tdb2: make sure records with extra padding have a 0 byte.

As detailed in doc/design.lyx section 2.16 "Record Headers Are Not
Expandible", we make sure that if there is padding at the end of a record,
the first byte of padding is a zero.

diff --git a/ccan/tdb2/check.c b/ccan/tdb2/check.c
index 9bbd12645f199ba6a5a62708caccc79d08cf16d1..b602d80d64777030aeb08fab853b1df9561561d2 100644 (file)
--- a/ccan/tdb2/check.c
+++ b/ccan/tdb2/check.c
@@ -30,7 +30,8 @@ static bool append(tdb_off_t **arr, size_t *num, tdb_off_t off)
        return true;
 }
 
-static enum TDB_ERROR check_header(struct tdb_context *tdb, tdb_off_t *recovery)
+static enum TDB_ERROR check_header(struct tdb_context *tdb, tdb_off_t *recovery,
+                                  uint64_t *features)
 {
        uint64_t hash_test;
        struct tdb_header hdr;
@@ -59,6 +60,16 @@ static enum TDB_ERROR check_header(struct tdb_context *tdb, tdb_off_t *recovery)
                                  hdr.magic_food);
        }
 
+       /* Features which are used must be a subset of features offered. */
+       if (hdr.features_used & ~hdr.features_offered) {
+               return tdb_logerr(tdb, TDB_ERR_CORRUPT, TDB_LOG_ERROR,
+                                 "check: features used (0x%llx) which"
+                                 " are not offered (0x%llx)",
+                                 (long long)hdr.features_used,
+                                 (long long)hdr.features_offered);
+       }
+
+       *features = hdr.features_offered;
        *recovery = hdr.recovery;
        if (*recovery) {
                if (*recovery < sizeof(hdr) || *recovery > tdb->map_size) {
@@ -568,7 +579,7 @@ tdb_off_t dead_space(struct tdb_context *tdb, tdb_off_t off)
 static enum TDB_ERROR check_linear(struct tdb_context *tdb,
                                   tdb_off_t **used, size_t *num_used,
                                   tdb_off_t **fr, size_t *num_free,
-                                  tdb_off_t recovery)
+                                  uint64_t features, tdb_off_t recovery)
 {
        tdb_off_t off;
        tdb_len_t len;
@@ -697,6 +708,28 @@ static enum TDB_ERROR check_linear(struct tdb_context *tdb,
                                                  (long long)len,
                                                  (long long)off);
                        }
+
+                       /* Check that records have correct 0 at end (but may
+                        * not in future). */
+                       if (extra && !features) {
+                               const char *p;
+                               char c;
+                               p = tdb_access_read(tdb, off + sizeof(rec.u)
+                                                   + klen + dlen, 1, false);
+                               if (TDB_PTR_IS_ERR(p))
+                                       return TDB_PTR_ERR(p);
+                               c = *p;
+                               tdb_access_release(tdb, p);
+
+                               if (c != '\0') {
+                                       return tdb_logerr(tdb, TDB_ERR_CORRUPT,
+                                                         TDB_LOG_ERROR,
+                                                         "tdb_check:"
+                                                         " non-zero extra"
+                                                         " at %llu",
+                                                         (long long)off);
+                               }
+                       }
                } else {
                        return tdb_logerr(tdb, TDB_ERR_CORRUPT,
                                          TDB_LOG_ERROR,
@@ -724,6 +757,7 @@ enum TDB_ERROR tdb_check_(struct tdb_context *tdb,
 {
        tdb_off_t *fr = NULL, *used = NULL, ft, recovery;
        size_t num_free = 0, num_used = 0, num_found = 0, num_ftables = 0;
+       uint64_t features;
        enum TDB_ERROR ecode;
 
        ecode = tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, false);
@@ -737,12 +771,13 @@ enum TDB_ERROR tdb_check_(struct tdb_context *tdb,
                return ecode;
        }
 
-       ecode = check_header(tdb, &recovery);
+       ecode = check_header(tdb, &recovery, &features);
        if (ecode != TDB_SUCCESS)
                goto out;
 
        /* First we do a linear scan, checking all records. */
-       ecode = check_linear(tdb, &used, &num_used, &fr, &num_free, recovery);
+       ecode = check_linear(tdb, &used, &num_used, &fr, &num_free, features,
+                            recovery);
        if (ecode != TDB_SUCCESS)
                goto out;
 

diff --git a/ccan/tdb2/doc/design.lyx b/ccan/tdb2/doc/design.lyx
index af94f6b9c1b1a84ae49d4997bbd6affbea964a85..dfbc19fa955127d4cdba1e5fcd97439faa443410 100644 (file)
--- a/ccan/tdb2/doc/design.lyx
+++ b/ccan/tdb2/doc/design.lyx
@@ -1434,7 +1434,13 @@ Status
 \end_layout
 
 \begin_layout Standard
+
+\change_deleted 0 1300360766
 Incomplete.
+\change_inserted 0 1300360767
+Complete.
+\change_unchanged
+
 \end_layout
 
 \begin_layout Subsection

diff --git a/ccan/tdb2/free.c b/ccan/tdb2/free.c
index d2285567eb0efdf56dfd8d96ec0650845ef403e0..e853d97eedce65dc8ae4b52313ac5a12ca2ba80d 100644 (file)
--- a/ccan/tdb2/free.c
+++ b/ccan/tdb2/free.c
@@ -523,6 +523,15 @@ again:
                        goto unlock_err;
                }
 
+               /* For futureproofing, we put a 0 in any unused space. */
+               if (rec_extra_padding(&rec)) {
+                       ecode = tdb->methods->twrite(tdb, best_off + sizeof(rec)
+                                                    + keylen + datalen, "", 1);
+                       if (ecode != TDB_SUCCESS) {
+                               goto unlock_err;
+                       }
+               }
+
                /* Bucket of leftover will be <= current bucket, so nested
                 * locking is allowed. */
                if (leftover) {

diff --git a/ccan/tdb2/tdb.c b/ccan/tdb2/tdb.c
index 4ffcebf56150bffe381018b4f401aa5503455196..42e75f9d1840b8f79875f01ac05b6637a6d9f358 100644 (file)
--- a/ccan/tdb2/tdb.c
+++ b/ccan/tdb2/tdb.c
@@ -510,6 +510,21 @@ static enum TDB_ERROR replace_data(struct tdb_context *tdb,
        return TDB_SUCCESS;
 }
 
+static enum TDB_ERROR update_data(struct tdb_context *tdb,
+                                 tdb_off_t off,
+                                 struct tdb_data dbuf,
+                                 tdb_len_t extra)
+{
+       enum TDB_ERROR ecode;
+
+       ecode = tdb->methods->twrite(tdb, off, dbuf.dptr, dbuf.dsize);
+       if (ecode == TDB_SUCCESS && extra) {
+               /* Put a zero in; future versions may append other data. */
+               ecode = tdb->methods->twrite(tdb, off + dbuf.dsize, "", 1);
+       }
+       return ecode;
+}
+
 enum TDB_ERROR tdb_store(struct tdb_context *tdb,
                         struct tdb_data key, struct tdb_data dbuf, int flag)
 {
@@ -542,11 +557,10 @@ enum TDB_ERROR tdb_store(struct tdb_context *tdb,
                                if (ecode != TDB_SUCCESS) {
                                        goto out;
                                }
-                               ecode = tdb->methods->twrite(tdb,
-                                                            off + sizeof(rec)
-                                                            + key.dsize,
-                                                            dbuf.dptr,
-                                                            dbuf.dsize);
+                               ecode = update_data(tdb,
+                                                   off + sizeof(rec)
+                                                   + key.dsize, dbuf,
+                                                   old_room - dbuf.dsize);
                                if (ecode != TDB_SUCCESS) {
                                        goto out;
                                }
@@ -602,8 +616,8 @@ enum TDB_ERROR tdb_append(struct tdb_context *tdb,
                        }
 
                        off += sizeof(rec) + key.dsize + old_dlen;
-                       ecode = tdb->methods->twrite(tdb, off, dbuf.dptr,
-                                                    dbuf.dsize);
+                       ecode = update_data(tdb, off, dbuf,
+                                           rec_extra_padding(&rec));
                        goto out;
                }
 

diff --git a/ccan/tdb2/test/layout.c b/ccan/tdb2/test/layout.c
index 16f077198dbea011a2e9510a9b6dc1aa1b938260..77ad1f8e12cc7dd27f2b69178e71226a003897a5 100644 (file)
--- a/ccan/tdb2/test/layout.c
+++ b/ccan/tdb2/test/layout.c
@@ -92,6 +92,12 @@ static void set_free_record(void *mem, tdb_len_t len)
        /* We do all the work in add_to_freetable */
 }
 
+static void add_zero_pad(struct tdb_used_record *u, size_t len, size_t extra)
+{
+       if (extra)
+               ((char *)(u + 1))[len] = '\0';
+}
+
 static void set_data_record(void *mem, struct tdb_context *tdb,
                            struct tle_used *used)
 {
@@ -103,6 +109,7 @@ static void set_data_record(void *mem, struct tdb_context *tdb,
        memcpy(u + 1, used->key.dptr, used->key.dsize);
        memcpy((char *)(u + 1) + used->key.dsize,
               used->data.dptr, used->data.dsize);
+       add_zero_pad(u, used->key.dsize + used->data.dsize, used->extra);
 }
 
 static void set_hashtable(void *mem, struct tdb_context *tdb,
@@ -113,6 +120,7 @@ static void set_hashtable(void *mem, struct tdb_context *tdb,
 
        set_header(tdb, u, TDB_HTABLE_MAGIC, 0, len, len + htable->extra, 0);
        memset(u + 1, 0, len);
+       add_zero_pad(u, len, htable->extra);
 }
 
 static void set_freetable(void *mem, struct tdb_context *tdb,