Import 898b5edfe757cb145960b8f3631029bfd5592119 from ctdb:

[ccan] / ccan / tdb / transaction.c

diff --git a/ccan/tdb/transaction.c b/ccan/tdb/transaction.c
index 4a49cc7b6421848a0eac1efbff2cb9e4913bb367..c5580f205d39c074fbbe52f9f5545659638112c3 100644 (file)
--- a/ccan/tdb/transaction.c
+++ b/ccan/tdb/transaction.c
@@ -86,12 +86,20 @@
     fsync/msync calls are made.
 
   - if TDB_ALLOW_NESTING is passed to flags in tdb open, or added using
-    tdb_add_flags() transaction is enabled.
-    The default is that transaction nesting is not allowed and an attempt
-    to create a nested transaction will fail with TDB_ERR_NESTING.
+    tdb_add_flags() transaction nesting is enabled.
+    It resets the TDB_DISALLOW_NESTING flag, as both cannot be used together.
+    The default is that transaction nesting is allowed.
+    Note: this default may change in future versions of tdb.
 
     Beware. when transactions are nested a transaction successfully
     completed with tdb_transaction_commit() can be silently unrolled later.
+
+  - if TDB_DISALLOW_NESTING is passed to flags in tdb open, or added using
+    tdb_add_flags() transaction nesting is disabled.
+    It resets the TDB_ALLOW_NESTING flag, as both cannot be used together.
+    An attempt create a nested transaction will fail with TDB_ERR_NESTING.
+    The default is that transaction nesting is allowed.
+    Note: this default may change in future versions of tdb.
 */
 
 
@@ -127,6 +135,9 @@ struct tdb_transaction {
        bool prepared;
        tdb_off_t magic_offset;
 
+       /* set when the GLOBAL_LOCK has been taken */
+       bool global_lock_taken;
+
        /* old file size before transaction */
        tdb_len_t old_map_size;
 
@@ -493,6 +504,11 @@ int _tdb_transaction_cancel(struct tdb_context *tdb, int ltype)
                }
        }
 
+       if (tdb->transaction->global_lock_taken) {
+               tdb_brunlock(tdb, F_WRLCK, GLOBAL_LOCK, 1);
+               tdb->transaction->global_lock_taken = false;
+       }
+
        /* remove any global lock created during the transaction */
        if (tdb->global_lock.count != 0) {
                tdb_brunlock(tdb, tdb->global_lock.ltype,
@@ -680,7 +696,7 @@ static int tdb_recovery_allocate(struct tdb_context *tdb,
                                 tdb_off_t *recovery_offset,
                                 tdb_len_t *recovery_max_size)
 {
-       struct list_struct rec;
+       struct tdb_record rec;
        const struct tdb_methods *methods = tdb->transaction->io_methods;
        tdb_off_t recovery_head;
 
@@ -766,7 +782,7 @@ static int transaction_setup_recovery(struct tdb_context *tdb,
        tdb_len_t recovery_size;
        unsigned char *data, *p;
        const struct tdb_methods *methods = tdb->transaction->io_methods;
-       struct list_struct *rec;
+       struct tdb_record *rec;
        tdb_off_t recovery_offset, recovery_max_size;
        tdb_off_t old_map_size = tdb->transaction->old_map_size;
        uint32_t magic, tailer;
@@ -786,7 +802,7 @@ static int transaction_setup_recovery(struct tdb_context *tdb,
                return -1;
        }
 
-       rec = (struct list_struct *)data;
+       rec = (struct tdb_record *)data;
        memset(rec, 0, sizeof(*rec));
 
        rec->magic    = 0;
@@ -869,7 +885,7 @@ static int transaction_setup_recovery(struct tdb_context *tdb,
        magic = TDB_RECOVERY_MAGIC;
        CONVERT(magic);
 
-       *magic_offset = recovery_offset + offsetof(struct list_struct, magic);
+       *magic_offset = recovery_offset + offsetof(struct tdb_record, magic);
 
        if (methods->tdb_write(tdb, *magic_offset, &magic, sizeof(magic)) == -1) {
                TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_setup_recovery: failed to write recovery magic\n"));
@@ -919,11 +935,6 @@ static int _tdb_transaction_prepare_commit(struct tdb_context *tdb)
                return 0;
        }               
 
-#ifdef TDB_TRACE
-       /* store seqnum now, before reading becomes illegal. */
-       tdb_ofs_read(tdb, TDB_SEQNUM_OFS, &tdb->transaction_prepare_seqnum);
-#endif
-
        /* check for a null transaction */
        if (tdb->transaction->blocks == NULL) {
                return 0;
@@ -957,11 +968,12 @@ static int _tdb_transaction_prepare_commit(struct tdb_context *tdb)
                return -1;
        }
 
+       tdb->transaction->global_lock_taken = true;
+
        if (!(tdb->flags & TDB_NOSYNC)) {
                /* write the recovery data to the end of the file */
                if (transaction_setup_recovery(tdb, &tdb->transaction->magic_offset) == -1) {
                        TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_prepare_commit: failed to setup recovery data\n"));
-                       tdb_brunlock(tdb, F_WRLCK, GLOBAL_LOCK, 1);
                        _tdb_transaction_cancel(tdb, F_WRLCK);
                        return -1;
                }
@@ -976,7 +988,6 @@ static int _tdb_transaction_prepare_commit(struct tdb_context *tdb)
                                             tdb->transaction->old_map_size) == -1) {
                        tdb->ecode = TDB_ERR_IO;
                        TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_prepare_commit: expansion failed\n"));
-                       tdb_brunlock(tdb, F_WRLCK, GLOBAL_LOCK, 1);
                        _tdb_transaction_cancel(tdb, F_WRLCK);
                        return -1;
                }
@@ -1012,13 +1023,7 @@ int tdb_transaction_commit(struct tdb_context *tdb)
                return -1;
        }
 
-       /* If we've prepared, can't read seqnum. */
-       if (tdb->transaction->prepared) {
-               tdb_trace_seqnum(tdb, tdb->transaction_prepare_seqnum,
-                                "tdb_transaction_commit");
-       } else {
-               tdb_trace(tdb, "tdb_transaction_commit");
-       }
+       tdb_trace(tdb, "tdb_transaction_commit");
 
        if (tdb->transaction->transaction_error) {
                tdb->ecode = TDB_ERR_IO;
@@ -1072,7 +1077,6 @@ int tdb_transaction_commit(struct tdb_context *tdb)
                        tdb_transaction_recover(tdb); 
 
                        _tdb_transaction_cancel(tdb, F_WRLCK);
-                       tdb_brunlock(tdb, F_WRLCK, GLOBAL_LOCK, 1);
 
                        TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_commit: write failed\n"));
                        return -1;
@@ -1088,8 +1092,6 @@ int tdb_transaction_commit(struct tdb_context *tdb)
                return -1;
        }
 
-       tdb_brunlock(tdb, F_WRLCK, GLOBAL_LOCK, 1);
-
        /*
          TODO: maybe write to some dummy hdr field, or write to magic
          offset without mmap, before the last sync, instead of the
@@ -1129,7 +1131,7 @@ int tdb_transaction_recover(struct tdb_context *tdb)
        tdb_off_t recovery_head, recovery_eof;
        unsigned char *data, *p;
        uint32_t zero = 0;
-       struct list_struct rec;
+       struct tdb_record rec;
 
        /* find the recovery area */
        if (tdb_ofs_read(tdb, TDB_RECOVERY_HEAD, &recovery_head) == -1) {
@@ -1216,7 +1218,7 @@ int tdb_transaction_recover(struct tdb_context *tdb)
        }
 
        /* remove the recovery magic */
-       if (tdb_ofs_write(tdb, recovery_head + offsetof(struct list_struct, magic), 
+       if (tdb_ofs_write(tdb, recovery_head + offsetof(struct tdb_record, magic),
                          &zero) == -1) {
                TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_transaction_recover: failed to remove recovery magic\n"));
                tdb->ecode = TDB_ERR_IO;