mdb_txn_renew0(): Fix un-mutexed me_flags update.

[openldap] / libraries / liblmdb / mdb.c

diff --git a/libraries/liblmdb/mdb.c b/libraries/liblmdb/mdb.c
index 7e2905891eb770c4d6eb53f751bd05b2738a0c7a..a3ac603f882f58cd01dad1807c35f7381c344b45 100644 (file)
--- a/libraries/liblmdb/mdb.c
+++ b/libraries/liblmdb/mdb.c
@@ -352,7 +352,7 @@ static txnid_t mdb_debug_start;
 
        /**     @brief The maximum size of a database page.
         *
-        *      This is 32k, since it must fit in #MDB_page.#mp_upper.
+        *      This is 32k, since it must fit in #MDB_page.%mp_upper.
         *
         *      LMDB will use database pages < OS pages if needed.
         *      That causes more I/O in write transactions: The OS must
@@ -941,7 +941,7 @@ struct MDB_txn {
 #define MDB_TXN_SPILLS         0x08            /**< txn or a parent has spilled pages */
 /** @} */
        unsigned int    mt_flags;               /**< @ref mdb_txn */
-       /** dirty_list room: Array size - #dirty pages visible to this txn.
+       /** #dirty_list room: Array size - \#dirty pages visible to this txn.
         *      Includes ancestor txns' dirty pages not hidden by other txns'
         *      dirty/spilled pages. Thus commit(nested txn) has room to merge
         *      dirty_list into mt_parent after freeing hidden mt_parent pages.
@@ -1034,8 +1034,6 @@ struct MDB_env {
 #define        MDB_ENV_ACTIVE  0x20000000U
        /** me_txkey is set */
 #define        MDB_ENV_TXKEY   0x10000000U
-       /** Have liveness lock in reader table */
-#define        MDB_LIVE_READER 0x08000000U
        uint32_t        me_flags;               /**< @ref mdb_env */
        unsigned int    me_psize;       /**< DB page size, inited from me_os_psize */
        unsigned int    me_os_psize;    /**< OS page size, from #GET_PAGESIZE */
@@ -1071,6 +1069,7 @@ struct MDB_env {
 #if !(MDB_MAXKEYSIZE)
        unsigned int    me_maxkey;      /**< max size of a key */
 #endif
+       int             me_live_reader;         /**< have liveness lock in reader table */
 #ifdef _WIN32
        int             me_pidquery;            /**< Used in OpenProcess */
        HANDLE          me_rmutex;              /* Windows mutexes don't reside in shared mem */
@@ -1096,7 +1095,7 @@ typedef struct MDB_ntxn {
 #define MDB_COMMIT_PAGES       IOV_MAX
 #endif
 
-       /* max bytes to write in one call */
+       /** max bytes to write in one call */
 #define MAX_WRITE              (0x80000000U >> (sizeof(ssize_t) == 4))
 
 static int  mdb_page_alloc(MDB_cursor *mc, int num, MDB_page **mp);
@@ -2268,11 +2267,11 @@ mdb_txn_renew0(MDB_txn *txn)
                                MDB_PID_T pid = env->me_pid;
                                pthread_t tid = pthread_self();
 
-                               if (!(env->me_flags & MDB_LIVE_READER)) {
+                               if (!env->me_live_reader) {
                                        rc = mdb_reader_pid(env, Pidset, pid);
                                        if (rc)
                                                return rc;
-                                       env->me_flags |= MDB_LIVE_READER;
+                                       env->me_live_reader = 1;
                                }
 
                                LOCK_MUTEX_R(env);
@@ -2728,10 +2727,11 @@ mdb_freelist_save(MDB_txn *txn)
                        MDB_ID save;
 
                        mdb_tassert(txn, len >= 0 && id <= env->me_pglast);
-                       key.mv_data = &id;
                        if (len > mop_len) {
                                len = mop_len;
                                data.mv_size = (len + 1) * sizeof(MDB_ID);
+                               /* Drop MDB_CURRENT when changing the data size */
+                               key.mv_data = &id;
                                flags = 0;
                        }
                        data.mv_data = mop -= len;
@@ -5761,7 +5761,7 @@ mdb_cursor_put(MDB_cursor *mc, MDB_val *key, MDB_val *data,
        unsigned int nflags;
        DKBUF;
 
-       if (mc == NULL || key == NULL)
+       if (mc == NULL)
                return EINVAL;
 
        env = mc->mc_txn->mt_env;
@@ -5782,8 +5782,16 @@ mdb_cursor_put(MDB_cursor *mc, MDB_val *key, MDB_val *data,
        if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_ERROR))
                return (mc->mc_txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN;
 
-       if (flags != MDB_CURRENT && key->mv_size-1 >= ENV_MAXKEY(env))
-               return MDB_BAD_VALSIZE;
+       if (flags != MDB_CURRENT) {
+               if (key == NULL)
+                       return EINVAL;
+               if (key->mv_size-1 >= ENV_MAXKEY(env))
+                       return MDB_BAD_VALSIZE;
+       } else {
+               /* Ignore key except in sub-cursor, where key holds the data */
+               if (!(mc->mc_flags & C_SUB))
+                       key = NULL;
+       }
 
 #if SIZE_MAX > MAXDATASIZE
        if (data->mv_size > ((mc->mc_db->md_flags & MDB_DUPSORT) ? ENV_MAXKEY(env) : MAXDATASIZE))
@@ -6107,7 +6115,7 @@ current:
                         */
                        if (F_ISSET(flags, MDB_RESERVE))
                                data->mv_data = olddata.mv_data;
-                       else if (data->mv_size)
+                       else if (!(mc->mc_flags & C_SUB))
                                memcpy(olddata.mv_data, data->mv_data, data->mv_size);
                        else
                                memcpy(NODEKEY(leaf), key->mv_data, key->mv_size);
@@ -6525,8 +6533,7 @@ full:
 }
 
 /** Delete the specified node from a page.
- * @param[in] mp The page to operate on.
- * @param[in] indx The index of the node to delete.
+ * @param[in] mc Cursor pointing to the node to delete.
  * @param[in] ksize The size of a node. Only used if the page is
  * part of a #MDB_DUPFIXED database.
  */
@@ -7100,6 +7107,7 @@ mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst)
  *     the \b csrc page will be freed.
  * @param[in] csrc Cursor pointing to the source page.
  * @param[in] cdst Cursor pointing to the destination page.
+ * @return 0 on success, non-zero on failure.
  */
 static int
 mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst)
@@ -7756,7 +7764,13 @@ mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno
                                mc->mc_ki[i] = mn.mc_ki[i];
                        }
                        mc->mc_pg[ptop] = mn.mc_pg[ptop];
-                       mc->mc_ki[ptop] = mn.mc_ki[ptop] - 1;
+                       if (mn.mc_ki[ptop]) {
+                               mc->mc_ki[ptop] = mn.mc_ki[ptop] - 1;
+                       } else {
+                               /* find right page's left sibling */
+                               mc->mc_ki[ptop] = mn.mc_ki[ptop];
+                               mdb_cursor_sibling(mc, 0);
+                       }
                }
        } else {
                mn.mc_top--;
@@ -7848,12 +7862,10 @@ mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno
                         */
                        if (mn.mc_pg[ptop] != mc->mc_pg[ptop] &&
                                mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) {
-                               for (i=0; i<ptop; i++) {
+                               for (i=0; i<=ptop; i++) {
                                        mc->mc_pg[i] = mn.mc_pg[i];
                                        mc->mc_ki[i] = mn.mc_ki[i];
                                }
-                               mc->mc_pg[ptop] = mn.mc_pg[ptop];
-                               mc->mc_ki[ptop] = mn.mc_ki[ptop] - 1;
                        }
                }
                /* return tmp page to freelist */