journal: introduce entry array chain cache

When traversing entry array chains for a bisection or for retrieving an
item by index we previously always started at the beginning of the
chain. Since we tend to look at the same chains repeatedly, let's cache
where we have been the last time, and maybe we can skip ahead with this
the next time.

This turns most bisections and index lookups from O(log(n)*log(n)) into
O(log(n)). More importantly however, we seek around on disk much less,
which is good to reduce buffer cache and seek times on rotational disks.

3 files changed, 157 insertions, 8 deletions

diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index edf8e7dd5e..6c9deacf2c 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -65,6 +65,9 @@
 /* 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
+
 void journal_file_close(JournalFile *f) {
         assert(f);
 
@@ -97,6 +100,8 @@ void journal_file_close(JournalFile *f) {
         if (f->mmap)
                 mmap_cache_unref(f->mmap);
 
+        hashmap_free_free(f->chain_cache);
+
 #ifdef HAVE_XZ
         free(f->compress_buffer);
 #endif
@@ -1307,37 +1312,89 @@ int journal_file_append_entry(JournalFile *f, const dual_timestamp *ts, const st
         return r;
 }
 
+typedef struct ChainCacheItem {
+        uint64_t first; /* the array at the begin of the chain */
+        uint64_t array; /* the cached array */
+        uint64_t begin; /* the first item in the cached array */
+        uint64_t total; /* the total number of items in all arrays before this one in the chain */
+} ChainCacheItem;
+
+static void chain_cache_put(
+                Hashmap *h,
+                ChainCacheItem *ci,
+                uint64_t first,
+                uint64_t array,
+                uint64_t begin,
+                uint64_t total) {
+
+        if (!ci) {
+                if (hashmap_size(h) >= CHAIN_CACHE_MAX)
+                        ci = hashmap_steal_first(h);
+                else {
+                        ci = new(ChainCacheItem, 1);
+                        if (!ci)
+                                return;
+                }
+
+                ci->first = first;
+
+                if (hashmap_put(h, &ci->first, ci) < 0) {
+                        free(ci);
+                        return;
+                }
+        } else
+                assert(ci->first == first);
+
+        ci->array = array;
+        ci->begin = begin;
+        ci->total = total;
+}
+
 static int generic_array_get(JournalFile *f,
                              uint64_t first,
                              uint64_t i,
                              Object **ret, uint64_t *offset) {
 
         Object *o;
-        uint64_t p = 0, a;
+        uint64_t p = 0, a, t = 0;
         int r;
+        ChainCacheItem *ci;
 
         assert(f);
 
         a = first;
+
+        /* Try the chain cache first */
+        ci = hashmap_get(f->chain_cache, &first);
+        if (ci && i > ci->total) {
+                a = ci->array;
+                i -= ci->total;
+                t = ci->total;
+        }
+
         while (a > 0) {
-                uint64_t n;
+                uint64_t k;
 
                 r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o);
                 if (r < 0)
                         return r;
 
-                n = journal_file_entry_array_n_items(o);
-                if (i < n) {
+                k = journal_file_entry_array_n_items(o);
+                if (i < k) {
                         p = le64toh(o->entry_array.items[i]);
-                        break;
+                        goto found;
                 }
 
-                i -= n;
+                i -= k;
+                t += k;
                 a = le64toh(o->entry_array.next_entry_array_offset);
         }
 
-        if (a <= 0 || p <= 0)
-                return 0;
+        return 0;
+
+found:
+        /* Let's cache this item for the next invocation */
+        chain_cache_put(f->chain_cache, ci, first, a, o->entry_array.items[0], t);
 
         r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o);
         if (r < 0)
@@ -1401,11 +1458,38 @@ static int generic_array_bisect(JournalFile *f,
         bool subtract_one = false;
         Object *o, *array = NULL;
         int r;
+        ChainCacheItem *ci;
 
         assert(f);
         assert(test_object);
 
+        /* Start with the first array in the chain */
         a = first;
+
+        ci = hashmap_get(f->chain_cache, &first);
+        if (ci && n > ci->total) {
+                /* Ah, we have iterated this bisection array chain
+                 * previously! Let's see if we can skip ahead in the
+                 * chain, as far as the last time. But we can't jump
+                 * backwards in the chain, so let's check that
+                 * first. */
+
+                r = test_object(f, ci->begin, needle);
+                if (r < 0)
+                        return r;
+
+                if (r == TEST_LEFT) {
+                        /* OK, what we are looking for is right of th
+                         * begin of this EntryArray, so let's jump
+                         * straight to previously cached array in the
+                         * chain */
+
+                        a = ci->array;
+                        n -= ci->total;
+                        t = ci->total;
+                }
+        }
+
         while (a > 0) {
                 uint64_t left, right, k, lp;
 
@@ -1486,6 +1570,9 @@ found:
         if (subtract_one && t == 0 && i == 0)
                 return 0;
 
+        /* Let's cache this item for the next invocation */
+        chain_cache_put(f->chain_cache, ci, first, a, array->entry_array.items[0], t);
+
         if (subtract_one && i == 0)
                 p = last_p;
         else if (subtract_one)
@@ -2265,6 +2352,12 @@ int journal_file_open(
                 goto fail;
         }
 
+        f->chain_cache = hashmap_new(uint64_hash_func, uint64_compare_func);
+        if (!f->chain_cache) {
+                r = -ENOMEM;
+                goto fail;
+        }
+
         f->fd = open(f->path, f->flags|O_CLOEXEC, f->mode);
         if (f->fd < 0) {
                 r = -errno;
diff --git a/src/journal/journal-file.h b/src/journal/journal-file.h
index d87cbe4876..cdbc8e41f6 100644
--- a/src/journal/journal-file.h
+++ b/src/journal/journal-file.h
@@ -33,6 +33,7 @@
 #include "journal-def.h"
 #include "util.h"
 #include "mmap-cache.h"
+#include "hashmap.h"
 
 typedef struct JournalMetrics {
         uint64_t max_use;
@@ -64,6 +65,8 @@ typedef struct JournalFile {
         JournalMetrics metrics;
         MMapCache *mmap;
 
+        Hashmap *chain_cache;
+
 #ifdef HAVE_XZ
         void *compress_buffer;
         uint64_t compress_buffer_size;
diff --git a/src/journal/test-journal-enum.c b/src/journal/test-journal-enum.c
new file mode 100644
index 0000000000..8a843ecdda
--- /dev/null
+++ b/src/journal/test-journal-enum.c
@@ -0,0 +1,53 @@
+/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
+
+/***
+  This file is part of systemd.
+
+  Copyright 2012 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU Lesser General Public License as published by
+  the Free Software Foundation; either version 2.1 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
+  Lesser General Public License for more details.
+
+  You should have received a copy of the GNU Lesser General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+#include <stdio.h>
+
+#include "log.h"
+#include "sd-journal.h"
+
+int main(int argc, char *argv[]) {
+        unsigned n = 0;
+        sd_journal *j;
+
+        log_set_max_level(LOG_DEBUG);
+
+        assert_se(sd_journal_open(&j, SD_JOURNAL_LOCAL_ONLY) >= 0);
+
+        assert_se(sd_journal_add_match(j, "_TRANSPORT=syslog", 0) >= 0);
+        assert_se(sd_journal_add_match(j, "_UID=0", 0) >= 0);
+
+        SD_JOURNAL_FOREACH_BACKWARDS(j) {
+                const void *d;
+                size_t l;
+
+                assert_se(sd_journal_get_data(j, "MESSAGE", &d, &l) >= 0);
+
+                printf("%.*s\n", (int) l, (char*) d);
+
+                n ++;
+                if (n >= 10)
+                        break;
+        }
+
+        sd_journal_close(j);
+        return 0;
+}