trx0trx.c

这是linux下运行的mysql软件包,可用于linux 下安装 php + mysql + apach 的网络配置

				trx->no = ut_dulint_max;
			}

			if (undo->dict_operation) {
				trx->dict_operation = undo->dict_operation;
				trx->table_id = undo->table_id;
			}

			if (!undo->empty) {
				trx->undo_no = ut_dulint_add(undo->top_undo_no,
									1);
			}

			trx_list_insert_ordered(trx);

			undo = UT_LIST_GET_NEXT(undo_list, undo);
		}

		undo = UT_LIST_GET_FIRST(rseg->update_undo_list);

		while (undo != NULL) {
			trx = trx_get_on_id(undo->trx_id);

			if (NULL == trx) {
				trx = trx_create(NULL); 

				trx->id = undo->trx_id;
				trx->xid = undo->xid;

				if (undo->state != TRX_UNDO_ACTIVE) {

					/* Prepared transactions are left in
					the prepared state waiting for a
					commit or abort decision from MySQL */

					if (undo->state == TRX_UNDO_PREPARED) {
 					    fprintf(stderr,
"InnoDB: Transaction %lu %lu was in the XA prepared state.\n",
					    ut_dulint_get_high(trx->id),
					    ut_dulint_get_low(trx->id));

					    if (srv_force_recovery == 0) {

						trx->conc_state = TRX_PREPARED;
					    } else {
 						fprintf(stderr,
"InnoDB: Since innodb_force_recovery > 0, we will rollback it anyway.\n");

						trx->conc_state = TRX_ACTIVE;
					    }
					} else {
						trx->conc_state =
						  TRX_COMMITTED_IN_MEMORY;
					}

					/* We give a dummy value for the trx
					number */

					trx->no = trx->id;
				} else {
					trx->conc_state = TRX_ACTIVE;

					/* A running transaction always has
					the number field inited to
					ut_dulint_max */

					trx->no = ut_dulint_max;
				}

				trx->rseg = rseg;
				trx_list_insert_ordered(trx);

				if (undo->dict_operation) {
					trx->dict_operation =
							undo->dict_operation;
					trx->table_id = undo->table_id;
				}
			}

			trx->update_undo = undo;

			if ((!undo->empty)
			    && (ut_dulint_cmp(undo->top_undo_no, trx->undo_no)
			        >= 0)) {

				trx->undo_no = ut_dulint_add(undo->top_undo_no,
									1);
			}
			
			undo = UT_LIST_GET_NEXT(undo_list, undo);
		}

		rseg = UT_LIST_GET_NEXT(rseg_list, rseg);
	}
}

/**********************************************************************
Assigns a rollback segment to a transaction in a round-robin fashion.
Skips the SYSTEM rollback segment if another is available. */
UNIV_INLINE
ulint
trx_assign_rseg(void)
/*=================*/
			/* out: assigned rollback segment id */
{
	trx_rseg_t*	rseg	= trx_sys->latest_rseg;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&kernel_mutex));
#endif /* UNIV_SYNC_DEBUG */
loop:
	/* Get next rseg in a round-robin fashion */

	rseg = UT_LIST_GET_NEXT(rseg_list, rseg);

	if (rseg == NULL) {
		rseg = UT_LIST_GET_FIRST(trx_sys->rseg_list);
	}

	/* If it is the SYSTEM rollback segment, and there exist others, skip
	it */

	if ((rseg->id == TRX_SYS_SYSTEM_RSEG_ID) 
			&& (UT_LIST_GET_LEN(trx_sys->rseg_list) > 1)) {
		goto loop;
	}			

	trx_sys->latest_rseg = rseg;

	return(rseg->id);
}

/********************************************************************
Starts a new transaction. */

ibool
trx_start_low(
/*==========*/
			/* out: TRUE */
	trx_t* 	trx,	/* in: transaction */
	ulint	rseg_id)/* in: rollback segment id; if ULINT_UNDEFINED
			is passed, the system chooses the rollback segment
			automatically in a round-robin fashion */
{
	trx_rseg_t*	rseg;

#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&kernel_mutex));
#endif /* UNIV_SYNC_DEBUG */
	ut_ad(trx->rseg == NULL);

	if (trx->type == TRX_PURGE) {
		trx->id = ut_dulint_zero;
		trx->conc_state = TRX_ACTIVE;
		trx->start_time = time(NULL);

		return(TRUE);
	}

	ut_ad(trx->conc_state != TRX_ACTIVE);
	
	if (rseg_id == ULINT_UNDEFINED) {

		rseg_id = trx_assign_rseg();
	}

	rseg = trx_sys_get_nth_rseg(trx_sys, rseg_id);

	trx->id = trx_sys_get_new_trx_id();

	/* The initial value for trx->no: ut_dulint_max is used in
	read_view_open_now: */

	trx->no = ut_dulint_max;

	trx->rseg = rseg;

	trx->conc_state = TRX_ACTIVE;
	trx->start_time = time(NULL);

	UT_LIST_ADD_FIRST(trx_list, trx_sys->trx_list, trx);

	return(TRUE);
}

/********************************************************************
Starts a new transaction. */

ibool
trx_start(
/*======*/
			/* out: TRUE */
	trx_t* 	trx,	/* in: transaction */
	ulint	rseg_id)/* in: rollback segment id; if ULINT_UNDEFINED
			is passed, the system chooses the rollback segment
			automatically in a round-robin fashion */
{
	ibool	ret;
	
	mutex_enter(&kernel_mutex);

	ret = trx_start_low(trx, rseg_id);

	mutex_exit(&kernel_mutex);

	return(ret);
}

/********************************************************************
Commits a transaction. */

void
trx_commit_off_kernel(
/*==================*/
	trx_t*	trx)	/* in: transaction */
{
	page_t*		update_hdr_page;
	dulint		lsn;
	trx_rseg_t*	rseg;
	trx_undo_t*	undo;
	ibool		must_flush_log	= FALSE;
	mtr_t		mtr;
	
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&kernel_mutex));
#endif /* UNIV_SYNC_DEBUG */

	trx->must_flush_log_later = FALSE;

	rseg = trx->rseg;
	
	if (trx->insert_undo != NULL || trx->update_undo != NULL) {

		mutex_exit(&kernel_mutex);

		mtr_start(&mtr);
		
		must_flush_log = TRUE;

		/* Change the undo log segment states from TRX_UNDO_ACTIVE
		to some other state: these modifications to the file data
		structure define the transaction as committed in the file
		based world, at the serialization point of the log sequence
		number lsn obtained below. */

		mutex_enter(&(rseg->mutex));
			
		if (trx->insert_undo != NULL) {
			trx_undo_set_state_at_finish(trx, trx->insert_undo,
									&mtr);
		}

		undo = trx->update_undo;

		if (undo) {
			mutex_enter(&kernel_mutex);
			trx->no = trx_sys_get_new_trx_no();
			
			mutex_exit(&kernel_mutex);

			/* It is not necessary to obtain trx->undo_mutex here
			because only a single OS thread is allowed to do the
			transaction commit for this transaction. */
					
			update_hdr_page = trx_undo_set_state_at_finish(trx,
								undo, &mtr);

			/* We have to do the cleanup for the update log while
			holding the rseg mutex because update log headers
			have to be put to the history list in the order of
			the trx number. */

			trx_undo_update_cleanup(trx, update_hdr_page, &mtr);
		}

		mutex_exit(&(rseg->mutex));

		/* Update the latest MySQL binlog name and offset info
		in trx sys header if MySQL binlogging is on or the database
		server is a MySQL replication slave */

		if (trx->mysql_log_file_name
				&& trx->mysql_log_file_name[0] != '\0') {
			trx_sys_update_mysql_binlog_offset(
					trx->mysql_log_file_name,
					trx->mysql_log_offset,
					TRX_SYS_MYSQL_LOG_INFO, &mtr);
			trx->mysql_log_file_name = NULL;
		}

		if (trx->mysql_master_log_file_name[0] != '\0') {
			/* This database server is a MySQL replication slave */ 
			trx_sys_update_mysql_binlog_offset(
				trx->mysql_master_log_file_name,
				trx->mysql_master_log_pos,
				TRX_SYS_MYSQL_MASTER_LOG_INFO, &mtr);
		}
				
		/* The following call commits the mini-transaction, making the
		whole transaction committed in the file-based world, at this
		log sequence number. The transaction becomes 'durable' when
		we write the log to disk, but in the logical sense the commit
		in the file-based data structures (undo logs etc.) happens
		here.

		NOTE that transaction numbers, which are assigned only to
		transactions with an update undo log, do not necessarily come
		in exactly the same order as commit lsn's, if the transactions
		have different rollback segments. To get exactly the same
		order we should hold the kernel mutex up to this point,
		adding to to the contention of the kernel mutex. However, if
		a transaction T2 is able to see modifications made by
		a transaction T1, T2 will always get a bigger transaction
		number and a bigger commit lsn than T1. */

		/*--------------*/
 		mtr_commit(&mtr);
 		/*--------------*/
 		lsn = mtr.end_lsn;

		mutex_enter(&kernel_mutex);
	}

	ut_ad(trx->conc_state == TRX_ACTIVE
					|| trx->conc_state == TRX_PREPARED);
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&kernel_mutex));
#endif /* UNIV_SYNC_DEBUG */
	
	/* The following assignment makes the transaction committed in memory
	and makes its changes to data visible to other transactions.
	NOTE that there is a small discrepancy from the strict formal
	visibility rules here: a human user of the database can see
	modifications made by another transaction T even before the necessary
	log segment has been flushed to the disk. If the database happens to
	crash before the flush, the user has seen modifications from T which
	will never be a committed transaction. However, any transaction T2
	which sees the modifications of the committing transaction T, and
	which also itself makes modifications to the database, will get an lsn
	larger than the committing transaction T. In the case where the log
	flush fails, and T never gets committed, also T2 will never get
	committed. */

	/*--------------------------------------*/
	trx->conc_state = TRX_COMMITTED_IN_MEMORY;
	/*--------------------------------------*/

	lock_release_off_kernel(trx);

	if (trx->global_read_view) {
		read_view_close(trx->global_read_view);
		mem_heap_empty(trx->global_read_view_heap);
		trx->global_read_view = NULL;
	}

	trx->read_view = NULL;

	if (must_flush_log) {

		mutex_exit(&kernel_mutex);
	
		if (trx->insert_undo != NULL) {

			trx_undo_insert_cleanup(trx);
		}

		/* NOTE that we could possibly make a group commit more
		efficient here: call os_thread_yield here to allow also other
		trxs to come to commit! */

		/*-------------------------------------*/

                /* Depending on the my.cnf options, we may now write the log
                buffer to the log files, making the transaction durable if
                the OS does not crash. We may also flush the log files to
                disk, making the transaction durable also at an OS crash or a
                power outage.

                The idea in InnoDB's group commit is that a group of
                transactions gather behind a trx doing a physical disk write
                to log files, and when that physical write has been completed,
                one of those transactions does a write which commits the whole
                group. Note that this group commit will only bring benefit if
                there are > 2 users in the database. Then at least 2 users can
                gather behind one doing the physical log write to disk.

                If we are calling trx_commit() under MySQL's binlog mutex, we
                will delay possible log write and flush to a separate function
                trx_commit_complete_for_mysql(), which is only called when the
                thread has released the binlog mutex. This is to make the
                group commit algorithm to work. Otherwise, the MySQL binlog
                mutex would serialize all commits and prevent a group of
                transactions from gathering. */

                if (trx->flush_log_later) {
                        /* Do nothing yet */
			trx->must_flush_log_later = TRUE;
                } else if (srv_flush_log_at_trx_commit == 0) {
                        /* Do nothing */
                } else if (srv_flush_log_at_trx_commit == 1) {
                        if (srv_unix_file_flush_method == SRV_UNIX_NOSYNC) {
                             	/* Write the log but do not flush it to disk */

                               	log_write_up_to(lsn, LOG_WAIT_ONE_GROUP,
									FALSE);
                        } else {
                               	/* Write the log to the log files AND flush
                               	them to disk */

                               	log_write_up_to(lsn, LOG_WAIT_ONE_GROUP, TRUE);
                        }
                } else if (srv_flush_log_at_trx_commit == 2) {

                        /* Write the log but do not flush it to disk */

                        log_write_up_to(lsn, LOG_WAIT_ONE_GROUP, FALSE);
                } else {
                        ut_error;
                }

		trx->commit_lsn = lsn;
		
		/*-------------------------------------*/
	
		mutex_enter(&kernel_mutex);
	}

	/* Free savepoints */
	trx_roll_savepoints_free(trx, NULL);

	trx->conc_state = TRX_NOT_STARTED;
	trx->rseg = NULL;
	trx->undo_no = ut_dulint_zero;
	trx->last_sql_stat_start.least_undo_no = ut_dulint_zero;

	ut_ad(UT_LIST_GET_LEN(trx->wait_thrs) == 0);
	ut_ad(UT_LIST_GET_LEN(trx->trx_locks) == 0);

	UT_LIST_REMOVE(trx_list, trx_sys->trx_list, trx);
}

/********************************************************************
Cleans up a transaction at database startup. The cleanup is needed if
the transaction already got to the middle of a commit when the database
crashed, andf we cannot roll it back. */

void
trx_cleanup_at_db_startup(
/*======================*/
	trx_t*	trx)	/* in: transaction */
{
	if (trx->insert_undo != NULL) {

		trx_undo_insert_cleanup(trx);
	}

	trx->conc_state = TRX_NOT_STARTED;
	trx->rseg = NULL;
	trx->undo_no = ut_dulint_zero;
	trx->last_sql_stat_start.least_undo_no = ut_dulint_zero;

	UT_LIST_REMOVE(trx_list, trx_sys->trx_list, trx);
}

/************************************************************************
Assigns a read view for a consistent read query. All the consistent reads
within the same transaction will get the same read view, which is created
when this function is first called for a new started transaction. */

read_view_t*
trx_assign_read_view(
/*=================*/
			/* out: consistent read view */
	trx_t*	trx)	/* in: active transaction */
{
	ut_ad(trx->conc_state == TRX_ACTIVE);

	if (trx->read_view) {
		return(trx->read_view);
	}
	
	mutex_enter(&kernel_mutex);

	if (!trx->read_view) {
		trx->read_view = read_view_open_now(trx,
					trx->global_read_view_heap);
		trx->global_read_view = trx->read_view;
	}

	mutex_exit(&kernel_mutex);
	
	return(trx->read_view);
}

/********************************************************************
Commits a transaction. NOTE that the kernel mutex is temporarily released. */
static
void
trx_handle_commit_sig_off_kernel(
/*=============================*/
	trx_t*		trx,		/* in: transaction */
	que_thr_t**	next_thr)	/* in/out: next query thread to run;
					if the value which is passed in is
					a pointer to a NULL pointer, then the
					calling function can start running
					a new query thread */
{
	trx_sig_t*	sig;
	trx_sig_t*	next_sig;
	
#ifdef UNIV_SYNC_DEBUG
	ut_ad(mutex_own(&kernel_mutex));
#endif /* UNIV_SYNC_DEBUG */

	trx->que_state = TRX_QUE_COMMITTING;