Merge remote-tracking branch 'origin/mdb.master' into OPENLDAP_REL_ENG_2_4

libraries/liblmdb/lmdb.h

@@ -194,7 +194,7 @@ typedef int mdb_filehandle_t;
 	MDB_VERINT(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH)
 
 /** The release date of this library version */
-#define MDB_VERSION_DATE	"July  8, 2014"
+#define MDB_VERSION_DATE	"July 24, 2014"
 
 /** A stringifier for the version info */
 #define MDB_VERSTR(a,b,c,d)	"LMDB " #a "." #b "." #c ": (" d ")"
@@ -411,7 +411,7 @@ typedef enum MDB_cursor_op {
 #define MDB_CURSOR_FULL	(-30787)
 	/** Page has not enough space - internal error */
 #define MDB_PAGE_FULL	(-30786)
-	/** Environment mapsize was changed by another process */
+	/** Database contents grew beyond environment mapsize */
 #define MDB_MAP_RESIZED	(-30785)
 	/** MDB_INCOMPATIBLE: Operation and DB incompatible, or DB flags changed */
 #define MDB_INCOMPATIBLE	(-30784)
@@ -785,12 +785,15 @@ int  mdb_env_get_fd(MDB_env *env, mdb_filehandle_t *fd);
 	 * This function should be called after #mdb_env_create() and before #mdb_env_open().
 	 * It may be called at later times if no transactions are active in
 	 * this process. Note that the library does not check for this condition,
-	 * the caller must ensure it explicitly. The new size takes effect
-	 * immediately for the current process but will not be persisted to
-	 * any others until a write transaction has been committed by the
-	 * current process.
+	 * the caller must ensure it explicitly.
 	 *
-	 * If the mapsize is changed by another process, #mdb_txn_begin() will
+	 * The new size takes effect immediately for the current process but
+	 * will not be persisted to any others until a write transaction has been
+	 * committed by the current process. Also, only mapsize increases are
+	 * persisted into the environment.
+	 *
+	 * If the mapsize is increased by another process, and data has grown
+	 * beyond the range of the current mapsize, #mdb_txn_begin() will
 	 * return #MDB_MAP_RESIZED. This function may be called with a size
 	 * of zero to adopt the new size.
 	 *

libraries/liblmdb/mdb.c

@@ -1074,8 +1074,6 @@ struct MDB_env {
 	HANDLE		me_mfd;			/**< just for writing the meta pages */
 	/** Failed to update the meta page. Probably an I/O error. */
 #define	MDB_FATAL_ERROR	0x80000000U
-	/** We're explicitly changing the mapsize. */
-#define	MDB_RESIZING	0x40000000U
 	/** Some fields are initialized. */
 #define	MDB_ENV_ACTIVE	0x20000000U
 	/** me_txkey is set */
@@ -2537,9 +2535,7 @@ mdb_txn_renew0(MDB_txn *txn)
 	}
 	txn->mt_dbflags[0] = txn->mt_dbflags[1] = DB_VALID;
 
-	/* If we didn't ask for a resize, but the size grew, fail */
-	if (!(env->me_flags & MDB_RESIZING)
-		&& env->me_mapsize < meta->mm_mapsize) {
+	if (env->me_maxpg < txn->mt_next_pgno) {
 		mdb_txn_reset0(txn, "renew0-mapfail");
 		if (new_notls) {
 			txn->mt_u.reader->mr_pid = 0;
@@ -3277,13 +3273,8 @@ mdb_txn_commit(MDB_txn *txn)
 	mdb_cursors_close(txn, 0);
 
 	if (!txn->mt_u.dirty_list[0].mid &&
-		!(txn->mt_flags & (MDB_TXN_DIRTY|MDB_TXN_SPILLS))) {
-		if ((env->me_flags & MDB_RESIZING)
-			&& (rc = mdb_env_write_meta(txn))) {
-			goto fail;
-		}
+		!(txn->mt_flags & (MDB_TXN_DIRTY|MDB_TXN_SPILLS)))
 		goto done;
-	}
 
 	DPRINTF(("committing txn %"Z"u %p on mdbenv %p, root page %"Z"u",
 	    txn->mt_txnid, (void*)txn, (void*)env, txn->mt_dbs[MAIN_DBI].md_root));
@@ -3324,6 +3315,7 @@ mdb_txn_commit(MDB_txn *txn)
 #endif
 
 	if ((rc = mdb_page_flush(txn, 0)) ||
+		(rc = mdb_env_sync(env, 0)) ||
 		(rc = mdb_env_write_meta(txn)))
 		goto fail;
 
@@ -3480,8 +3472,9 @@ mdb_env_init_meta(MDB_env *env, MDB_meta *meta)
 static int
 mdb_env_write_meta(MDB_txn *txn)
 {
-	MDB_env *env = txn->mt_env;
+	MDB_env *env;
 	MDB_meta	meta, metab, *mp;
+	size_t mapsize;
 	off_t off;
 	int rc, len, toggle;
 	char *ptr;
@@ -3492,22 +3485,19 @@ mdb_env_write_meta(MDB_txn *txn)
 	int r2;
 #endif
 
-	/* Sync data and previous metapage before writing a new metapage */
-	if ((rc = mdb_env_sync(env, 0)) != MDB_SUCCESS)
-		return rc;
-
 	toggle = txn->mt_txnid & 1;
 	DPRINTF(("writing meta page %d for root page %"Z"u",
 		toggle, txn->mt_dbs[MAIN_DBI].md_root));
 
+	env = txn->mt_env;
 	mp = env->me_metas[toggle];
+	mapsize = env->me_metas[toggle ^ 1]->mm_mapsize;
+	/* Persist any increases of mapsize config */
+	if (mapsize < env->me_mapsize)
+		mapsize = env->me_mapsize;
 
 	if (env->me_flags & MDB_WRITEMAP) {
-		/* Persist any changes of mapsize config */
-		if (env->me_flags & MDB_RESIZING) {
-			mp->mm_mapsize = env->me_mapsize;
-			env->me_flags ^= MDB_RESIZING;
-		}
+		mp->mm_mapsize = mapsize;
 		mp->mm_dbs[0] = txn->mt_dbs[0];
 		mp->mm_dbs[1] = txn->mt_dbs[1];
 		mp->mm_last_pg = txn->mt_next_pgno - 1;
@@ -3534,23 +3524,15 @@ mdb_env_write_meta(MDB_txn *txn)
 	metab.mm_txnid = env->me_metas[toggle]->mm_txnid;
 	metab.mm_last_pg = env->me_metas[toggle]->mm_last_pg;
 
-	ptr = (char *)&meta;
-	if (env->me_flags & MDB_RESIZING) {
-		/* Persist any changes of mapsize config */
-		meta.mm_mapsize = env->me_mapsize;
-		off = offsetof(MDB_meta, mm_mapsize);
-		env->me_flags ^= MDB_RESIZING;
-	} else {
-		off = offsetof(MDB_meta, mm_dbs[0].md_depth);
-	}
-	len = sizeof(MDB_meta) - off;
-
-	ptr += off;
+	meta.mm_mapsize = mapsize;
 	meta.mm_dbs[0] = txn->mt_dbs[0];
 	meta.mm_dbs[1] = txn->mt_dbs[1];
 	meta.mm_last_pg = txn->mt_next_pgno - 1;
 	meta.mm_txnid = txn->mt_txnid;
 
+	off = offsetof(MDB_meta, mm_mapsize);
+	ptr = (char *)&meta + off;
+	len = sizeof(MDB_meta) - off;
 	if (toggle)
 		off += env->me_psize;
 	off += PAGEHDRSZ;
@@ -3728,25 +3710,19 @@ mdb_env_set_mapsize(MDB_env *env, size_t size)
 	 * sure there are no active txns.
 	 */
 	if (env->me_map) {
-		int rc, change = 0;
+		int rc;
 		void *old;
 		if (env->me_txn)
 			return EINVAL;
 		if (!size)
 			size = env->me_metas[mdb_env_pick_meta(env)]->mm_mapsize;
-		else {
-			if (size < env->me_mapsize) {
-				/* If the configured size is smaller, make sure it's
-				 * still big enough. Silently round up to minimum if not.
-				 */
-				size_t minsize = (env->me_metas[mdb_env_pick_meta(env)]->mm_last_pg + 1) * env->me_psize;
-				if (size < minsize)
-					size = minsize;
-			}
-			/* nothing actually changed */
-			if (size == env->me_mapsize)
-				return MDB_SUCCESS;
-			change = 1;
+		else if (size < env->me_mapsize) {
+			/* If the configured size is smaller, make sure it's
+			 * still big enough. Silently round up to minimum if not.
+			 */
+			size_t minsize = (env->me_metas[mdb_env_pick_meta(env)]->mm_last_pg + 1) * env->me_psize;
+			if (size < minsize)
+				size = minsize;
 		}
 		munmap(env->me_map, env->me_mapsize);
 		env->me_mapsize = size;
@@ -3754,8 +3730,6 @@ mdb_env_set_mapsize(MDB_env *env, size_t size)
 		rc = mdb_env_map(env, old);
 		if (rc)
 			return rc;
-		if (change)
-			env->me_flags |= MDB_RESIZING;
 	}
 	env->me_mapsize = size;
 	if (env->me_psize)
@@ -3828,17 +3802,13 @@ mdb_env_open2(MDB_env *env)
 		 * else use the size recorded in the existing env.
 		 */
 		env->me_mapsize = newenv ? DEFAULT_MAPSIZE : meta.mm_mapsize;
-	} else {
-		if (env->me_mapsize < meta.mm_mapsize) {
-			/* If the configured size is smaller, make sure it's
-			 * still big enough. Silently round up to minimum if not.
-			 */
-			size_t minsize = (meta.mm_last_pg + 1) * meta.mm_psize;
-			if (env->me_mapsize < minsize)
-				env->me_mapsize = minsize;
-		}
-		if (env->me_mapsize != meta.mm_mapsize)
-			env->me_flags |= MDB_RESIZING;
+	} else if (env->me_mapsize < meta.mm_mapsize) {
+		/* If the configured size is smaller, make sure it's
+		 * still big enough. Silently round up to minimum if not.
+		 */
+		size_t minsize = (meta.mm_last_pg + 1) * meta.mm_psize;
+		if (env->me_mapsize < minsize)
+			env->me_mapsize = minsize;
 	}
 
 	rc = mdb_env_map(env, (flags & MDB_FIXEDMAP) ? meta.mm_address : NULL);
@@ -6992,6 +6962,9 @@ mdb_cursor_count(MDB_cursor *mc, size_t *countp)
 	if (!(mc->mc_flags & C_INITIALIZED))
 		return EINVAL;
 
+	if (!mc->mc_snum || (mc->mc_flags & C_EOF))
+		return MDB_NOTFOUND;
+
 	leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]);
 	if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) {
 		*countp = 1;
@@ -8440,13 +8413,13 @@ mdb_env_copyfd1(MDB_env *env, HANDLE fd)
 #ifdef _WIN32
 	my.mc_mutex = CreateMutex(NULL, FALSE, NULL);
 	my.mc_cond = CreateEvent(NULL, FALSE, FALSE, NULL);
-	my.mc_wbuf[0] = _aligned_malloc(MDB_WBUF*2, env->me_psize);
+	my.mc_wbuf[0] = _aligned_malloc(MDB_WBUF*2, env->me_os_psize);
 	if (my.mc_wbuf[0] == NULL)
 		return errno;
 #else
 	pthread_mutex_init(&my.mc_mutex, NULL);
 	pthread_cond_init(&my.mc_cond, NULL);
-	rc = posix_memalign((void **)&my.mc_wbuf[0], env->me_psize, MDB_WBUF*2);
+	rc = posix_memalign((void **)&my.mc_wbuf[0], env->me_os_psize, MDB_WBUF*2);
 	if (rc)
 		return rc;
 #endif
@@ -8684,6 +8657,7 @@ mdb_env_copy2(MDB_env *env, const char *path, unsigned int flags)
 		goto leave;
 	}
 
+	if (env->me_psize >= env->me_os_psize) {
 #ifdef O_DIRECT
 	/* Set O_DIRECT if the file system supports it */
 	if ((rc = fcntl(newfd, F_GETFL)) != -1)
@@ -8696,6 +8670,7 @@ mdb_env_copy2(MDB_env *env, const char *path, unsigned int flags)
 		goto leave;
 	}
 #endif
+	}
 
 	rc = mdb_env_copyfd2(env, newfd, flags);