               JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "",
+               JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "",
+               (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "",
+               le64toh(f->header->header_size),
+               le64toh(f->header->arena_size),
+               le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+               le64toh(f->header->field_hash_table_size) / sizeof(HashItem),
                yes_no(journal_file_rotate_suggested(f, 0)),
-               (unsigned long long) le64toh(f->header->head_entry_seqnum),
-               (unsigned long long) le64toh(f->header->tail_entry_seqnum),
-               format_timestamp(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
-               format_timestamp(y, sizeof(y), le64toh(f->header->tail_entry_realtime)),
-               (unsigned long long) le64toh(f->header->n_objects),
-               (unsigned long long) le64toh(f->header->n_entries));
+               le64toh(f->header->head_entry_seqnum),
+               le64toh(f->header->tail_entry_seqnum),
+               format_timestamp_safe(x, sizeof(x), le64toh(f->header->head_entry_realtime)),
+               format_timestamp_safe(y, sizeof(y), le64toh(f->header->tail_entry_realtime)),
+               format_timespan(z, sizeof(z), le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC),
+               le64toh(f->header->n_objects),
+               le64toh(f->header->n_entries));
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
-                printf("Data Objects: %llu\n"
+                printf("Data Objects: %"PRIu64"\n"
                        "Data Hash Table Fill: %.1f%%\n",
-                       (unsigned long long) le64toh(f->header->n_data),
+                       le64toh(f->header->n_data),
                        100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))));
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
-                printf("Field Objects: %llu\n"
+                printf("Field Objects: %"PRIu64"\n"
                        "Field Hash Table Fill: %.1f%%\n",
-                       (unsigned long long) le64toh(f->header->n_fields),
+                       le64toh(f->header->n_fields),
                        100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))));
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_tags))
-                printf("Tag Objects: %llu\n",
-                       (unsigned long long) le64toh(f->header->n_tags));
+                printf("Tag Objects: %"PRIu64"\n",
+                       le64toh(f->header->n_tags));
         if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays))
-                printf("Entry Array Objects: %llu\n",
-                       (unsigned long long) le64toh(f->header->n_entry_arrays));
+                printf("Entry Array Objects: %"PRIu64"\n",
+                       le64toh(f->header->n_entry_arrays));
 
         if (fstat(f->fd, &st) >= 0)
                 printf("Disk usage: %s\n", format_bytes(bytes, sizeof(bytes), (off_t) st.st_blocks * 512ULL));
@@ -2217,9 +2534,10 @@ int journal_file_open(
                 JournalFile *template,
                 JournalFile **ret) {
 
+        bool newly_created = false;
         JournalFile *f;
+        void *h;
         int r;
-        bool newly_created = false;
 
         assert(fname);
         assert(ret);
@@ -2242,8 +2560,10 @@ int journal_file_open(
         f->flags = flags;
         f->prot = prot_from_flags(flags);
         f->writable = (flags & O_ACCMODE) != O_RDONLY;
-#ifdef HAVE_XZ
-        f->compress = compress;
+#if defined(HAVE_LZ4)
+        f->compress_lz4 = compress;
+#elif defined(HAVE_XZ)
+        f->compress_xz = compress;
 #endif
 #ifdef HAVE_GCRYPT
         f->seal = seal;
@@ -2265,20 +2585,34 @@ int journal_file_open(
                 goto fail;
         }
 
+        f->chain_cache = ordered_hashmap_new(&uint64_hash_ops);
+        if (!f->chain_cache) {
+                r = -ENOMEM;
+                goto fail;
+        }
+
         f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode);
         if (f->fd < 0) {
                 r = -errno;
                 goto fail;
         }
 
-        if (fstat(f->fd, &f->last_stat) < 0) {
-                r = -errno;
+        r = journal_file_fstat(f);
+        if (r < 0)
                 goto fail;
-        }
 
         if (f->last_stat.st_size == 0 && f->writable) {
-#ifdef HAVE_XATTR
-                uint64_t crtime;
+
+                /* Before we write anything, turn off COW logic. Given
+                 * our write pattern that is quite unfriendly to COW
+                 * file systems this should greatly improve
+                 * performance on COW file systems, such as btrfs, at
+                 * the expense of data integrity features (which
+                 * shouldn't be too bad, given that we do our own
+                 * checksumming). */
+                r = chattr_fd(f->fd, true, FS_NOCOW_FL);
+                if (r < 0)
+                        log_warning_errno(errno, "Failed to set file attributes: %m");
 
                 /* Let's attach the creation time to the journal file,
                  * so that the vacuuming code knows the age of this
@@ -2288,11 +2622,9 @@ int journal_file_open(
                  * currently no usable API to query this, hence let's
                  * emulate this via extended attributes. If extended
                  * attributes are not supported we'll just skip this,
-                 * and rely solely on mtime/atime/ctime of the file.*/
+                 * and rely solely on mtime/atime/ctime of the file. */
 
-                crtime = htole64((uint64_t) now(CLOCK_REALTIME));
-                fsetxattr(f->fd, "user.crtime_usec", &crtime, sizeof(crtime), XATTR_CREATE);
-#endif
+                fd_setcrtime(f->fd, 0);
 
 #ifdef HAVE_GCRYPT
                 /* Try to load the FSPRG state, and if we can't, then
@@ -2308,10 +2640,9 @@ int journal_file_open(
                 if (r < 0)
                         goto fail;
 
-                if (fstat(f->fd, &f->last_stat) < 0) {
-                        r = -errno;
+                r = journal_file_fstat(f);
+                if (r < 0)
                         goto fail;
-                }
 
                 newly_created = true;
         }
@@ -2321,13 +2652,14 @@ int journal_file_open(
                 goto fail;
         }
 
-        f->header = mmap(NULL, PAGE_ALIGN(sizeof(Header)), prot_from_flags(flags), MAP_SHARED, f->fd, 0);
-        if (f->header == MAP_FAILED) {
-                f->header = NULL;
+        r = mmap_cache_get(f->mmap, f->fd, f->prot, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h);
+        if (r < 0) {
                 r = -errno;
                 goto fail;
         }
 
+        f->header = h;
+
         if (!newly_created) {
                 r = journal_file_verify_header(f);
                 if (r < 0)
@@ -2384,17 +2716,25 @@ int journal_file_open(
         if (r < 0)
                 goto fail;
 
+        if (mmap_cache_got_sigbus(f->mmap, f->fd)) {
+                r = -EIO;
+                goto fail;
+        }
+
         *ret = f;
         return 0;
 
 fail:
+        if (f->fd >= 0 && mmap_cache_got_sigbus(f->mmap, f->fd))
+                r = -EIO;
+
         journal_file_close(f);
 
         return r;
 }
 
 int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
-        char *p;
+        _cleanup_free_ char *p = NULL;
         size_t l;
         JournalFile *old_file, *new_file = NULL;
         int r;
@@ -2411,27 +2751,28 @@ int journal_file_rotate(JournalFile **f, bool compress, bool seal) {
                 return -EINVAL;
 
         l = strlen(old_file->path);
-
-        p = new(char, l + 1 + 32 + 1 + 16 + 1 + 16 + 1);
-        if (!p)
+        r = asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal",
+                     (int) l - 8, old_file->path,
+                     SD_ID128_FORMAT_VAL(old_file->header->seqnum_id),
+                     le64toh((*f)->header->head_entry_seqnum),
+                     le64toh((*f)->header->head_entry_realtime));
+        if (r < 0)
                 return -ENOMEM;
 
-        memcpy(p, old_file->path, l - 8);
-        p[l-8] = '@';
-        sd_id128_to_string(old_file->header->seqnum_id, p + l - 8 + 1);
-        snprintf(p + l - 8 + 1 + 32, 1 + 16 + 1 + 16 + 8 + 1,
-                 "-%016llx-%016llx.journal",
-                 (unsigned long long) le64toh((*f)->header->head_entry_seqnum),
-                 (unsigned long long) le64toh((*f)->header->head_entry_realtime));
-
+        /* Try to rename the file to the archived version. If the file
+         * already was deleted, we'll get ENOENT, let's ignore that
+         * case. */
         r = rename(old_file->path, p);
-        free(p);
-
-        if (r < 0)
+        if (r < 0 && errno != ENOENT)
                 return -errno;
 
         old_file->header->state = STATE_ARCHIVED;
 
+        /* Currently, btrfs is not very good with out write patterns
+         * and fragments heavily. Let's defrag our journal files when
+         * we archive them */
+        old_file->defrag_on_close = true;
+
         r = journal_file_open(old_file->path, old_file->flags, old_file->mode, compress, seal, NULL, old_file->mmap, old_file, &new_file);
         journal_file_close(old_file);
 
@@ -2452,7 +2793,7 @@ int journal_file_open_reliably(
 
         int r;
         size_t l;
-        char *p;
+        _cleanup_free_ char *p = NULL;
 
         r = journal_file_open(fname, flags, mode, compress, seal,
                               metrics, mmap_cache, template, ret);
@@ -2461,7 +2802,9 @@ int journal_file_open_reliably(
             r != -EHOSTDOWN && /* other machine */
             r != -EPROTONOSUPPORT && /* incompatible feature */
             r != -EBUSY && /* unclean shutdown */
-            r != -ESHUTDOWN /* already archived */)
+            r != -ESHUTDOWN && /* already archived */
+            r != -EIO && /* IO error, including SIGBUS on mmap */
+            r != -EIDRM /* File has been deleted */)
                 return r;
 
         if ((flags & O_ACCMODE) == O_RDONLY)
@@ -2476,24 +2819,28 @@ int journal_file_open_reliably(
         /* The file is corrupted. Rotate it away and try it again (but only once) */
 
         l = strlen(fname);
-        if (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
-                     (int) (l-8), fname,
+        if (asprintf(&p, "%.*s@%016llx-%016" PRIx64 ".journal~",
+                     (int) l - 8, fname,
                      (unsigned long long) now(CLOCK_REALTIME),
-                     random_ull()) < 0)
+                     random_u64()) < 0)
                 return -ENOMEM;
 
         r = rename(fname, p);
-        free(p);
         if (r < 0)
                 return -errno;
 
+        /* btrfs doesn't cope well with our write pattern and
+         * fragments heavily. Let's defrag all files we rotate */
+
+        (void) chattr_path(p, false, FS_NOCOW_FL);
+        (void) btrfs_defrag(p);
+
         log_warning("File %s corrupted or uncleanly shut down, renaming and replacing.", fname);
 
         return journal_file_open(fname, flags, mode, compress, seal,
                                  metrics, mmap_cache, template, ret);
 }
 
-
 int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p, uint64_t *seqnum, Object **ret, uint64_t *offset) {
         uint64_t i, n;
         uint64_t q, xor_hash = 0;
@@ -2512,12 +2859,9 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
         ts.monotonic = le64toh(o->entry.monotonic);
         ts.realtime = le64toh(o->entry.realtime);
 
-        if (to->tail_entry_monotonic_valid &&
-            ts.monotonic < le64toh(to->header->tail_entry_monotonic))
-                return -EINVAL;
-
         n = journal_file_entry_n_items(o);
-        items = alloca(sizeof(EntryItem) * n);
+        /* alloca() can't take 0, hence let's allocate at least one */
+        items = alloca(sizeof(EntryItem) * MAX(1u, n));
 
         for (i = 0; i < n; i++) {
                 uint64_t l, h;
@@ -2543,12 +2887,14 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
                 if ((uint64_t) t != l)
                         return -E2BIG;
 
-                if (o->object.flags & OBJECT_COMPRESSED) {
-#ifdef HAVE_XZ
-                        uint64_t rsize;
+                if (o->object.flags & OBJECT_COMPRESSION_MASK) {
+#if defined(HAVE_XZ) || defined(HAVE_LZ4)
+                        size_t rsize;
 
-                        if (!uncompress_blob(o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize))
-                                return -EBADMSG;
+                        r = decompress_blob(o->object.flags & OBJECT_COMPRESSION_MASK,
+                                            o->data.payload, l, &from->compress_buffer, &from->compress_buffer_size, &rsize, 0);
+                        if (r < 0)
+                                return r;
 
                         data = from->compress_buffer;
                         l = rsize;
@@ -2571,7 +2917,12 @@ int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint6
                         return r;
         }
 
-        return journal_file_append_entry_internal(to, &ts, xor_hash, items, n, seqnum, ret, offset);
+        r = journal_file_append_entry_internal(to, &ts, xor_hash, items, n, seqnum, ret, offset);
+
+        if (mmap_cache_got_sigbus(to->mmap, to->fd))
+                return -EIO;
+
+        return r;
 }
 
 void journal_default_metrics(JournalMetrics *m, int fd) {
@@ -2633,7 +2984,7 @@ void journal_default_metrics(JournalMetrics *m, int fd) {
         if (m->keep_free == (uint64_t) -1) {
 
                 if (fs_size > 0) {
-                        m->keep_free = PAGE_ALIGN(fs_size / 20); /* 5% of file system size */
+                        m->keep_free = PAGE_ALIGN(fs_size * 3 / 20); /* 15% of file system size */
 
                         if (m->keep_free > DEFAULT_KEEP_FREE_UPPER)
                                 m->keep_free = DEFAULT_KEEP_FREE_UPPER;
@@ -2671,7 +3022,6 @@ int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *from, usec_t *
 }
 
 int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *from, usec_t *to) {
-        char t[9+32+1] = "_BOOT_ID=";
         Object *o;
         uint64_t p;
         int r;
@@ -2679,9 +3029,7 @@ int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, u
         assert(f);
         assert(from || to);
 
-        sd_id128_to_string(boot_id, t + 9);
-
-        r = journal_file_find_data_object(f, t, strlen(t), &o, &p);
+        r = find_data_object_by_boot_id(f, boot_id, &o, &p);
         if (r <= 0)
                 return r;
 
@@ -2733,26 +3081,33 @@ bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec) {
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_data))
                 if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) {
-                        log_debug("Data hash table of %s has a fill level at %.1f (%llu of %llu items, %llu file size, %llu bytes per hash table item), suggesting rotation.",
+                        log_debug("Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" bytes per hash table item), suggesting rotation.",
                                   f->path,
                                   100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))),
-                                  (unsigned long long) le64toh(f->header->n_data),
-                                  (unsigned long long) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)),
-                                  (unsigned long long) (f->last_stat.st_size),
-                                  (unsigned long long) (f->last_stat.st_size / le64toh(f->header->n_data)));
+                                  le64toh(f->header->n_data),
+                                  le64toh(f->header->data_hash_table_size) / sizeof(HashItem),
+                                  (unsigned long long) f->last_stat.st_size,
+                                  f->last_stat.st_size / le64toh(f->header->n_data));
                         return true;
                 }
 
         if (JOURNAL_HEADER_CONTAINS(f->header, n_fields))
                 if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) {
-                        log_debug("Field hash table of %s has a fill level at %.1f (%llu of %llu items), suggesting rotation.",
+                        log_debug("Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.",
                                   f->path,
                                   100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))),
-                                  (unsigned long long) le64toh(f->header->n_fields),
-                                  (unsigned long long) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)));
+                                  le64toh(f->header->n_fields),
+                                  le64toh(f->header->field_hash_table_size) / sizeof(HashItem));
                         return true;
                 }
 
+        /* Are the data objects properly indexed by field objects? */
+        if (JOURNAL_HEADER_CONTAINS(f->header, n_data) &&
+            JOURNAL_HEADER_CONTAINS(f->header, n_fields) &&
+            le64toh(f->header->n_data) > 0 &&
+            le64toh(f->header->n_fields) == 0)
+                return true;
+
         if (max_file_usec > 0) {
                 usec_t t, h;