]> git.ozlabs.org Git - ccan/blobdiff - ccan/tdb2/tdb1_summary.c
tdb2: import TDB1 code.
[ccan] / ccan / tdb2 / tdb1_summary.c
diff --git a/ccan/tdb2/tdb1_summary.c b/ccan/tdb2/tdb1_summary.c
new file mode 100644 (file)
index 0000000..61e1853
--- /dev/null
@@ -0,0 +1,201 @@
+ /*
+   Trivial Database: human-readable summary code
+   Copyright (C) Rusty Russell 2010
+
+   This library 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 3 of the License, or (at your option) any later version.
+
+   This library 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 this library; if not, see <http://www.gnu.org/licenses/>.
+*/
+#include "tdb1_private.h"
+
+#define SUMMARY_FORMAT1 \
+       "Size of file/data: %u/%zu\n" \
+       "Number of records: %zu\n" \
+       "Smallest/average/largest keys: %zu/%zu/%zu\n" \
+       "Smallest/average/largest data: %zu/%zu/%zu\n" \
+       "Smallest/average/largest padding: %zu/%zu/%zu\n" \
+       "Number of dead records: %zu\n" \
+       "Smallest/average/largest dead records: %zu/%zu/%zu\n" \
+       "Number of free records: %zu\n" \
+       "Smallest/average/largest free records: %zu/%zu/%zu\n" \
+       "Number of hash chains: %zu\n" \
+       "Smallest/average/largest hash chains: %zu/%zu/%zu\n" \
+       "Number of uncoalesced records: %zu\n" \
+       "Smallest/average/largest uncoalesced runs: %zu/%zu/%zu\n" \
+       "Percentage keys/data/padding/free/dead/rechdrs&tailers/hashes: %.0f/%.0f/%.0f/%.0f/%.0f/%.0f/%.0f\n"
+
+/* We don't use tally module, to keep upstream happy. */
+struct tally {
+       size_t min, max, total;
+       size_t num;
+};
+
+static void tally1_init(struct tally *tally)
+{
+       tally->total = 0;
+       tally->num = 0;
+       tally->min = tally->max = 0;
+}
+
+static void tally1_add(struct tally *tally, size_t len)
+{
+       if (tally->num == 0)
+               tally->max = tally->min = len;
+       else if (len > tally->max)
+               tally->max = len;
+       else if (len < tally->min)
+               tally->min = len;
+       tally->num++;
+       tally->total += len;
+}
+
+static size_t tally1_mean(const struct tally *tally)
+{
+       if (!tally->num)
+               return 0;
+       return tally->total / tally->num;
+}
+
+static size_t get_hash_length(struct tdb1_context *tdb, unsigned int i)
+{
+       tdb1_off_t rec_ptr;
+       size_t count = 0;
+
+       if (tdb1_ofs_read(tdb, TDB1_HASH_TOP(i), &rec_ptr) == -1)
+               return 0;
+
+       /* keep looking until we find the right record */
+       while (rec_ptr) {
+               struct tdb1_record r;
+               ++count;
+               if (tdb1_rec_read(tdb, rec_ptr, &r) == -1)
+                       return 0;
+               rec_ptr = r.next;
+       }
+       return count;
+}
+
+_PUBLIC_ char *tdb1_summary(struct tdb1_context *tdb)
+{
+       tdb1_off_t off, rec_off;
+       struct tally freet, keys, data, dead, extra, hash, uncoal;
+       struct tdb1_record rec;
+       char *ret = NULL;
+       bool locked;
+       size_t len, unc = 0;
+       struct tdb1_record recovery;
+
+       /* Read-only databases use no locking at all: it's best-effort.
+        * We may have a write lock already, so skip that case too. */
+       if (tdb->read_only || tdb->allrecord_lock.count != 0) {
+               locked = false;
+       } else {
+               if (tdb1_lockall_read(tdb) == -1)
+                       return NULL;
+               locked = true;
+       }
+
+       if (tdb1_recovery_area(tdb, tdb->methods, &rec_off, &recovery) != 0) {
+               goto unlock;
+       }
+
+       tally1_init(&freet);
+       tally1_init(&keys);
+       tally1_init(&data);
+       tally1_init(&dead);
+       tally1_init(&extra);
+       tally1_init(&hash);
+       tally1_init(&uncoal);
+
+       for (off = TDB1_DATA_START(tdb->header.hash_size);
+            off < tdb->map_size - 1;
+            off += sizeof(rec) + rec.rec_len) {
+               if (tdb->methods->tdb1_read(tdb, off, &rec, sizeof(rec),
+                                          TDB1_DOCONV()) == -1)
+                       goto unlock;
+               switch (rec.magic) {
+               case TDB1_MAGIC:
+                       tally1_add(&keys, rec.key_len);
+                       tally1_add(&data, rec.data_len);
+                       tally1_add(&extra, rec.rec_len - (rec.key_len
+                                                        + rec.data_len));
+                       if (unc > 1)
+                               tally1_add(&uncoal, unc - 1);
+                       unc = 0;
+                       break;
+               case TDB1_FREE_MAGIC:
+                       tally1_add(&freet, rec.rec_len);
+                       unc++;
+                       break;
+               /* If we crash after ftruncate, we can get zeroes or fill. */
+               case TDB1_RECOVERY_INVALID_MAGIC:
+               case 0x42424242:
+                       unc++;
+                       /* If it's a valid recovery, we can trust rec_len. */
+                       if (off != rec_off) {
+                               rec.rec_len = tdb1_dead_space(tdb, off)
+                                       - sizeof(rec);
+                       }
+                       /* Fall through */
+               case TDB1_DEAD_MAGIC:
+                       tally1_add(&dead, rec.rec_len);
+                       break;
+               default:
+                       TDB1_LOG((tdb, TDB1_DEBUG_ERROR,
+                                "Unexpected record magic 0x%x at offset %d\n",
+                                rec.magic, off));
+                       goto unlock;
+               }
+       }
+       if (unc > 1)
+               tally1_add(&uncoal, unc - 1);
+
+       for (off = 0; off < tdb->header.hash_size; off++)
+               tally1_add(&hash, get_hash_length(tdb, off));
+
+       /* 20 is max length of a %zu. */
+       len = strlen(SUMMARY_FORMAT1) + 35*20 + 1;
+       ret = (char *)malloc(len);
+       if (!ret)
+               goto unlock;
+
+       snprintf(ret, len, SUMMARY_FORMAT1,
+                tdb->map_size, keys.total+data.total,
+                keys.num,
+                keys.min, tally1_mean(&keys), keys.max,
+                data.min, tally1_mean(&data), data.max,
+                extra.min, tally1_mean(&extra), extra.max,
+                dead.num,
+                dead.min, tally1_mean(&dead), dead.max,
+                freet.num,
+                freet.min, tally1_mean(&freet), freet.max,
+                hash.num,
+                hash.min, tally1_mean(&hash), hash.max,
+                uncoal.total,
+                uncoal.min, tally1_mean(&uncoal), uncoal.max,
+                keys.total * 100.0 / tdb->map_size,
+                data.total * 100.0 / tdb->map_size,
+                extra.total * 100.0 / tdb->map_size,
+                freet.total * 100.0 / tdb->map_size,
+                dead.total * 100.0 / tdb->map_size,
+                (keys.num + freet.num + dead.num)
+                * (sizeof(struct tdb1_record) + sizeof(uint32_t))
+                * 100.0 / tdb->map_size,
+                tdb->header.hash_size * sizeof(tdb1_off_t)
+                * 100.0 / tdb->map_size);
+
+unlock:
+       if (locked) {
+               tdb1_unlockall_read(tdb);
+       }
+       return ret;
+}